<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Exclusive-Analysis on Korea Invest Insights</title><link>https://koreainvestinsights.com/series/exclusive-analysis/</link><description>Recent content in Exclusive-Analysis on Korea Invest Insights</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>koreainvestinsights.com · @korea_invest_insights</copyright><lastBuildDate>Sun, 08 Nov 2026 19:53:15 +0900</lastBuildDate><atom:link href="https://koreainvestinsights.com/series/exclusive-analysis/feed.xml" rel="self" type="application/rss+xml"/><item><title>AI is raising the price of every ad: Meta's 14% and AppLovin's 55% come from the same law, and agents decide the next round</title><link>https://koreainvestinsights.com/post/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/</link><pubDate>Sun, 08 Nov 2026 19:45:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/</guid><description>&lt;img src="https://koreainvestinsights.com/" alt="Featured image of post AI is raising the price of every ad: Meta's 14% and AppLovin's 55% come from the same law, and agents decide the next round" /&gt;&lt;p&gt;In the third quarter, Meta served 10% more ad impressions than a year earlier, and the average price per ad rose &lt;strong&gt;14%&lt;/strong&gt;. This is the first time since 3Q25 that price growth beat impression growth. In the same quarter AppLovin&amp;rsquo;s installs grew 2% while its net revenue per install grew &lt;strong&gt;55%&lt;/strong&gt;. Unity&amp;rsquo;s ad model, Vector, grew 40% from the prior quarter.&lt;/p&gt;
&lt;p&gt;The companies that grew fastest did not sell many more ads. They sold each ad for more. Advertisers pay more because the ads fit better. In an October engineering post, Meta said its ads foundation model GEM raised conversions by 18.9% and click-through rate by 12% across its ad surfaces.&lt;/p&gt;
&lt;p&gt;Our &lt;a class="link" href="https://koreainvestinsights.com/post/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/" &gt;AppLovin and Unity earnings post&lt;/a&gt; traced both beats to a step up in their prediction models. This follow-up uses Meta&amp;rsquo;s primary sources to explain why these models became more accurate now, and judges what that means for the ad industry.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our view: ad prediction models got more accurate because model size, the ability to reuse what a model learns, and the decisions advertisers hand over all grew at once. Before 2024, adding compute to recommendation models did little. Now accuracy rises predictably as the models grow. In advertising, AI therefore raises the value of each ad more than the number of ads, and money gathers in a few engines that run the auction and see the conversion outcome. Compute can be bought; outcome data cannot. The size of outcome data decides the winners in 2027 and 2028. The variable that could overturn this view is agents. If the purchase decision moves from feeds and search to personal agents, outcome data changes owner.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="a-short-glossary"&gt;A short glossary
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Term&lt;/th&gt;
 &lt;th&gt;Plain meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Prediction model (ad engine)&lt;/td&gt;
 &lt;td&gt;The AI that estimates, for each ad slot, how likely this person is to act on this ad, and sets the bid. Meta&amp;rsquo;s GEM, AppLovin&amp;rsquo;s Axon and Unity&amp;rsquo;s Vector are examples.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Conversion&lt;/td&gt;
 &lt;td&gt;The action the advertiser wants after the ad: an install, a purchase, a sign-up.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Scaling law&lt;/td&gt;
 &lt;td&gt;The rule that performance improves at a predictable rate as compute, model size and data grow. It was first shown for large language models (LLMs).&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundation model&lt;/td&gt;
 &lt;td&gt;A large central model that teaches many smaller task models. GEM is Meta&amp;rsquo;s foundation model for ads.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Knowledge transfer (distillation)&lt;/td&gt;
 &lt;td&gt;Moving what a large model learned into small, fast, cheap models.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Outcome data&lt;/td&gt;
 &lt;td&gt;The record of whether an install or purchase actually happened after the ad. The model learns again only when this record comes back.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Annual run-rate&lt;/td&gt;
 &lt;td&gt;Recent quarterly or monthly revenue scaled up to a full year.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Agentic commerce&lt;/td&gt;
 &lt;td&gt;AI agents that find, compare and pay for products on a person&amp;rsquo;s behalf.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="growth-came-from-the-value-of-each-ad-not-the-number-of-ads"&gt;Growth came from the value of each ad, not the number of ads
&lt;/h2&gt;&lt;p&gt;Ad revenue is the number of ads times the value of each ad. Meta discloses the growth of both every quarter. Set the quarters since 3Q25 side by side and the weight of growth moves toward price.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Meta&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3Q25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;4Q25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3Q26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Ad revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$50.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$58.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$55.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$59.4B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$63.1B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ad revenue growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+33%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ad impression growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+19%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price per ad growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Share of growth from price (our calculation)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;42%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;26%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;39%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;46%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;58%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;img alt="Growth came from the value of each ad, not the number of ads. Meta’s ad impression growth from 3Q25 was 14, 18, 19, 14 and 10%, and price per ad growth was 10, 6, 12, 12 and 14%. The share of growth from price was 42, 26, 39, 46 and 58%. AppLovin’s installs grew 2% and net revenue per install 55% in 3Q26, so 96% of its growth came from price." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/growth-from-price.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Meta&amp;rsquo;s impression growth fell from 19% in 1Q26 to 10% in 3Q26. Its ad revenue growth barely slowed, at 27% in 2Q26 and 26% in 3Q26. Price filled the gap. AppLovin is the extreme case: almost all of its revenue growth (96% by our calculation) came from the value of each install. AppLovin&amp;rsquo;s two figures multiply to about 3 points more than its revenue growth, so we use this value only to read direction.&lt;/p&gt;
&lt;p&gt;Meta named both better ad performance and better macro conditions than a year ago as reasons for the higher price, and did not split the two. The 1Q26 and 2Q26 prices also include a currency tailwind. So the whole price share cannot be read as AI&amp;rsquo;s share. Still, Meta&amp;rsquo;s CFO said on the 3Q25 call that value-weighted conversions were growing faster than impressions. Advertisers are getting more results from the same impressions.&lt;/p&gt;
&lt;p&gt;Google Search looks different. In 3Q26 Google Search revenue grew 12.0%, paid clicks grew 8% and cost per click grew 4%. As our &lt;a class="link" href="https://koreainvestinsights.com/post/meta-ad-edge-google-amazon-share-shift-2026-10-30/" &gt;October 30 post&lt;/a&gt; showed, Meta took 57% of the combined ad revenue increase of Meta, Google and Amazon in the quarter.&lt;/p&gt;
&lt;h2 id="since-2024-recommendation-models-also-get-more-accurate-as-they-grow"&gt;Since 2024, recommendation models also get more accurate as they grow
&lt;/h2&gt;&lt;p&gt;To see why accuracy rose now, start before 2024. In a 2024 paper (HSTU), Meta researchers wrote that most industry recommendation models, trained on thousands of features, failed to scale with compute. The models were also split finely by surface and objective. Since 2023 alone, Meta has folded about 100 ad ranking models into larger ones.&lt;/p&gt;
&lt;p&gt;The paper recast recommendation as a generative task, predicting the next action the way a language model predicts the next word. A model with 1.5 trillion parameters improved metrics by 12.4% in online tests. Model quality rose as a power law of training compute across three orders of magnitude. Another Meta paper the same year (Wukong) found the same law across two orders of magnitude of model complexity.&lt;/p&gt;
&lt;p&gt;GEM brought this finding into ads. In November 2025 Meta introduced GEM as the industry&amp;rsquo;s largest foundation model for recommendation. It trains on thousands of GPUs and, Meta said, is 4x more efficient at driving ad performance than its earlier ranking models for the same data and compute. In 2Q25 it lifted ad conversions by 5% on Instagram and 3% on Facebook Feed.&lt;/p&gt;
&lt;p&gt;Since then Meta has changed its ad models one step every quarter.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Meta changed its ad models one step every quarter. In 3Q25 Lattice extended to app ads and Andromeda merged stages, lifting app ad conversions about 3% and Facebook ads quality 14%. In 4Q25 GPUs for GEM training doubled and Facebook ad clicks rose 3.5%. In 1Q26 adaptive ranking lifted conversion rates on major surfaces 1.6%. In 2Q26 the Generative Recommender arrived and Facebook conversions rose 15.7%. In 3Q26 a wider GEM lifted app ad conversions 9.6%." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/meta-model-ladder.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;In 3Q25 Meta extended its unified Lattice architecture to app ads and lifted conversions by about 3%. In 4Q25 it doubled the GPUs used to train GEM. In 2Q26 it introduced the Generative Recommender, which uses LLMs to reason over ad content and user preferences, and called it a paradigm shift in how its ads system works. Each effect applies to a specific surface and objective, so they cannot be added. The direction is clear: a few percent of conversion gains stacked up every quarter.&lt;/p&gt;
&lt;p&gt;Meta&amp;rsquo;s GEM follow-up post on October 13 puts numbers on the mechanism. Over the past 12 months Meta raised training compute 4x, and model complexity is now 2.5x that of earlier GEM versions. The new architecture delivered twice the performance of the previous one at the same compute. Meta also found that prediction quality follows a power law in the length of a user&amp;rsquo;s behaviour history.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Recommendation models now improve as they grow. Over the past 12 months Meta raised GEM’s training compute 4x, and model complexity is 2.5x that of earlier GEM versions. Bigger models make longer behaviour histories useful, and accuracy follows a power law in history length. Conversions rose 18.9% and click-through rate 12% across surfaces, and the extra ad revenue pays for the next round of compute." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/scaling-loop.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;The two laws reinforce each other. Bigger models gain more from longer histories, and longer histories justify bigger models. This is why Meta said on its 3Q25 call that its ads business runs in a compute-starved state. On the 3Q26 call, Zuckerberg said Meta has not yet found the limit of how effectively it can apply compute to its core business.&lt;/p&gt;
&lt;p&gt;For investors, the change reads like this. Ad accuracy has become something you can buy with compute. Gains arrive in steps each time a new model ships, and the timing of the next step moves quarterly numbers. AppLovin&amp;rsquo;s weak 2Q26 and strong 3Q26 follow the same pattern.&lt;/p&gt;
&lt;h2 id="the-models-now-fit-small-advertisers-and-new-posts-too"&gt;The models now fit small advertisers and new posts too
&lt;/h2&gt;&lt;p&gt;The second effect of large models is generalization. Meta says GEM transfers what it learns from click prediction, where data is rich, to conversion prediction, where data is sparse. Clicks pile up every day; purchases are rare. To predict rare events, a model must borrow from common ones.&lt;/p&gt;
&lt;p&gt;Other companies show the same effect. AppLovin&amp;rsquo;s CEO said the next-generation model improved most in data-sparse segments: new and small advertisers and new categories. AppLovin&amp;rsquo;s install growth moved from -18% in 1Q26 and -2% in 2Q26 to +2% in 3Q26. Unity&amp;rsquo;s advertiser accounts grew 27% from the prior quarter.&lt;/p&gt;
&lt;p&gt;Meta&amp;rsquo;s generative retrieval works in the same direction. In organic recommendations it labels posts with semantic representations, so the system can recommend new posts before they gather engagement. In ads, Meta applied its Generative Recommender, which uses LLMs to reason over ad content and user preferences, to the retrieval stage of Facebook Feed ads, lifting conversions by nearly 2% in targeted segments in 3Q26.&lt;/p&gt;
&lt;p&gt;This widens the market. Small advertisers and new products that lacked the data to run efficient ads can now enter the auction. Meta estimates that at least 500 million small businesses use its apps to reach customers. Ad market growth no longer depends only on large advertisers raising budgets.&lt;/p&gt;
&lt;h2 id="the-big-model-learns-small-models-serve-each-ad-in-real-time"&gt;The big model learns; small models serve each ad in real time
&lt;/h2&gt;&lt;p&gt;Large models bring a cost problem. Every ad request must be ranked in a very short time. Meta has said on several calls that it does not use large models such as GEM for live inference because they are too expensive.&lt;/p&gt;
&lt;p&gt;Instead Meta moves what GEM learns into hundreds of small serving models. Beyond distillation, where small models learn from the big model&amp;rsquo;s answers, Meta now passes the big model&amp;rsquo;s internal representations and doubled the transfer effect. It also built a system that precomputes the big model&amp;rsquo;s work just before a request arrives. In 1Q26 Meta extended a ranking model that assigns heavy compute only to requests likely to convert to offsite conversion ads, lifting conversion rates across major Facebook and Instagram surfaces by 1.6%.&lt;/p&gt;
&lt;p&gt;&lt;img alt="The big model learns; small models serve each ad in real time. The central foundation model GEM trains on thousands of GPUs but is not served live. Distillation, structured representations and precomputation move its knowledge into hundreds of serving models. Representations learned on data-rich clicks are used for data-sparse conversions, and AppLovin’s new model also gained most with new and small advertisers." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/teacher-student.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;This design keeps serving costs small relative to revenue. In 3Q26 AppLovin&amp;rsquo;s cost of revenue rose only $0.06 to $0.09 for each extra dollar of revenue, and 87% of its revenue increase became adjusted EBITDA. In 4Q25 Meta made its ad retrieval engine Andromeda run on Nvidia, AMD and its own MTIA chips and nearly tripled its compute efficiency. In 3Q26 Meta began moving content-understanding models onto its own Muse models, which cut per-unit processing cost by more than 80% in offline tests.&lt;/p&gt;
&lt;p&gt;This is why ad AI is the first AI investment to pay. Training is expensive but is spread over billions of people&amp;rsquo;s ad requests, and small models serve cheaply.&lt;/p&gt;
&lt;h2 id="advertisers-are-handing-targeting-and-creative-to-the-engine"&gt;Advertisers are handing targeting and creative to the engine
&lt;/h2&gt;&lt;p&gt;The last ingredient of accuracy is the advertiser&amp;rsquo;s choice. When advertisers decide whom to reach, where, and with which creative, the model has little room to choose. When advertisers hand these decisions to the engine, the model picks from many more combinations.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Advertisers are handing targeting and creative to the engine. Advantage+ automated campaigns grew from a $60 billion annual run-rate in 3Q25 to over $75 billion in 2Q26. Advertisers using generative AI creative tools grew from over 8 million in 1Q26 to about 10 million in 3Q26. Creator partnership ads grew from a $10 billion run-rate in 1Q26 to over $15 billion in 3Q26." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/automation-adoption.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;The annual run-rate of Meta&amp;rsquo;s automated Advantage+ campaigns rose from $60 billion in 3Q25 to over $75 billion in 2Q26. Nearly 10 million advertisers used at least one generative AI creative tool in 3Q26. More creative means a better fit for each person. Meta estimates that advertisers who adopted its business AI assistant&amp;rsquo;s advice got about 5% more conversions at the same cost per conversion and raised ad spend by 9% compared with similar non-adopters.&lt;/p&gt;
&lt;p&gt;Measurement moved to the engine too. Meta&amp;rsquo;s incremental attribution, described on its 4Q25 call, targets the conversions an ad actually adds. Meta said the model drove 24% more incremental conversions than its standard attribution model.&lt;/p&gt;
&lt;p&gt;Google is moving the same way. When it added exploration to Smart Bidding in May 2025, Google said campaigns using it saw 19% more conversions on average, based on its internal data.&lt;/p&gt;
&lt;h2 id="cash-from-ads-goes-into-bigger-ai"&gt;Cash from ads goes into bigger AI
&lt;/h2&gt;&lt;p&gt;The money ad AI makes is large and visible. On its 3Q26 call Meta said its AI investments already deliver substantial incremental ad revenue. But the cash goes somewhere much bigger than ad models.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Cash from ads went into bigger AI infrastructure. Meta’s quarterly ad revenue grew from $50.1 billion in 3Q25 to $63.1 billion in 3Q26. Over the same period capital expenditure rose from $19.4 billion to $34.9 billion, and free cash flow fell from $10.6 billion to minus $2.1 billion." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/meta-capex-fcf.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Meta&amp;rsquo;s quarterly capital expenditure rose from $19.4 billion to $34.9 billion in a year, and free cash flow went from $10.6 billion to minus $2.1 billion. Its 2026 capex plan is $135 billion to $145 billion. On the 2Q26 call Meta said a substantial amount of its compute goes to training models to be a leading lab. In late October it also raised outside capital by issuing common stock and mandatory convertible preferred stock.&lt;/p&gt;
&lt;p&gt;Investors should separate two returns. The return on ad AI is already visible in the numbers. The return on frontier models and the personal agent Muse is not yet visible. The ad business funds the second bet.&lt;/p&gt;
&lt;p&gt;The two bets also meet. On the 3Q26 call Zuckerberg said Meta built an evaluation of how well its newest model, Muse Wonder, can autonomously improve the recommendation and ads systems. He expects the rate of improvement to rise as new models arrive. If AI starts improving ad AI, the cycle between model steps could shorten. For now this is the company&amp;rsquo;s expectation, not a measured result.&lt;/p&gt;
&lt;p&gt;We also share new analyses on Telegram as they are published. You are welcome to join us at &lt;a class="link" href="https://t.me/korea_invest_insights" target="_blank" rel="noopener"
 &gt;@korea_invest_insights&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="ad-money-is-gathering-in-engines-that-own-outcome-data"&gt;Ad money is gathering in engines that own outcome data
&lt;/h2&gt;&lt;p&gt;In its June forecast, WPP Media expected global ad revenue to grow 8.9% in 2026 to about $1.3 trillion, excluding US political ads, up from the 7.1% it forecast in December. The same report said Alphabet, Meta and Amazon control 57.6% of the market outside China, and no traditional media owner is in the global top ten.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Growth in ad-related revenue in 3Q26: AppLovin revenue 55%, Reddit ad revenue 54%, Meta ad revenue 26%, Pinterest revenue 12%, Amazon ad revenue 11%, Google ad revenue 10.6%, Omnicom organic 7.2%, Publicis organic 5%, The Trade Desk 2Q revenue 3%, WPP net revenue minus 2.1%. Agencies and open-web buying platforms grew slowest." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/growth-compare.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Side by side, 3Q26 growth rates differ widely inside the ad market. Platforms that run the auction and see outcomes sit at the top. Agencies and open-web buying platforms, which buy space on behalf of advertisers, sit at the bottom. Google and Amazon trail Meta for their own reasons: Google&amp;rsquo;s search clicks grew less, and Amazon&amp;rsquo;s Prime Day moved into 2Q.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Reddit&amp;rsquo;s ad revenue grew 54%, and the company named both pricing and impressions as drivers. Revenue per user grew 36%.&lt;/li&gt;
&lt;li&gt;The Trade Desk grew revenue 3% in 2Q26 and guided 3Q revenue to at least $650 million, below its 2Q level. It reports 3Q on November 9 (US time).&lt;/li&gt;
&lt;li&gt;Agencies are not moving in one direction. Omnicom grew 7.2% organically and Publicis 5%. WPP&amp;rsquo;s net revenue fell 2.1%, with its creative business down 5.1%.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;We think creative production and manual targeting lose value first inside agency work, because engines now make the creative and pick the audience. Agencies that shifted to media buying and data are still growing. It is too early to read AI&amp;rsquo;s impact from agency results alone; Omnicom&amp;rsquo;s 3Q also includes World Cup and US midterm spending (per secondary reports).&lt;/p&gt;
&lt;p&gt;Search faces structural pressure. Google said search queries grew at the fastest pace in five years while paid click growth slowed from 13% in 2Q to 8% in 3Q, and explained that AI Mode monetized at a lower rate. WPP Media expects generative search ads to grow from $5.1 billion in 2026 to over $100 billion by 2030. The place of search ads is moving inside AI answers.&lt;/p&gt;
&lt;p&gt;AI chatbots have become ad space as well. OpenAI began a ChatGPT ads pilot in the US on February 9 and opened a self-serve ads manager to US advertisers in May. It said tests would start in the UK, Brazil, Japan, South Korea and Mexico. Ad tracker AdClarity estimated, as of early May, average monthly spend since the pilot at about $109 million. OpenAI has not published this number.&lt;/p&gt;
&lt;p&gt;AI is moving ad money in these directions:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;From intermediaries to engines: as engines take over targeting, creative and measurement, agencies and open-web intermediaries lose share.&lt;/li&gt;
&lt;li&gt;From large advertisers to small ones: as ads work even with little data, more advertisers enter the auction.&lt;/li&gt;
&lt;li&gt;From link clicks to AI answers and agents: search clicks move into AI answers, and new ad space appears inside them.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-variable-that-could-flip-the-view-is-how-fast-purchase-decisions-move-to-agents"&gt;The variable that could flip the view is how fast purchase decisions move to agents
&lt;/h2&gt;&lt;p&gt;An ad engine&amp;rsquo;s power comes from outcome data. The model learns again only when it knows whether the viewer actually bought. If agents search and pay on a person&amp;rsquo;s behalf, that data lands first with whoever owns the agent.&lt;/p&gt;
&lt;p&gt;&lt;img alt="If agents make the purchase decision, outcome data changes owner. Today a person sees an ad in a feed or search and buys on the advertiser’s site; money is paid per impression, click or conversion, and outcome data returns to the ad platform. Through agents, a personal agent connects to a business agent or authorised API and pays; money becomes transaction fees, subscriptions and ads inside agents, and outcome data goes to whoever owns the agent." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/agent-shift.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Meta is betting on both sides. More than 10 million people in the US and Canada set up its personal agent Muse within a month and a half of its September launch. Over 25% of weekly users shop with it, with an average purchase above $200. Zuckerberg said Meta will work on Muse monetization sooner than usual and described a model in which Meta takes a percent of the value it helps create. A transaction-fee model could sit beside the ad-fee model.&lt;/p&gt;
&lt;p&gt;CFO Susan Li expects the market to move away from agents navigating every website like a human. She expects permissioned, structured interactions instead: authenticated APIs, approved partnerships and direct links between personal agents and business agents. Then whoever holds the standards and payment rails for those links will own the outcome data.&lt;/p&gt;
&lt;p&gt;We think this variable is small for 2027 ad numbers and grows after 2028. Muse users are still few next to Meta&amp;rsquo;s 3.6 billion daily users. ChatGPT ads are still adding countries. If the share of purchases made through agents is disclosed and transaction fees start to show up as revenue, the timeline moves forward.&lt;/p&gt;
&lt;h2 id="set-side-by-side-outcome-data-is-the-dominant-variable"&gt;Set side by side, outcome data is the dominant variable
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Variable&lt;/th&gt;
 &lt;th&gt;Evidence now&lt;/th&gt;
 &lt;th&gt;Our view&lt;/th&gt;
 &lt;th&gt;Signal that changes the view&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Size of outcome data&lt;/td&gt;
 &lt;td&gt;Meta took 57% of the three companies&amp;rsquo; 3Q ad revenue increase; top three hold 57.6% outside China&lt;/td&gt;
 &lt;td&gt;Dominant. Money keeps gathering in engines that run the auction and see outcomes.&lt;/td&gt;
 &lt;td&gt;Advertisers pull budgets after a gap appears between platform-reported conversions and real incremental sales&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cadence of model steps&lt;/td&gt;
 &lt;td&gt;Meta changed models every quarter; GEM training compute 4x in 12 months; AppLovin&amp;rsquo;s 2Q/3Q step&lt;/td&gt;
 &lt;td&gt;Steps continue in 2027, and quarterly numbers move in steps&lt;/td&gt;
 &lt;td&gt;Meta&amp;rsquo;s price growth falls below impression growth for two quarters in a row&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Advertiser base&lt;/td&gt;
 &lt;td&gt;About 10 million advertisers on generative creative tools; Unity advertiser accounts +27%; AppLovin installs turned positive&lt;/td&gt;
 &lt;td&gt;Small advertisers support market growth&lt;/td&gt;
 &lt;td&gt;Advertiser growth stalls and growth relies only on price from existing advertisers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shift of ad space&lt;/td&gt;
 &lt;td&gt;Google paid clicks +8%; ChatGPT ads adding countries; Muse shopping&lt;/td&gt;
 &lt;td&gt;Search pressure is gradual; agents are a post-2028 variable&lt;/td&gt;
 &lt;td&gt;Share of purchases through agents is disclosed, and Meta or OpenAI books transaction fees&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Economy&lt;/td&gt;
 &lt;td&gt;Meta includes better macro among reasons for price gains&lt;/td&gt;
 &lt;td&gt;Price-led growth reacts to the economy first&lt;/td&gt;
 &lt;td&gt;Price growth turns down first as the economy slows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Outcome data is the dominant variable. Compute can be bought: AppLovin&amp;rsquo;s cost of revenue rose by less than $0.10 for each extra dollar of revenue. Auction and purchase records cannot be bought. The cadence of model steps sets how much quarterly numbers swing, and the economy sets the level of prices. Agents are the variable that could change who owns the dominant variable itself.&lt;/p&gt;
&lt;h2 id="links-for-korean-investors"&gt;Links for Korean investors
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Naver is a Korean engine that holds outcome data from both search and shopping (Smart Store). Its 3Q results, expected on November 12, will show the effect of its AI ad products and how AI Briefing affects search clicks.&lt;/li&gt;
&lt;li&gt;Kakao&amp;rsquo;s Talk Biz revenue grew 13% in 3Q, and KakaoTalk display ads within it grew 35% (per Korean press reports).&lt;/li&gt;
&lt;li&gt;If ChatGPT ads begin testing in Korea, Korean search and display budgets gain a new competitor.&lt;/li&gt;
&lt;li&gt;For mobile game publishers such as Krafton and Netmarble, user acquisition becomes more efficient, but the price of each install rises too. Spending grows while the return target stays the same.&lt;/li&gt;
&lt;li&gt;For AI chips, ad models are a steady source of compute demand. GEM&amp;rsquo;s training compute rose 4x in 12 months. But a substantial share of Meta&amp;rsquo;s compute goes to frontier model training, so ad models alone do not explain Meta&amp;rsquo;s capex.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="what-to-check-next"&gt;What to check next
&lt;/h2&gt;&lt;p&gt;&lt;img alt="What to check next: November 9 (US) The Trade Desk 3Q, November 12 Naver 3Q, late January Meta 4Q and 2027 plan, early February Alphabet and Amazon 4Q, February AppLovin and Unity 4Q. Dates not yet confirmed are marked TBC." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/ad-ai-scaling-law-price-per-ad-meta-applovin-agents-2026-11-08/next-checks.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;In Meta&amp;rsquo;s 4Q results, check whether price growth keeps beating impression growth, whether new model gains keep coming, and how large 2027 capex will be. For Google, the test is whether paid click growth catches up with query growth. In The Trade Desk and agency results, check whether intermediaries keep slowing.&lt;/p&gt;
&lt;h2 id="data-limits"&gt;Data limits
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Meta&amp;rsquo;s per-model effects apply to specific surfaces, objectives and segments, so they cannot be added. GEM&amp;rsquo;s 18.9% conversion and 12% CTR gains are company figures without a stated baseline or period.&lt;/li&gt;
&lt;li&gt;The price share is our calculation, splitting impression and price growth with logarithms. Meta&amp;rsquo;s price includes macro and currency effects, so AI&amp;rsquo;s share cannot be isolated.&lt;/li&gt;
&lt;li&gt;Meta figures come from official transcripts, and GEM technical details from Meta&amp;rsquo;s engineering blog and arXiv papers. Google and Amazon figures follow the primary sources of our &lt;a class="link" href="https://koreainvestinsights.com/post/meta-ad-edge-google-amazon-share-shift-2026-10-30/" &gt;October 30 post&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;ChatGPT ad spend is a third-party (AdClarity) estimate; OpenAI has not disclosed ad revenue.&lt;/li&gt;
&lt;li&gt;The WPP Media figure is a forecast. One outlet reported 4.4%; we used the 8.9% in WPP Media&amp;rsquo;s own release.&lt;/li&gt;
&lt;li&gt;Publicis and Kakao figures come from secondary reports. Naver and The Trade Desk have not yet reported 3Q.&lt;/li&gt;
&lt;li&gt;AppLovin&amp;rsquo;s comment on data-sparse segments and Unity&amp;rsquo;s advertiser account growth are call statements, not filed figures.&lt;/li&gt;
&lt;li&gt;We did not use broker estimates or price targets.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Meta official call transcripts: &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/content_files/META-Q3-2026-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;3Q26&lt;/a&gt;, &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/doc_financials/2026/q2/META-Q2-2026-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;2Q26&lt;/a&gt;, &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/doc_financials/2026/q1/META-Q1-2026-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;1Q26&lt;/a&gt;, &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/doc_financials/2025/q4/META-Q4-2025-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;4Q25&lt;/a&gt;, &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/doc_financials/2025/q3/META-Q3-2025-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;3Q25&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Meta 3Q26 filings: &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1326801/000162828026066700/meta-09302026xexhibit991.htm" target="_blank" rel="noopener"
 &gt;press release (SEC)&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1326801/000119312526436410/d94886d424b5.htm" target="_blank" rel="noopener"
 &gt;common stock prospectus (424B5)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Meta Engineering: &lt;a class="link" href="https://engineering.fb.com/2025/11/10/ml-applications/metas-generative-ads-model-gem-the-central-brain-accelerating-ads-recommendation-ai-innovation/" target="_blank" rel="noopener"
 &gt;GEM introduction (2025-11-10)&lt;/a&gt;, &lt;a class="link" href="https://engineering.fb.com/2026/10/13/ml-applications/generative-ads-model-gem-scaling-frontier-ads-ai/" target="_blank" rel="noopener"
 &gt;GEM follow-up (2026-10-13)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Papers: &lt;a class="link" href="https://arxiv.org/abs/2402.17152" target="_blank" rel="noopener"
 &gt;HSTU, arXiv 2402.17152&lt;/a&gt;, &lt;a class="link" href="https://arxiv.org/abs/2403.02545" target="_blank" rel="noopener"
 &gt;Wukong, arXiv 2403.02545&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;AppLovin and Unity: &lt;a class="link" href="https://d18rn0p25nwr6d.cloudfront.net/CIK-0001751008/5d38e9cf-1f99-43b1-adcc-36979436fce4.pdf" target="_blank" rel="noopener"
 &gt;AppLovin 3Q 10-Q&lt;/a&gt;, &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/11/05/applovin-app-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;AppLovin 3Q call transcript (secondary)&lt;/a&gt;, &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/11/05/unity-software-u-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;Unity 3Q call transcript (secondary)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Google: &lt;a class="link" href="https://blog.google/products/ads-commerce/smart-bidding-exploration-ai/" target="_blank" rel="noopener"
 &gt;Smart Bidding exploration (2025-05-21)&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000080/googexhibit991q32026.htm" target="_blank" rel="noopener"
 &gt;Alphabet 3Q release (SEC)&lt;/a&gt;, &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/10/29/alphabet-googl-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;Alphabet 3Q call transcript (secondary)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Other platforms: &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001713445/000171344526000114/exhibit992q326.htm" target="_blank" rel="noopener"
 &gt;Reddit 3Q shareholder letter (SEC)&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001506293/000150629326000151/q3-26xpressrelease.htm" target="_blank" rel="noopener"
 &gt;Pinterest 3Q release (SEC)&lt;/a&gt;, &lt;a class="link" href="https://www.thetradedesk.com/press-room/the-trade-desk-reports-second-quarter-2026-financial-results" target="_blank" rel="noopener"
 &gt;The Trade Desk 2Q results&lt;/a&gt;, &lt;a class="link" href="https://www.thetradedesk.com/press-room/the-trade-desk-announces-date-of-third-quarter-2026-financial-results-and-conference-call" target="_blank" rel="noopener"
 &gt;The Trade Desk 3Q date&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Agencies: &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0000029989/000002998926000024/a2026q3earningsrelease.htm" target="_blank" rel="noopener"
 &gt;Omnicom 3Q (SEC)&lt;/a&gt;, &lt;a class="link" href="https://www.exchange4media.com/advertising-news/omnicom-q3-organic-growth-hits-72-as-adjusted-eps-climbs-22-159420.html" target="_blank" rel="noopener"
 &gt;Omnicom 3Q coverage (exchange4media, secondary)&lt;/a&gt;, &lt;a class="link" href="https://www.directorstalkinterviews.com/wp-content/uploads/2026/10/WPP-Third-Quarter-Trading-Update-2026.pdf" target="_blank" rel="noopener"
 &gt;WPP 3Q trading update copy&lt;/a&gt;, &lt;a class="link" href="https://www.adweek.com/agencies/publicis-says-three-horse-holdco-race-is-good-for-business-as-q3-growth-hits-5/" target="_blank" rel="noopener"
 &gt;Publicis 3Q (Adweek, secondary)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Industry forecast: &lt;a class="link" href="https://www.wppmedia.com/news/report-this-year-next-year-midyear-2026" target="_blank" rel="noopener"
 &gt;WPP Media This Year Next Year 2026 midyear&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;ChatGPT ads: &lt;a class="link" href="https://digiday.com/media-buying/expand-thoughtfully-openai-offers-chatgpt-ads-to-new-markets-including-the-u-k-brazil-and-japan/" target="_blank" rel="noopener"
 &gt;Digiday (secondary)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Korea: &lt;a class="link" href="https://news.bizwatch.co.kr/article/mobile/2026/11/05/0003" target="_blank" rel="noopener"
 &gt;Kakao 3Q (Bizwatch, secondary)&lt;/a&gt;, &lt;a class="link" href="https://www.fnnews.com/news/202611021502551535" target="_blank" rel="noopener"
 &gt;Naver 3Q date (Financial News, secondary)&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;As AI redistributes ad money, who loses share first? Leave A, B or C and one data point in the comments.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;A. Ad agencies and open-web intermediaries: engines take over targeting and creative, and fees shrink.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;B. Google Search: as AI answers grow, paid clicks grow less and search ads lose share.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;C. Meta&amp;rsquo;s feed ads: as purchase decisions move to agents, outcome data changes owner.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;This is for research and information only and is not investment advice. Views and decision thresholds are analytical assumptions, not trading instructions or personal portfolio decisions. Figures have reference dates, and some come from secondary sources. We used Anthropic&amp;rsquo;s Claude for research and editing. Alphabet and Amazon are investors in Anthropic, and Anthropic competes with Meta, OpenAI and Google in AI models, so Anthropic has interests related to several companies in this post. We did not judge the size of that effect.&lt;/p&gt;</description></item><item><title>AppLovin's Revenue Rose 55% on 2% More Installs: Both AppLovin and Unity Beat Because One Engine Jumped a Step</title><link>https://koreainvestinsights.com/post/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/</link><pubDate>Sun, 08 Nov 2026 18:50:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/</guid><description>&lt;img src="https://koreainvestinsights.com/" alt="Featured image of post AppLovin's Revenue Rose 55% on 2% More Installs: Both AppLovin and Unity Beat Because One Engine Jumped a Step" /&gt;&lt;p&gt;On November 5 the Nasdaq Composite fell 5.5% in one day. AppLovin rose 31.0% and Unity rose 14.9% the same day. Over two days AppLovin gained &lt;strong&gt;44.0%&lt;/strong&gt; and Unity &lt;strong&gt;21.3%&lt;/strong&gt;, while the Nasdaq lost 4.5%.&lt;/p&gt;
&lt;p&gt;AppLovin reported third-quarter revenue of &lt;strong&gt;$2.185 billion (+55%)&lt;/strong&gt;, $100 million above the top of its own guidance. Unity&amp;rsquo;s strategic revenue, the part tied to advertising, was &lt;strong&gt;$597 million (+59%)&lt;/strong&gt;, 8.5% above the top of its range. Unity also reported its first positive operating income under GAAP (accounting rules) since its IPO.&lt;/p&gt;
&lt;p&gt;One number stands out more. AppLovin&amp;rsquo;s installs grew 2% from a year earlier, but the money it earns per install grew 55%. Both beats came from how accurately each company can match an ad to a person. Their ad prediction models took a step up during the quarter, and advertisers raised budgets to match.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our call: this beat is not one-off demand. It is a step change in a prediction model showing up in the numbers. The same structure means next quarter can swing hard in either direction. Whether it lasts depends on when the next model improvement arrives. AppLovin&amp;rsquo;s consumer ads and Unity&amp;rsquo;s Spark matter for 2028 and later, more than for 2027.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="how-to-read-the-unfamiliar-terms"&gt;How to read the unfamiliar terms
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Term&lt;/th&gt;
 &lt;th&gt;In plain words&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Guidance&lt;/td&gt;
 &lt;td&gt;The range a company gives for next quarter&amp;rsquo;s numbers. A result above the top of the range is &amp;ldquo;better than guided&amp;rdquo;.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;User acquisition (UA) ads&lt;/td&gt;
 &lt;td&gt;Ads a game maker buys to find new players. The maker pays when a player installs the game.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ROAS&lt;/td&gt;
 &lt;td&gt;Revenue earned per dollar of ad spend. Advertisers keep raising budgets as long as they hit their ROAS target.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted EBITDA&lt;/td&gt;
 &lt;td&gt;Profit before interest, taxes and depreciation, with some further costs (such as stock pay) removed by the company. Useful for comparing one company&amp;rsquo;s quarters, not different companies.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Quarter-on-quarter, year-on-year&lt;/td&gt;
 &lt;td&gt;Growth versus the previous quarter, and versus the same quarter a year earlier.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CPM&lt;/td&gt;
 &lt;td&gt;The price of 1,000 ad impressions.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="one-ad-is-decided-by-one-auction-and-a-prediction-model-sets-the-price"&gt;One ad is decided by one auction, and a prediction model sets the price
&lt;/h2&gt;&lt;p&gt;A mobile game maker needs new players. So it buys ads inside other apps and games. Each time an ad slot opens, an auction decides which ad to show and how much to bid. AppLovin and Unity build and sell the machinery that runs that auction.&lt;/p&gt;
&lt;p&gt;The model that sets the bid predicts one thing: how likely this person is to install the game and spend money. AppLovin&amp;rsquo;s model is called Axon. Unity&amp;rsquo;s is called Vector.&lt;/p&gt;
&lt;p&gt;Think of an appraiser at an auction house. The sharper the appraisal, the higher the price the house can bid, and the advertiser still comes out ahead, so it is glad to spend. A bad appraisal works the other way.&lt;/p&gt;
&lt;p&gt;&lt;img alt="How one ad gets matched to one game slot. The game maker (advertiser) hands a budget to the ad engine. The engine’s prediction model sets the bid and sends the ad to an app or game that owns the slot. Installs and in-app purchases flow back and the model keeps retraining. AppLovin learns from auction records on its ad switchboard MAX plus install and purchase events. Unity learns from game engine runtime information and install and billing records on its ad network." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/ad-matching.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;The two companies are built a little differently.&lt;/p&gt;
&lt;p&gt;AppLovin also runs MAX, software that connects advertisers with the apps and games that own ad slots. It books as revenue only what is left after paying app owners their share. That is why its revenue growth splits into &amp;ldquo;money left per install × number of installs&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;Unity sells game-making tools (an engine) and also learns from information produced by games built on that engine. It has two businesses: Grow for advertising and Create for the tools. Vector is the ad model inside Grow, and it switched to a self-learning design in 2025.&lt;/p&gt;
&lt;h2 id="both-beat-the-top-of-their-guidance-and-the-stocks-moved-against-the-market"&gt;Both beat the top of their guidance, and the stocks moved against the market
&lt;/h2&gt;&lt;p&gt;First the results. Both companies beat the top of their own guidance. Almost all of Unity&amp;rsquo;s excess came from ads (Grow). Its tools business (Create) beat the top by only 1.3%.&lt;/p&gt;
&lt;p&gt;&lt;img alt="How far above the top of their own guidance. AppLovin revenue 4.8%, AppLovin adjusted EBITDA 5.9%, Unity strategic revenue 8.5%, Unity strategic Grow 12.1%, Unity adjusted EBITDA 35%." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/beat-vs-guidance.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Here are AppLovin&amp;rsquo;s numbers.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;AppLovin&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;vs a year ago&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.185 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted EBITDA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.842 billion (84% margin)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+59%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.439 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+72%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Diluted earnings per share&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$4.30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;4Q26 revenue guidance&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.365 to $2.395 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+43 to 44%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;And Unity&amp;rsquo;s.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Unity&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;vs a year ago&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Total revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$622 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+32%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Strategic revenue (Grow + Create)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$597 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+59%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Strategic Grow (ads)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$432 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+90%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Strategic Create (tools)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$165 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted EBITDA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$257 million (41% margin)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+134%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP operating income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$101 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−$126 million a year ago&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;4Q26 guidance&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Total revenue about $688 to $696 million, adjusted EBITDA $290 to $295 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Unity&amp;rsquo;s total revenue grew less (+32%) than its strategic revenue (+59%) because of the businesses it wound down. It closed its ironSource ad network on April 30 and sold Supersonic on August 4. Their revenue fell from $96 million to $26 million in a year and will be zero in 2027. From then on, total growth and strategic growth match.&lt;/p&gt;
&lt;p&gt;Pre-release averages on aggregator sites (they differ by provider, so treat them as a rough guide) were about $2.07 billion of revenue and $4.01 to $4.04 of earnings per share for AppLovin, and about $565 to $567 million of total revenue for Unity. Actual results were $2.185 billion and $4.30 for AppLovin and $622 million for Unity.&lt;/p&gt;
&lt;p&gt;&lt;img alt="On the day the Nasdaq fell 5.5%, both stocks rose. With the November 4 close set to 100, the November 6 close was 144.0 for AppLovin, 121.3 for Unity and 95.5 for the Nasdaq Composite." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/stock-reaction.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Take out the index move and the reaction is bigger than the raw numbers show. An analyst on the call said AppLovin had been down about 40% from the start of the year before the report. Its October 30 close of $228.50 was near the low.&lt;/p&gt;
&lt;h2 id="almost-all-of-applovins-growth-came-from-the-value-of-each-install"&gt;Almost all of AppLovin&amp;rsquo;s growth came from the value of each install
&lt;/h2&gt;&lt;p&gt;AppLovin&amp;rsquo;s quarterly filing (10-Q) explains revenue growth as &amp;ldquo;net revenue per install × number of installs&amp;rdquo;. Put three quarters side by side and the nature of this quarter shows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Net revenue per install (YoY)&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Installs (YoY)&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue (YoY)&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+93%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−18%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+59%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+58%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+53%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3Q26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;9 months&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+56%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;img alt="Almost all of AppLovin’s growth came from the value of each install. Net revenue per install was +93%, +58% and +55% in 1Q26, 2Q26 and 3Q26, while installs were −18%, −2% and +2%." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/price-vs-volume.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;In the first half, installs fell and growth came from price alone. In the third quarter installs turned positive for the first time while net revenue per install held at 55%. Multiply the two and you get about +58%, which is 3 points off the +55% revenue growth. We could not confirm the reason in the filing, so we use the two numbers only to read direction.&lt;/p&gt;
&lt;p&gt;The CEO said the improvement from the next-generation model was largest where &amp;ldquo;data is scarce&amp;rdquo;, meaning new and small advertisers and new categories. We read this as a sign that the model did not only lift bids for large existing advertisers. It widened the range of advertisers and installs the company can serve. That fits installs turning positive.&lt;/p&gt;
&lt;p&gt;By region, the United States brought $1.098 billion (+59.6%) and the rest of the world $1.087 billion (+51.5%). The direction matches the second-quarter call remark that consumer (web e-commerce) advertising is concentrated in the West.&lt;/p&gt;
&lt;p&gt;The CEO said gaming ad revenue grew about 50% year on year and re-accelerated. The company does not report segment revenue, so we cannot confirm that number. The next-generation model is only partly deployed, and the company made no promise on when it will be fully switched. The CEO described the change as comparable in size to the earlier Axon 2.0.&lt;/p&gt;
&lt;h2 id="a-better-model-turns-into-revenue-along-two-paths"&gt;A better model turns into revenue along two paths
&lt;/h2&gt;&lt;p&gt;Advertisers on both platforms keep raising budgets as long as they hit their ROAS target. So a better model opens two paths to revenue.&lt;/p&gt;
&lt;p&gt;The first path is a higher value for each install. If the model predicts higher value on the same impression, the bid goes up, and the advertiser earns the same or more for each ad dollar. The AppLovin CEO&amp;rsquo;s remark that &amp;ldquo;better targeting lifts CPM&amp;rdquo; points to this path. AppLovin&amp;rsquo;s +55% net revenue per install belongs here.&lt;/p&gt;
&lt;p&gt;The second path is more advertisers the company can serve. Advertisers, categories and regions that did not pay off before start to. AppLovin&amp;rsquo;s installs moving from −18% and −2% to +2% belongs here, and so does Unity&amp;rsquo;s advertiser accounts rising 27% quarter on quarter. Unity&amp;rsquo;s CFO said 90% of that growth came from existing advertisers.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Two ways a better prediction model turns into revenue. A sharper model raises bids, advertisers still hit their target and raise budgets. Revenue grows along two paths: each install is worth more, and more advertisers become viable." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/model-to-revenue.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;Unity&amp;rsquo;s CEO said Vector&amp;rsquo;s growth came from more installs and more player value, and that the take rate is not a variable the company manages. The two CEOs describe the same mechanism in different words.&lt;/p&gt;
&lt;h2 id="applovin-kept-87-of-its-added-revenue-as-profit"&gt;AppLovin kept 87% of its added revenue as profit
&lt;/h2&gt;&lt;p&gt;When AppLovin&amp;rsquo;s revenue rose $261 million in the quarter, adjusted EBITDA rose $228 million. The share of added revenue that stayed as profit (the incremental margin) was 87.5%. Computing cost barely followed the revenue increase.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;AppLovin item&lt;/th&gt;
 &lt;th&gt;3Q26&lt;/th&gt;
 &lt;th&gt;How to read it&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Cost of revenue (including network infrastructure)&lt;/td&gt;
 &lt;td&gt;$241 million (11.0% of revenue, 12.4% a year ago)&lt;/td&gt;
 &lt;td&gt;Revenue jumped and the ratio fell.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Added cost of revenue per added revenue dollar&lt;/td&gt;
 &lt;td&gt;$0.06 vs prior quarter, $0.09 vs prior year&lt;/td&gt;
 &lt;td&gt;Inside the $0.10 the company gave in the second quarter.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP operating income&lt;/td&gt;
 &lt;td&gt;$1.709 billion (78.2% margin)&lt;/td&gt;
 &lt;td&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Free cash flow (FCF)&lt;/td&gt;
 &lt;td&gt;$1.418 billion&lt;/td&gt;
 &lt;td&gt;77% of adjusted EBITDA. The second quarter&amp;rsquo;s 53% reflected the timing of cash tax payments.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Share repurchases&lt;/td&gt;
 &lt;td&gt;4.1 million shares for $1.4 billion (about $341 on average)&lt;/td&gt;
 &lt;td&gt;Up from 1.1 million shares and $550 million in the second quarter. The CFO called it opportunistic buying of a price gap. The authorization grew by $3.2 billion to $3.6 billion.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Balance sheet&lt;/td&gt;
 &lt;td&gt;Cash $3.17 billion, long-term debt $3.52 billion&lt;/td&gt;
 &lt;td&gt;Net debt about $340 million, 0.05 times the last four quarters of adjusted EBITDA.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;R&amp;amp;D expense rose 132% year on year. The 10-Q gives $55 million of added stock-pay-related labor cost as the reason, and the added cost of computing for training does not show up much in this line. The cost to watch next is the renewal of its minimum cloud purchase commitment: $1.3 billion over three years as of August 2024, of which $1.1 billion had been paid by the end of September.&lt;/p&gt;
&lt;h2 id="unitys-profit-jumped-because-the-same-engine-met-a-clean-up"&gt;Unity&amp;rsquo;s profit jumped because the same engine met a clean-up
&lt;/h2&gt;&lt;p&gt;Unity&amp;rsquo;s strategic Grow rose 31.2% from the prior quarter and 90% from a year earlier. The company does not report Vector revenue, but it gave growth rates. Vector grew 23% quarter on quarter in the second quarter (guided 12 to 13%) and 40% in the third (guided 20%). That is nearly double the guide two quarters in a row.&lt;/p&gt;
&lt;p&gt;So we estimated Vector&amp;rsquo;s revenue. We assume Grow revenue outside Vector (LevelPlay mediation and similar) is roughly constant from quarter to quarter. Backing the $103 million rise in strategic Grow out of Vector&amp;rsquo;s +40% gives third-quarter Vector revenue of about $359 million. The same assumption gives $256 million for the second quarter, which matches the &amp;ldquo;annualized above $1 billion&amp;rdquo; remark on the second-quarter call ($256 million × 4 = $1.02 billion).&lt;/p&gt;
&lt;p&gt;&lt;img alt="Unity revenue mix. Estimated Vector revenue grew from $159 million in 3Q25 to $359 million in 3Q26, from 34% to 58% of total revenue, while the closed or sold businesses fell from $96 million to $26 million." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/unity-revenue-stack.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;If the estimate holds, two things follow. Unity is already a Vector company. And the fourth-quarter guide (Vector up 21 to 23% on the quarter) assumes almost no growth outside Vector. The estimate rests on the assumption that non-Vector Grow is constant, and we put the error at about ±$15 million.&lt;/p&gt;
&lt;p&gt;The profit side is in the next figure. Adjusted EBITDA rose $96 million from the prior quarter. About 76% of that rise came from higher gross profit and about 23% from lower marketing cost. The lower marketing cost reflects user-acquisition spend that disappeared with Supersonic, sold in August.&lt;/p&gt;
&lt;p&gt;&lt;img alt="Where Unity’s profit came from. Adjusted EBITDA rose from $160.2 million in the second quarter through gross profit +72.9, lower marketing +22.1, lower R and D spending +1.6 and higher G and A spending −0.2 to $256.6 million in the third quarter." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/unity-profit-waterfall.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;GAAP operating income turned positive later than adjusted EBITDA because amortization of intangible assets and stock pay were large. Both fell, and the first profit arrived in the third quarter.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Unity item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3Q25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3Q26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted EBITDA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$110 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$160 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$257 million&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted EBITDA margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;29%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;41%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP operating income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−$126 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−$32 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+$101 million&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amortization of intangibles&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$125 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$78 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$77 million&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stock pay&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$92 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$76 million&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$63 million&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The company guided amortization to $50 million in the fourth quarter and $15 million a quarter in 2027. It guided stock pay to $75 million a quarter in 2027 and the effective tax rate to 5 to 10%. If that holds, GAAP profit in 2027 sits closer to adjusted EBITDA.&lt;/p&gt;
&lt;p&gt;Unity&amp;rsquo;s tools business (Create) has settled into a business growing about 9% a year. The fourth-quarter guide is +4 to 6%. The drivers are price and mix more than seat count.&lt;/p&gt;
&lt;p&gt;On the balance sheet, Unity has $2.65 billion of cash. Its convertible notes are $558 million due November 15, 2026, $1 billion due 2027 and $690 million due 2030. Repaying the $558 million in cash on November 15 leaves cash of about $2.1 billion and $1.69 billion of convertible notes. Shares outstanding are 443.8 million, up 2.5% from the start of the year.&lt;/p&gt;
&lt;p&gt;China is a new risk factor. China revenue was $157 million (+74%), 25% of the total, up from 19% a year ago and 21% in the second quarter. The company did not explain the mix. We guess it is Vector&amp;rsquo;s demand from Chinese publishers buying overseas users, plus strong Create sales in China.&lt;/p&gt;
&lt;p&gt;We also share new analyses on Telegram as they are published. You are welcome to join us at &lt;a class="link" href="https://t.me/korea_invest_insights" target="_blank" rel="noopener"
 &gt;@korea_invest_insights&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="whether-it-lasts-depends-on-when-the-next-model-improvement-arrives"&gt;Whether it lasts depends on when the next model improvement arrives
&lt;/h2&gt;&lt;p&gt;Both companies put only model improvements that are already running into their guidance. They leave out models not yet deployed. AppLovin&amp;rsquo;s CFO said in the second and third quarters that guidance counts &amp;ldquo;improvements already running, further consumer expansion, normal seasonality and rising training cost&amp;rdquo;. Unity&amp;rsquo;s CFO also builds guidance from run rate, seasonality and planned improvements, and said &amp;ldquo;unpredictable step changes like the third quarter&amp;rdquo; can happen.&lt;/p&gt;
&lt;p&gt;So whether next quarter beats the top of guidance comes down to whether a new improvement lands inside the quarter. AppLovin&amp;rsquo;s second and third quarters are the two sides of one variable.&lt;/p&gt;
&lt;p&gt;&lt;img alt="AppLovin’s quarterly growth moves in steps. Quarter-on-quarter revenue growth was 11.6% in 3Q25, 18.0% in 4Q25, 11.1% in 1Q26, 4.4% in 2Q26 and 13.6% in 3Q26, and the 4Q26 guide is 8.2 to 9.6%." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/applovin-growth-steps.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;In the second quarter, as the CEO put it, the pace of improvement was lighter than usual and the next jump came just after quarter-end. Revenue came in slightly below the guidance midpoint and the stock fell 16%. In the third quarter that jump landed early, revenue beat the top of the range by $100 million and the stock rose 44% in two days. Averaged over the two quarters, growth returns to normal.&lt;/p&gt;
&lt;p&gt;AppLovin stresses that it beat the top of its range in 14 of 16 quarters. Still, one late quarter like the second moves the stock a lot. The fourth-quarter guide (+8 to 10% on the quarter) is below past fourth quarters (+20%, +18%). That looks conservative given the consumer-ad peak season, but the CEO warned of &amp;ldquo;a harder first quarter after a big fourth&amp;rdquo;. The bigger the consumer-ad share, the bigger the seasonality.&lt;/p&gt;
&lt;p&gt;Unity&amp;rsquo;s CEO drew a line: &amp;ldquo;not 40% every quarter&amp;rdquo;. The fourth-quarter Vector guide is +21 to 23%.&lt;/p&gt;
&lt;p&gt;Put the five variables side by side and the ranking shows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Variable&lt;/th&gt;
 &lt;th&gt;AppLovin&lt;/th&gt;
 &lt;th&gt;Unity&lt;/th&gt;
 &lt;th&gt;Our read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1. Model improvement cycle&lt;/td&gt;
 &lt;td&gt;Next-generation model partly deployed, no date for full switch&lt;/td&gt;
 &lt;td&gt;6 to 10 product changes a quarter; CEO: &amp;ldquo;not 40% every quarter&amp;rdquo;&lt;/td&gt;
 &lt;td&gt;Drives 2027 numbers most. Both are mid-switch, so room for another jump remains through the first half.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2. Demand pool (game ad budgets)&lt;/td&gt;
 &lt;td&gt;Most revenue is gaming. Installs turning positive shows a wider reach&lt;/td&gt;
 &lt;td&gt;Mobile game ad market above $40 billion, double-digit growth (CEO claim)&lt;/td&gt;
 &lt;td&gt;Both accelerate at once, so there is no sign yet of one taking from the other.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3. Supply of ad slots (impressions)&lt;/td&gt;
 &lt;td&gt;MAX publisher revenue up double digits; market-wide impressions up single to low double digits&lt;/td&gt;
 &lt;td&gt;Creator daily users up 25% from early August&lt;/td&gt;
 &lt;td&gt;Impressions are not growing fast, so growth leans on value per impression. It lasts only while advertiser ROAS holds.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;4. New demand pools&lt;/td&gt;
 &lt;td&gt;Consumer (web e-commerce), then non-game apps, open web and CTV&lt;/td&gt;
 &lt;td&gt;Spark&amp;rsquo;s own ad slots, asset store&lt;/td&gt;
 &lt;td&gt;Small in 2027 numbers. A 2028-and-later condition.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5. Cost and cash&lt;/td&gt;
 &lt;td&gt;Cloud commitment renewal, stock pay&lt;/td&gt;
 &lt;td&gt;Convertible repayment, stock pay 10% of revenue, China share 25%&lt;/td&gt;
 &lt;td&gt;Cash generation covers cost risk at both. Unity&amp;rsquo;s China share is a new risk.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The dominant variable is the first. The second and third set where growth converges if the first stops. The AppLovin CEO&amp;rsquo;s long-run 30% compounding, Unity&amp;rsquo;s Create at 9% and Vector&amp;rsquo;s &amp;ldquo;healthy growth&amp;rdquo; are the companies&amp;rsquo; own versions of that landing point.&lt;/p&gt;
&lt;p&gt;For two companies in the same market to grow near 50% and 90%, the market has to grow while Meta, Google and older networks lose share. We covered ad budgets moving between platforms, with Google and Meta, in our &lt;a class="link" href="https://koreainvestinsights.com/post/meta-ad-edge-google-amazon-share-shift-2026-10-30/" &gt;October 30 post&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="the-new-growth-engines-matter-for-2028-and-later-not-2027"&gt;The new growth engines matter for 2028 and later, not 2027
&lt;/h2&gt;&lt;h3 id="applovin-consumer-web-e-commerce-ads"&gt;AppLovin: consumer (web e-commerce) ads
&lt;/h3&gt;&lt;p&gt;AppLovin is widening beyond games, and online shopping ads come first. Visibility is still low. The company does not disclose advertiser counts or revenue share, and the CEO said it does not give KPIs because the business is &amp;ldquo;not at a stable stage&amp;rdquo;.&lt;/p&gt;
&lt;p&gt;The public clues are these. In the second quarter, consumer advertiser spend was up 28% from the fourth quarter of 2025. A new customer spends on average $70,000 in its first year. The company goes after mid-sized advertisers first and has put the long tail on hold. The biggest bottleneck is ad creative. It is also widening integrations with analytics platforms such as Triple Whale.&lt;/p&gt;
&lt;p&gt;The CEO said that supply expansion beyond games (non-game apps, then the open web, then CTV) is &amp;ldquo;at pilot stage and will not contribute to numbers for now&amp;rdquo;. Sell-side estimates vary widely, so we did not use them.&lt;/p&gt;
&lt;h3 id="unity-spark-and-the-google-partnership"&gt;Unity: Spark and the Google partnership
&lt;/h3&gt;&lt;p&gt;Spark is a browser-based, no-code, collaborative creation tool announced on October 7. It deploys to Google Playground, aims for a private beta by the end of 2026 and a full launch in 2027, and is non-exclusive. Guidance contains no revenue from it. We read it as a door for new users more than as revenue.&lt;/p&gt;
&lt;p&gt;The CFO stressed two sets of economics. Unity gets its own (O&amp;amp;O) ad inventory outside the mediation ecosystem. And Unity&amp;rsquo;s own assets sold in its asset store carry a 100% margin.&lt;/p&gt;
&lt;p&gt;Creator daily users rose 25% from early August after Unity opened its coding agent and CLI. That opening includes a Claude Code plugin. We disclose our interest below.&lt;/p&gt;
&lt;h2 id="how-to-read-the-signs-that-it-is-breaking"&gt;How to read the signs that it is breaking
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;AppLovin: two quarters in a row that the company explains by &amp;ldquo;no new model landed&amp;rdquo;, with quarter-on-quarter growth at or below the seasonal average, would say growth has not escaped its dependence on step jumps. The first test is the first quarter of 2027. If net revenue per install and installs both slow to single digits, the pricing engine has run dry.&lt;/li&gt;
&lt;li&gt;Unity: if Vector&amp;rsquo;s quarterly growth falls below 15% while advertiser account growth stalls, the share shift is over. If strategic Grow outside Vector (about $70 million) starts to shrink, LevelPlay&amp;rsquo;s mediation position is weakening.&lt;/li&gt;
&lt;li&gt;Common 1: if one company accelerates in a quarter when the other slows, that is a zero-sum phase. Today both accelerate. What Meta and Google say about gaming ads, and publishers&amp;rsquo; in-app purchase trends (secondary sources such as Sensor Tower), are side indicators.&lt;/li&gt;
&lt;li&gt;Common 2: if advertisers&amp;rsquo; ROAS does not hold, the higher CPM reverses. For AppLovin&amp;rsquo;s consumer ads there is a measurement debate between platform-reported ROAS and truly incremental revenue, and it is not yet confirmed that more advertiser spend leads to more purchases by end consumers.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-prices-may-already-hold-part-of-the-next-jump"&gt;The prices may already hold part of the next jump
&lt;/h2&gt;&lt;p&gt;Both stocks rose a lot in two days. We think the prices have taken in some of the next jump. These numbers are our own calculation from the November 6 close.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;At the November 6 close&lt;/th&gt;
 &lt;th style="text-align: right"&gt;AppLovin&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Unity&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Closing price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$391.22&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$51.49&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Market value&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $129.6 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $22.8 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Enterprise value (EV)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $130 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $22.7 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 estimated adjusted EBITDA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$7.02 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $850 million&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EV ÷ adjusted EBITDA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 18.5x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 26.8x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;On fourth-quarter annualized adjusted EBITDA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 19.4x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EV ÷ 2026 estimated revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 9.6x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 estimated FCF yield&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 3.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Method: market value is shares outstanding times the closing price, and enterprise value is market value plus debt minus cash. AppLovin&amp;rsquo;s 2026 adjusted EBITDA is nine months of actual results ($5.01 billion) plus the midpoint of the fourth-quarter guide ($2.01 billion), and the FCF yield assumes about $5 billion of 2026 FCF (a 72% conversion). Unity&amp;rsquo;s is nine months of actuals plus the guidance midpoint, with convertible notes counted only as debt.&lt;/p&gt;
&lt;h2 id="what-to-check-next"&gt;What to check next
&lt;/h2&gt;&lt;p&gt;&lt;img alt="What to check next. November 15, Unity convertible repayment. By December 31, Spark private beta. Late January, Meta and Google fourth-quarter calls. Early February, AppLovin and Unity fourth-quarter results." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/applovin-unity-3q26-prediction-model-step-up-price-per-install-2026-11-08/next-checks.en.svg"&gt;
&lt;/p&gt;
&lt;p&gt;In the early-February results, we check whether AppLovin beats the top of its fourth-quarter guide ($2.395 billion) again, how it guides the first quarter on the prior quarter, and whether it discloses consumer-ad metrics. The annual numbers for net revenue per install and installs arrive in the 10-K. For Unity, the tests are whether Vector hits +21 to 23% on the quarter, whether advertiser account growth continues, whether the 2027 amortization and stock-pay guidance holds, and how China&amp;rsquo;s share is explained.&lt;/p&gt;
&lt;h2 id="what-this-means-for-korean-investors"&gt;What this means for Korean investors
&lt;/h2&gt;&lt;p&gt;For mobile game publishers (Krafton, Netmarble and others), this change means better user-acquisition efficiency, but the same model also lifts CPM. Translating it into simple marketing savings gets it wrong. For a publisher, spend rises when ROAS stays the same.&lt;/p&gt;
&lt;p&gt;The link to memory and AI infrastructure is small. AppLovin&amp;rsquo;s computing cost is under 10% of its added revenue. Ad prediction models are often cited as proof that AI investment pays back, but their contribution to hyperscaler revenue is small. AppLovin&amp;rsquo;s $1.3 billion three-year cloud commitment works out to about $430 million a year.&lt;/p&gt;
&lt;p&gt;The two stocks share exposure to mobile-game ad budgets, the multiple paid for AI growth stocks, and step changes in models. Holding both may give less diversification than it appears. AppLovin stands out for pricing power that is disclosed in numbers in its 10-Q and for cash returns. Unity stands out for more room left in growth acceleration and operating leverage.&lt;/p&gt;
&lt;h2 id="limits-of-the-data"&gt;Limits of the data
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Vector revenue and AppLovin&amp;rsquo;s split of gaming versus consumer revenue are not disclosed. Our quarterly Vector estimate rests on non-Vector strategic Grow staying constant, and the roughly 50% gaming growth at AppLovin is a CEO remark.&lt;/li&gt;
&lt;li&gt;Consensus figures are pre-release values from aggregator sites (MarketBeat, Zacks, Benzinga) and differ by provider. We could not obtain raw LSEG or FactSet values.&lt;/li&gt;
&lt;li&gt;Earnings call transcripts come from Motley Fool and StockAnalysis (Quartr). They are not the companies&amp;rsquo; official transcripts. Quotes are short and the numbers were checked against filings.&lt;/li&gt;
&lt;li&gt;Market share figures (an iOS ad revenue share of 39 to 44% and similar) and Triple Whale&amp;rsquo;s ROAS compilations come from blogs and vendors, so we did not use them as a basis.&lt;/li&gt;
&lt;li&gt;AppLovin&amp;rsquo;s shareholder letter holds only charts and financial tables, so we relied on the earnings call for narrative. Unity has issued no shareholder letter since the third quarter of 2024.&lt;/li&gt;
&lt;li&gt;We did not recheck, in this 10-Q, the gross-versus-net recognition basis for Unity&amp;rsquo;s advertising revenue.&lt;/li&gt;
&lt;li&gt;This post restates our November 8 analysis note for readers. We did not rerun source checks for publication, and the numbers were tied back to the note&amp;rsquo;s tables one line at a time. Prices are US regular-session closes of November 4 to 6.&lt;/li&gt;
&lt;li&gt;A single AI vendor reviewed this analysis. No second vendor checked it.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;AppLovin third-quarter release (8-K): &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001751008/000175100826000068/exhibit991-3q26earningspre.htm" target="_blank" rel="noopener"
 &gt;SEC Ex.99.1&lt;/a&gt;, &lt;a class="link" href="https://s21.q4cdn.com/165405286/files/content_files/3Q26-Financial-Update.pdf" target="_blank" rel="noopener"
 &gt;3Q26 Financial Update&lt;/a&gt;, &lt;a class="link" href="https://d18rn0p25nwr6d.cloudfront.net/CIK-0001751008/5d38e9cf-1f99-43b1-adcc-36979436fce4.pdf" target="_blank" rel="noopener"
 &gt;third-quarter 10-Q&lt;/a&gt;, &lt;a class="link" href="https://d18rn0p25nwr6d.cloudfront.net/CIK-0001751008/6d342606-71ad-4f30-891c-39ec65255384.pdf" target="_blank" rel="noopener"
 &gt;second-quarter 10-Q&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001751008/000175100826000057/exhibit991-2q26earningspre.htm" target="_blank" rel="noopener"
 &gt;second-quarter 8-K&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Unity third-quarter release (8-K): &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001810806/000181080626000053/a2026q3ex-991.htm" target="_blank" rel="noopener"
 &gt;SEC Ex.99.1&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1810806/000181080626000055/unity-20260930.htm" target="_blank" rel="noopener"
 &gt;third-quarter 10-Q&lt;/a&gt;, &lt;a class="link" href="https://s205.q4cdn.com/200788659/files/doc_earnings/2026/q3/generic/2026_Q3-Spreadsheet_FINAL.xlsx" target="_blank" rel="noopener"
 &gt;3Q26 Earnings Spreadsheet&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001810806/000181080626000041/a2026q2ex-991.htm" target="_blank" rel="noopener"
 &gt;second-quarter 8-K&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Earnings call transcripts (secondary): &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/11/05/applovin-app-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;AppLovin third quarter&lt;/a&gt;, &lt;a class="link" href="https://stockanalysis.com/stocks/app/transcripts/661674-q2-2026/" target="_blank" rel="noopener"
 &gt;AppLovin second quarter&lt;/a&gt;, &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/11/05/unity-software-u-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;Unity third quarter&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Unity first-quarter preliminary results: &lt;a class="link" href="https://investors.unity.com/news/news-details/2026/Unity-Releases-Preliminary-First-Quarter-Results-Exceeding-Guidance-Will-Enhance-Growth-and-Profitability-by-Exiting-Non-Strategic-Ad-Businesses/default.aspx" target="_blank" rel="noopener"
 &gt;Unity release&lt;/a&gt;. Spark and the Google partnership: &lt;a class="link" href="https://finance.yahoo.com/technology/ai/articles/google-unity-partner-ai-gaming-120000661.html" target="_blank" rel="noopener"
 &gt;Yahoo Finance&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Consensus (secondary): &lt;a class="link" href="https://www.marketbeat.com/earnings/reports/2026-11-4-applovin-co-stock/" target="_blank" rel="noopener"
 &gt;MarketBeat&lt;/a&gt;, &lt;a class="link" href="https://stockstory.org/us/stocks/nasdaq/app/news/earnings/applovin-nasdaqapp-delivers-strong-q3-2026-numbers-stock-jumps-249percent" target="_blank" rel="noopener"
 &gt;StockStory&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/markets/stocks/articles/unity-q3-earnings-beat-estimates-152300876.html" target="_blank" rel="noopener"
 &gt;Zacks&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/markets/stocks/articles/unity-software-stock-soars-22-124617311.html" target="_blank" rel="noopener"
 &gt;Investing.com&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Which one cracks first in the February reports? Reply with A, B or C and one data point that backs your pick.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;A. AppLovin: consumer ads raise seasonality and its first-quarter 2027 growth on the prior quarter weakens.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;B. Unity: Vector, now about $440 million a quarter, meets the limits of size and growth on the quarter falls below 15%.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;C. Neither: new models keep arriving and beats above the top of guidance continue through the first half of 2027.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;This is for research and information only, not investment advice. Estimates and scenarios are analytical assumptions, not trading instructions or individual portfolio decisions. Figures are dated data, and some are secondary compilations. We used Anthropic&amp;rsquo;s Claude for research and editing. Unity&amp;rsquo;s opening of its coding agent and CLI included a Claude Code plugin, and its CEO linked the rise in creator daily users to that opening, so Anthropic is a name with an interest in this post. We do not judge the size of its effect.&lt;/p&gt;</description></item><item><title>ServiceNow Fell 27% in a Day. Its 98% Renewal Rate Cannot See the Seat Freeze.</title><link>https://koreainvestinsights.com/post/servicenow-renewal-rate-seat-freeze-saas-agent-bypass-test-2026-10-30/</link><pubDate>Fri, 30 Oct 2026 10:55:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/servicenow-renewal-rate-seat-freeze-saas-agent-bypass-test-2026-10-30/</guid><description>&lt;img src="https://koreainvestinsights.com/" alt="Featured image of post ServiceNow Fell 27% in a Day. Its 98% Renewal Rate Cannot See the Seat Freeze." /&gt;&lt;p&gt;ServiceNow (NOW) closed at $109.67 on October 29, down 26.95%. CNBC called it the stock&amp;rsquo;s worst day ever. The company beat every line of its third-quarter guidance and raised its year-end AI annual contract value (ACV) target from $1 billion at the start of the year to $1.7 billion or more. In the same release it guided fourth-quarter cRPO growth to 18.5% in constant currency. cRPO is contracted revenue that will be recognized over the next 12 months, and it grew 21% in the third quarter. &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000092/erq3fy26.htm" target="_blank" rel="noopener"
 &gt;ServiceNow Q3 2026 release (8-K)&lt;/a&gt;, &lt;a class="link" href="https://www.cnbc.com/2026/10/29/servicenow-now-earnings-q3-2026.html" target="_blank" rel="noopener"
 &gt;CNBC&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The market read it fast: AI is eating software. We think that reading has the direction right and the path wrong. Customers did not leave. The numbers point to customers that stopped buying more seats.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our view: the threat this quarter revealed is an expansion freeze. Customers are not adding people, so they are not adding seats. Churn is not visible yet. ServiceNow&amp;rsquo;s 98% renewal rate cannot see this change, by definition. For the next one to two years the seat freeze will drive results. After that, usage revenue will decide which companies win. So SaaS stocks should be sorted less by &amp;ldquo;seat pricing or usage pricing&amp;rdquo; and more by one question: can an AI agent route around this system?&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;Data are as of the US regular-session close on October 29, 2026. Company figures come from ServiceNow&amp;rsquo;s 8-K and 10-Q filings with the SEC. Management quotes come from the Q3 earnings call transcript.&lt;/p&gt;
&lt;h2 id="the-numbers-beat-the-direction-turned"&gt;The numbers beat. The direction turned
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Q3 2026&lt;/th&gt;
 &lt;th&gt;How to read it&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Subscription revenue&lt;/td&gt;
 &lt;td&gt;$3.988bn, +21% (+20.5% constant currency)&lt;/td&gt;
 &lt;td&gt;0.5 points above the top of guidance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;cRPO&lt;/td&gt;
 &lt;td&gt;$13.65bn, +21% constant currency&lt;/td&gt;
 &lt;td&gt;1 point above guidance. The CFO said half of the beat was Q4 renewals pulled forward&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Year-end AI ACV target&lt;/td&gt;
 &lt;td&gt;$1.7bn or more&lt;/td&gt;
 &lt;td&gt;Up from $1.0bn in January and $1.5bn in May. Q3 AI net new ACV +65%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Renewal rate&lt;/td&gt;
 &lt;td&gt;98%&lt;/td&gt;
 &lt;td&gt;97% a year earlier&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customers above $5m ACV&lt;/td&gt;
 &lt;td&gt;680, +23%&lt;/td&gt;
 &lt;td&gt;Large customers are still expanding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Share buybacks&lt;/td&gt;
 &lt;td&gt;Effectively zero&lt;/td&gt;
 &lt;td&gt;$584m in Q3 2025. Borrowing rose after the Veza (March) and Armis (April) acquisitions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 cRPO guidance&lt;/td&gt;
 &lt;td&gt;+18.5% constant currency&lt;/td&gt;
 &lt;td&gt;Q3: +21%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;img alt="ServiceNow cRPO growth in constant currency: 21% in Q4 2025, 21% in Q1 2026, 21.5% in Q2, 21% in Q3 and 18.5% guided for Q4. Adding back every item the CFO cited, 1.25 points, gives 19.75%. Over the same period the year-end AI ACV target rose from $1.0 billion to $1.5 billion and then to $1.7 billion or more." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/servicenow-renewal-rate-seat-freeze-saas-agent-bypass-test-2026-10-30/crpo-vs-ai-target.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;The CFO explained the Q4 slowdown with one-off items. Early renewals pulled into Q3 cost about 0.5 points. A higher mix of on-premises public-sector deals costs 0.25 points. Backlog under termination-for-convenience (T4C) clauses, which revenue includes but cRPO excludes, costs about 0.5 points. She also called investment in deployment staff for large deals a small headwind. &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/10/29/servicenow-now-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;Q3 call transcript (Motley Fool)&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Add back every item the CFO quantified and Q4 is still about 19.75%. That is below Q3&amp;rsquo;s 21%. The one-off story does not explain the whole slowdown.&lt;/p&gt;
&lt;h2 id="the-ai-target-rose-70-why-did-the-total-slow"&gt;The AI target rose 70%. Why did the total slow?
&lt;/h2&gt;&lt;p&gt;A $1.7 billion AI ACV is about 10% of roughly $17 billion, the Q4 subscription guidance annualized. If that 10% grows fast and total contracted growth still falls, the business outside AI is likely slowing. We lined up three explanations for that slowdown against this quarter&amp;rsquo;s evidence.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Explanation&lt;/th&gt;
 &lt;th&gt;Mechanism&lt;/th&gt;
 &lt;th&gt;Evidence this quarter&lt;/th&gt;
 &lt;th&gt;Verdict&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Seat cuts at renewal&lt;/td&gt;
 &lt;td&gt;Customers renew with fewer human seats&lt;/td&gt;
 &lt;td&gt;A 98% renewal rate cannot rule this out. Per the 10-Q, the rate excludes changes in price or users and does not reflect lower purchases by customers who stay&lt;/td&gt;
 &lt;td&gt;Open. Filings cannot settle it&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Seat expansion freeze plus AI repricing&lt;/td&gt;
 &lt;td&gt;Customers freeze headcount, so new seats stop; premium AI tiers raise the price of existing seats&lt;/td&gt;
 &lt;td&gt;AI ACV grows while total cRPO slows. Customers above $5m +23%. In 2024 the CFO said the AI tier, Pro Plus, carries a 60% list uplift over Pro&lt;/td&gt;
 &lt;td&gt;Best fit&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Lag in consumption billing&lt;/td&gt;
 &lt;td&gt;Agent calls grow, but only usage above fixed credits is billed&lt;/td&gt;
 &lt;td&gt;Agentic AI consumption up 13x; Action Fabric assists up 500% month over month. The 10-K says consumption pricing applies when usage exceeds the fixed number of service credits&lt;/td&gt;
 &lt;td&gt;Volume confirmed. Revenue not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The renewal-rate definition is in the &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000096/now-20260930.htm" target="_blank" rel="noopener"
 &gt;10-Q&lt;/a&gt;, the consumption terms in the &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000007/now-20251231.htm" target="_blank" rel="noopener"
 &gt;2025 10-K&lt;/a&gt;, and the Pro Plus pricing in &lt;a class="link" href="https://www.constellationr.com/insights/news/servicenow-cfo-outlines-method-pro-plus-sku-pricing" target="_blank" rel="noopener"
 &gt;Constellation Research&amp;rsquo;s report&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;If the second explanation is right, time is not on ServiceNow&amp;rsquo;s side. Repricing is a one-time step that stops once customers finish upgrading. Seat compression compounds every year as AI gets better. A step happens once; a trend keeps going.&lt;/p&gt;
&lt;p&gt;Read management&amp;rsquo;s own example with care. The CFO said internal AI now delivers more than $1 billion of annualized value, and that ServiceNow grew quota-carrying headcount at double digits while total headcount fell slightly. Yet the 10-Q attributes higher cost of revenue and sales expense &amp;ldquo;primarily&amp;rdquo; to increased headcount. Staff from this year&amp;rsquo;s acquisitions may be mixed in. ServiceNow&amp;rsquo;s internal seat-compression story is still a management claim.&lt;/p&gt;
&lt;h2 id="how-the-seat-threat-unfolds"&gt;How the seat threat unfolds
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stage&lt;/th&gt;
 &lt;th&gt;What happens&lt;/th&gt;
 &lt;th&gt;What shows up&lt;/th&gt;
 &lt;th&gt;ServiceNow now&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Expansion freeze&lt;/td&gt;
 &lt;td&gt;Customer headcount stops growing, so fewer new seats&lt;/td&gt;
 &lt;td&gt;Net new ACV and cRPO slow; renewal rate holds&lt;/td&gt;
 &lt;td&gt;Here&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Renewal cuts&lt;/td&gt;
 &lt;td&gt;Customers cut seats at renewal&lt;/td&gt;
 &lt;td&gt;Invisible in the company&amp;rsquo;s renewal rate. Falling net revenue retention (NRR) at companies that disclose it; management talk of downsells&lt;/td&gt;
 &lt;td&gt;Not confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price pressure&lt;/td&gt;
 &lt;td&gt;AI-native rivals and self-cannibalization push seat prices down&lt;/td&gt;
 &lt;td&gt;Lower price per seat; more self-serve, consumption-priced products&lt;/td&gt;
 &lt;td&gt;Early signal: ServiceNow launched Flow, a conversational service desk that runs outside its main platform&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Exposure differs by seat type. The most exposed seats belong to people who work through requests: IT help desk, customer support and HR case handling. Next come products billed per employee, such as HR and payroll. Developer seats could go either way.&lt;/p&gt;
&lt;p&gt;On the call, ServiceNow said its Level 1 service desk agent resolves incidents 85% faster than human agents. The work that agent does is the work of ServiceNow&amp;rsquo;s own ITSM fulfiller seats. ServiceNow is selling a product that competes with its own seats.&lt;/p&gt;
&lt;p&gt;CEO Bill McDermott said customers like seat pricing because it makes budgeting predictable. That predictability cuts both ways. When customers cut seats, ServiceNow&amp;rsquo;s revenue falls just as predictably.&lt;/p&gt;
&lt;h2 id="what-usage-revenue-must-cover"&gt;What usage revenue must cover
&lt;/h2&gt;&lt;p&gt;We used a simple break-even formula: new revenue / old revenue = (1 − seat cut) × (1 + AI price uplift on remaining seats) + usage revenue (as a share of old revenue). We assume every remaining seat upgrades.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Seats cut&lt;/th&gt;
 &lt;th&gt;Usage revenue needed at a 30% uplift&lt;/th&gt;
 &lt;th&gt;Usage revenue needed at a 60% uplift&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;20%&lt;/td&gt;
 &lt;td&gt;None (revenue 1.04x)&lt;/td&gt;
 &lt;td&gt;None (1.28x)&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;30%&lt;/td&gt;
 &lt;td&gt;9%&lt;/td&gt;
 &lt;td&gt;None (1.12x)&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;40%&lt;/td&gt;
 &lt;td&gt;22%&lt;/td&gt;
 &lt;td&gt;4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;50%&lt;/td&gt;
 &lt;td&gt;35%&lt;/td&gt;
 &lt;td&gt;20%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;img alt="Usage revenue needed to keep total revenue flat, by seat cut and AI price uplift on remaining seats. A 30% seat cut with a 30% uplift needs 9%. A 40% cut needs 22% at a 30% uplift and 4% at 60%. A 50% cut needs 35% and 20%." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/servicenow-renewal-rate-seat-freeze-saas-agent-bypass-test-2026-10-30/seat-cut-breakeven.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;If the full 60% Pro Plus list uplift sticks, repricing alone holds revenue up to a 30% seat cut. If the uplift is only 30%, the limit is a 20% cut. At a 40% cut, usage revenue must reach 4% to 22% of today&amp;rsquo;s revenue; at 50%, 20% to 35%. ServiceNow&amp;rsquo;s AI ACV is about 10% of subscription revenue today, and that already includes repricing. If fewer than all customers upgrade, or if the goal is 20% growth instead of flat revenue, the usage revenue needed is larger than the table shows.&lt;/p&gt;
&lt;p&gt;Our reading is that the market is now discounting the chance that seat cuts pass 40% within three to five years. This quarter offered no evidence that lowers that chance.&lt;/p&gt;
&lt;h2 id="when-an-agent-cannot-route-around-the-system"&gt;When an agent cannot route around the system
&lt;/h2&gt;&lt;p&gt;For SaaS to become the agent&amp;rsquo;s tool and replace lost seat revenue, three conditions must hold.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;First, the agent must not be able to route around it.&lt;/strong&gt; The work cannot finish without passing through this system&amp;rsquo;s authoritative records, write permissions and audit trail. ServiceNow&amp;rsquo;s CMDB and its change and approval records likely meet this test. A CMDB is the system of record for a company&amp;rsquo;s IT assets and how they connect.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Second, usage revenue must exceed the seat revenue it replaces.&lt;/strong&gt; That is the table above. Consumption up 13x and Action Fabric assists up 500% a month are volume measures. The company did not disclose the revenue this volume produced. McDermott himself placed &amp;ldquo;the true acceleration&amp;rdquo; in 2027.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Third, the tool must keep pricing power against the front door.&lt;/strong&gt; If ChatGPT, Gemini or Copilot becomes where work requests start, those front doors can squeeze the price of the tools behind them. This quarter ServiceNow widened its partnership with OpenAI and launched a ServiceNow plugin for ChatGPT. Users will run ServiceNow tasks from inside ChatGPT. ServiceNow is giving up the front door to defend the point of execution. That choice works only if it gets paid for every action.&lt;/p&gt;
&lt;p&gt;ServiceNow is also going after the control layer above the front door. AI Control Tower discovers and governs the AI agents inside a company, and its net new ACV grew more than sixfold. The company did not disclose the amount. If that control layer bills per agent, whichever front door the agent uses, it becomes the strongest path to replacing seat revenue.&lt;/p&gt;
&lt;h2 id="a-grid-for-sorting-saas-stocks"&gt;A grid for sorting SaaS stocks
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;&lt;/th&gt;
 &lt;th&gt;An agent can route around it&lt;/th&gt;
 &lt;th&gt;An agent cannot route around it&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Priced per human seat&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Most at risk: screen-centric collaboration tools, simple workflow tools, support-seat tools&lt;/td&gt;
 &lt;td&gt;A one-to-two-year growth gap: systems of record such as ServiceNow, Salesforce, Workday and SAP. The question is whether they can re-price on usage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Priced on usage or data volume&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Price compression risk: generic APIs and easily swapped tools&lt;/td&gt;
 &lt;td&gt;Structural winners: hyperscalers, data, observability, networking and agent identity security. They rose sharply this year, so the price already carries expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;On October 25 we wrote that the unit of pricing is moving from seats to usage. &lt;a class="link" href="https://koreainvestinsights.com/post/who-pays-for-ai-demand-agents-capex-roi-2026-10-25/" &gt;KII, October 25&lt;/a&gt; We keep that view. ServiceNow adds one condition. The profit from usage pricing flows to execution points an agent cannot skip and to the front-door owners. Usage pricing that an agent can route around faces price compression.&lt;/p&gt;
&lt;p&gt;ServiceNow is moving from the top-right cell to the bottom-right cell of that grid. This quarter showed the gap that opens in the middle of the move.&lt;/p&gt;
&lt;h2 id="on-october-29-the-market-still-sold-with-a-blunt-knife"&gt;On October 29 the market still sold with a blunt knife
&lt;/h2&gt;&lt;p&gt;&lt;img alt="October 29 one-day moves for software stocks. Seat-priced apps: Atlassian -7.3%, Salesforce -4.6%, Workday -4.4%, HubSpot -3.8%. Usage-priced infrastructure: Datadog -4.5%, Snowflake -2.8%, Cloudflare -2.6%. Security: Palo Alto -3.5%, CrowdStrike -2.1%. Microsoft and Palantir -0.2%, Accenture +1.6%, IBM +2.0%, and the IGV software ETF -3.5%." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/servicenow-renewal-rate-seat-freeze-saas-agent-bypass-test-2026-10-30/oct29-software-reaction.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;Seat-priced apps fell 3.8% to 7.3%. They face the same test. Usage-priced infrastructure also fell 2.6% to 4.5%. Datadog was up 95% this year and Cloudflare 57%. A favorable structure does not protect a price like that from a guidance slowdown.&lt;/p&gt;
&lt;p&gt;Microsoft and Palantir barely moved. Accenture and IBM rose, which fits the view that putting AI into enterprise work still needs human services. In the same week F5 fell 20.4% on October 28 and Fortinet fell 20.1% on October 29, each after its own results. The IGV software ETF fell 3.5% and is up 2.2% this year. ServiceNow is down 28.4% this year, more than 30 points behind IGV.&lt;/p&gt;
&lt;p&gt;The market has not yet sorted stocks by whether an agent can route around them. It sold them together on a shared risk: slowing leading indicators. We expect that sorting to show up in prices from the next earnings season.&lt;/p&gt;
&lt;h2 id="what-would-change-our-view"&gt;What would change our view
&lt;/h2&gt;&lt;p&gt;Our emphasis is on the seat freeze. For seat-priced apps that an agent can route around, a lower multiple is structural. For systems of record that an agent cannot route around, the risk is a one-to-two-year growth gap; churn risk is small. These are analytical statements, not trading instructions.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Signs that usage revenue will cover seats&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;ServiceNow discloses total net new ACV alongside the AI share, and the total accelerates.&lt;/li&gt;
&lt;li&gt;It discloses revenue from usage above fixed credits, or the growth of credit commitments at renewal.&lt;/li&gt;
&lt;li&gt;Evidence emerges that Control Tower bills by the number of third-party agents or actions.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Signs that the seat freeze is turning into renewal cuts&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Seat-linked companies such as Workday or Salesforce mention seat cuts at renewal, or their NRR falls.&lt;/li&gt;
&lt;li&gt;ServiceNow&amp;rsquo;s Q4 cRPO comes in below 18.5%, or its 2027 subscription revenue guidance is below 17%.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;If Q4 cRPO is 19.5% or higher and 2027 subscription guidance is 19% or higher, we will lower the weight we put on the seat-freeze reading. ServiceNow reports Q4 at the end of January.&lt;/p&gt;
&lt;h2 id="a-question-for-readers"&gt;A question for readers
&lt;/h2&gt;&lt;p&gt;At your company, which software will an AI agent route around first? And which system will an agent never be able to skip? Tell us what you see at work in the comments. We will use your examples in our next sort of the sector.&lt;/p&gt;
&lt;h2 id="key-sources"&gt;Key sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000092/erq3fy26.htm" target="_blank" rel="noopener"
 &gt;ServiceNow Q3 2026 results (8-K Exhibit 99.1), October 28&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000096/now-20260930.htm" target="_blank" rel="noopener"
 &gt;ServiceNow Q3 2026 10-Q, renewal-rate definition&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000007/now-20251231.htm" target="_blank" rel="noopener"
 &gt;ServiceNow 2025 10-K, consumption pricing&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000072/erq2fy26.htm" target="_blank" rel="noopener"
 &gt;ServiceNow Q2 2026 results&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000054/erq1fy26.htm" target="_blank" rel="noopener"
 &gt;Q1 2026&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1373715/000137371526000005/erq4fy25.htm" target="_blank" rel="noopener"
 &gt;Q4 2025&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/10/29/servicenow-now-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;Motley Fool, Q3 2026 earnings call transcript, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.cnbc.com/2026/10/29/servicenow-now-earnings-q3-2026.html" target="_blank" rel="noopener"
 &gt;CNBC, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.constellationr.com/insights/news/servicenow-cfo-outlines-method-pro-plus-sku-pricing" target="_blank" rel="noopener"
 &gt;Constellation Research, Pro Plus pricing, March 2024&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Nasdaq historical quotes, October 29 regular-session closes&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Related facts: the founding partners of ServiceNow&amp;rsquo;s Control Tower Alliance include Anthropic, AWS, Cisco, NVIDIA and Okta. This quarter ServiceNow widened its ChatGPT plugin partnership with OpenAI.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Research and information only; not investment advice. Estimates and scenarios are analytical assumptions, not company guidance, trade instructions or personalized portfolio decisions. Figures are date-bound. AI tools assisted research and editing.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>Google's Ad Budget Is Moving to Meta. Six Straight Quarters of an 11-17 Point Growth Gap Is Not Luck</title><link>https://koreainvestinsights.com/post/meta-ad-edge-google-amazon-share-shift-2026-10-30/</link><pubDate>Fri, 30 Oct 2026 10:40:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/meta-ad-edge-google-amazon-share-shift-2026-10-30/</guid><description>&lt;img src="https://koreainvestinsights.com/" alt="Featured image of post Google's Ad Budget Is Moving to Meta. Six Straight Quarters of an 11-17 Point Growth Gap Is Not Luck" /&gt;&lt;p&gt;The three largest ad sellers reported third-quarter results in late October. Meta&amp;rsquo;s ad revenue &lt;strong&gt;grew 26% to $63.1 billion&lt;/strong&gt;. In the same quarter Google&amp;rsquo;s ad revenue grew 10.6% to $82.0 billion, and Amazon&amp;rsquo;s grew 10.9% to $19.6 billion. Together the three sold $164.7 billion of ads, 16.0% more than a year earlier, and &lt;strong&gt;Meta took 57% of that increase.&lt;/strong&gt; &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1326801/000162828026066700/meta-09302026xexhibit991.htm" target="_blank" rel="noopener"
 &gt;Meta release&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000080/googexhibit991q32026.htm" target="_blank" rel="noopener"
 &gt;Alphabet release&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001018724/000101872426000038/amzn-20260930xex991.htm" target="_blank" rel="noopener"
 &gt;Amazon release&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The market read Google&amp;rsquo;s miss as an AI search story and explained Amazon&amp;rsquo;s miss with the Prime Day calendar. Each explanation holds inside its own company. Put the three companies in one table, and a larger pattern appears.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our judgment: Meta is taking Google&amp;rsquo;s performance ad budget, and the shift has run for six quarters. It is structural. The cause lies more in the strength of Meta&amp;rsquo;s ad machine than in Google&amp;rsquo;s weakness. The evidence is that advertisers now pay 14% more per Meta ad than a year ago. The contest with Amazon is still open, and Q4 results are the first real test.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;This article is dated October 30, 2026. Figures come from each company&amp;rsquo;s SEC earnings releases and quarterly reports (10-Q). Share means share of the three companies&amp;rsquo; combined ad revenue. It is not share of the whole digital ad market.&lt;/p&gt;
&lt;h2 id="the-budget-shift-in-numbers"&gt;The budget shift in numbers
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Six straight quarters:&lt;/strong&gt; since Q2 2025, Meta&amp;rsquo;s ad revenue has grown 11 to 17 percentage points faster than Google&amp;rsquo;s every quarter. A one-quarter calendar quirk or currency swing cannot explain a run this long.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Share:&lt;/strong&gt; Meta&amp;rsquo;s share of the three companies&amp;rsquo; ad revenue was 33.2% in 2024, 35.1% in 2025 and 37.6% in 2026. The annual gain widened from 1.9 to 2.5 points.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Acceleration:&lt;/strong&gt; on a two-year compound basis, which removes last year&amp;rsquo;s base, Meta sped up from 24.4% in Q2 to 25.7% in Q3. Google slowed from 12.4% to 11.6%.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Amazon:&lt;/strong&gt; on the same basis Amazon&amp;rsquo;s share was 12.1% in both 2025 and 2026. Meta has not yet taken share from Amazon.&lt;/li&gt;
&lt;/ul&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Q2+Q3 share of combined ad revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2024&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2025&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;33.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;37.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Google&lt;/td&gt;
 &lt;td style="text-align: right"&gt;55.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;52.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;11.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;img alt="Share of the three companies’ combined ad revenue, Q2 and Q3 combined, 2024 to 2026. Meta rose from 33.2% to 35.1% and 37.6%. Google fell from 55.3% to 52.8% and 50.3%. Amazon went from 11.5% to 12.1% and stayed at 12.1% in 2026." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/meta-ad-edge-google-amazon-share-shift-2026-10-30/ad-share-shift.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;We combine Q2 and Q3 because of Prime Day. Amazon held Prime Day in July in 2024 and 2025, and in June in 2026. With the two quarters combined, the event falls in the same window every year.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Ad revenue growth gap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q2 25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q3 25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q4 25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q1 26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q2 26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q3 26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta minus Google&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.1pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.9pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.7pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.4pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+13.1pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.4pt&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta minus Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;−1.4pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.0pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.0pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.1pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.3pt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.0pt&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;img alt="Meta’s year-over-year ad revenue growth minus Google’s and Amazon’s, Q2 2025 to Q3 2026. The gap with Google was 11.1, 12.9, 10.7, 17.4, 13.1 and 15.4 points. The gap with Amazon was -1.4, 2.0, 1.0, 9.1, 1.3 and 15.0 points. Q1 2026 includes a currency tailwind of about 4 points for Meta, and the Q3 2026 Amazon gap includes Prime Day timing." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/meta-ad-edge-google-amazon-share-shift-2026-10-30/ad-growth-gap.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;The Q1 2026 gap includes a currency tailwind of about 4 points for Meta. Without it, the gap with Google still exceeds 10 points in every quarter. The 15.0-point gap with Amazon in Q3 2026 is mostly Prime Day timing.&lt;/p&gt;
&lt;h2 id="why-meta-is-stronger-advertisers-pay-for-results"&gt;Why Meta is stronger: advertisers pay for results
&lt;/h2&gt;&lt;p&gt;In Q3, Meta&amp;rsquo;s ad impressions grew 10% and its average price per ad rose 14%. For prices to rise while supply also grows, advertiser demand must grow faster than supply. CFO Susan Li attributed the price increase to ad performance gains and to better macro conditions than a year earlier. In the same quarter, Google Search&amp;rsquo;s cost per click rose 4%. &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/content_files/META-Q3-2026-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;Meta call transcript&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000084/goog-20260930.htm" target="_blank" rel="noopener"
 &gt;Alphabet 10-Q&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;img alt="Year-over-year change in Meta’s ad impressions and average price per ad. From Q3 2025 to Q3 2026, impressions grew 14, 18, 19, 14 and 10%, and price per ad rose 10, 6, 12, 12 and 14%. In Q3 2026 price growth passed impression growth." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/meta-ad-edge-google-amazon-share-shift-2026-10-30/meta-price-vs-impressions.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;Meta can raise prices because what advertisers buy from it is closer to a conversion than to an impression. On the Q3 call, Meta put numbers on the effect of its ad ranking models:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;GEM expansion:&lt;/strong&gt; conversions rose 1.7% for off-site conversion campaigns and 9.6% for app ads.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Andromeda improvements:&lt;/strong&gt; conversions on Facebook rose 1.4%.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Partnership ads:&lt;/strong&gt; grew more than 97% year over year and passed a $15 billion annual revenue run rate.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Advertisers split budgets by measured conversion results. Meta&amp;rsquo;s 10-Q names online commerce as the largest contributor to its ad revenue growth. Google said retail and finance led its Search revenue growth. The same advertisers compare the two companies on the same performance metrics, and we judge that the result shows up as the growth gap. None of the three companies said budgets are moving, so this link is our inference. &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001326801/000162828026066773/meta-20260930.htm" target="_blank" rel="noopener"
 &gt;Meta 10-Q&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Meta accelerated in its largest market.&lt;/strong&gt; By advertiser location, Meta&amp;rsquo;s US and Canada revenue grew 33.0% in Q3, up from 29.3% in Q2. The region is 41% of Meta&amp;rsquo;s total revenue. Meta said strong impression growth on lower-monetizing surfaces and regions partly offset the price gain. The regional mix pushed the average price down, and the average price still rose 14%. That supports the reading that the same advertisers are paying more.&lt;/p&gt;
&lt;h2 id="googles-searches-grew-faster-than-its-paid-clicks"&gt;Google&amp;rsquo;s searches grew faster than its paid clicks
&lt;/h2&gt;&lt;p&gt;CEO Sundar Pichai said Q3 search query growth was the highest in the last five years. Yet Search paid-click growth fell from 13% in Q2 to 8% in Q3. Search revenue grew 12.0%. &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/10/29/alphabet-googl-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;Alphabet call transcript (Motley Fool)&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Chief Business Officer Philipp Schindler gave several causes for the slowdown. Currency was the single largest. The EU began charging duties on small parcels on July 1, and some Asia-Pacific retail advertisers pulled back. Media spending that had bunched around the June World Cup normalized in Q3. He also said AI Mode queries grew strongly while AI Mode monetized at a lower rate.&lt;/p&gt;
&lt;p&gt;Currency, the World Cup and the Asia-Pacific advertisers are temporary. Currency changes revenue but not the number of clicks. What remains is that each search query now produces fewer ad clicks. The effect is still small. But it comes from a structure in which Google must receive a query before it can sell an ad. The more users get answers straight from AI, the fewer ads there are to click.&lt;/p&gt;
&lt;p&gt;Meta&amp;rsquo;s feed does not have this problem. Meta creates its own ad inventory, and it uses AI to rank the ads in that inventory. The same technology is cutting clicks at Google and lifting conversions at Meta. Still, we treat Google&amp;rsquo;s AI Mode gap as a secondary cause. The growth gap dates back to Q2 2025, long before this quarter, when Google first cited AI Mode&amp;rsquo;s lower monetization.&lt;/p&gt;
&lt;h2 id="meta-has-not-beaten-amazon-yet"&gt;Meta has not beaten Amazon yet
&lt;/h2&gt;&lt;p&gt;Amazon&amp;rsquo;s ad growth fell from 26% in Q2 to 12% in Q3, both excluding currency. CEO Andy Jassy said moving Prime Day into Q2 in most countries had an 850-basis-point effect on ad growth. A change to Amazon&amp;rsquo;s search marketing program in Q3 2025 also moved into the base this quarter. &lt;a class="link" href="https://www.investing.com/news/transcripts/earnings-call-transcript-amazon-q3-2026-misses-eps-as-shares-fall-after-hours-93CH-4985489" target="_blank" rel="noopener"
 &gt;Amazon call transcript (Investing.com)&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Combining Q2 and Q3 removes the calendar effect and gives 18.6% growth excluding currency. Meta grew faster than Amazon, but it has not yet taken Amazon&amp;rsquo;s share. Amazon measures whether people who saw an ad actually bought, using its own store data. Nothing in these results shows that this advantage has weakened.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Q4 is the first test.&lt;/strong&gt; Amazon guided Q4 ad growth to the high teens. Meta&amp;rsquo;s Q4 revenue guidance of $73-76 billion implies 21.9-26.9% growth. If both land on guidance, a 5-7 point gap appears for the first time in a quarter with no calendar distortion. If the gap is 2 points or less, we will conclude that Meta has no edge over Amazon.&lt;/p&gt;
&lt;h2 id="where-meta-does-not-win"&gt;Where Meta does not win
&lt;/h2&gt;&lt;p&gt;Meta&amp;rsquo;s edge is limited to the performance budgets of online commerce and app advertisers. Meta is not taking Google&amp;rsquo;s search queries.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Search intent:&lt;/strong&gt; Google is still strongest for ads shown to people who search in order to buy.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;TV:&lt;/strong&gt; in Q3, living-room viewing accounted for more than a third of YouTube ad revenue. Meta has no such surface.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Purchase data:&lt;/strong&gt; Amazon is strongest for ads measured all the way to a purchase in its own store.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="what-could-break-this-judgment"&gt;What could break this judgment
&lt;/h2&gt;&lt;p&gt;&lt;strong&gt;Meta&amp;rsquo;s growth is now price-led.&lt;/strong&gt; Impression growth fell from 19% in Q1 to 14% in Q2 and 10% in Q3. Daily active people grew 3% to 3.64 billion. Prices react first to advertiser returns and the economy, and Meta included better macro conditions among the reasons for its price gain. If the economy turns, price is where growth falls first.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Asia-Pacific advertisers are wobbling.&lt;/strong&gt; By advertiser location, Asia-Pacific is 26.3% of Meta&amp;rsquo;s revenue. Its growth slowed from 25.0% in Q2 to 18.8% in Q3. We estimate these are the same advertisers Google linked to the EU small-parcel duty. Meta did not discuss this on its call.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Ad cash is flowing into AI capex.&lt;/strong&gt; Meta&amp;rsquo;s Q3 capital expenditure was $34.9 billion, and its free cash flow was negative $2.1 billion. Its 2026 capex guidance is $135-145 billion. In late October Meta raised outside capital with $25 billion of common stock (through forward sales) and mandatory convertible preferred stock. The strength of the ad machine and the cash left for shareholders are separate questions. &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1326801/000119312526436410/d94886d424b5.htm" target="_blank" rel="noopener"
 &gt;Meta prospectus (424B5)&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="the-numbers-that-would-change-our-view"&gt;The numbers that would change our view
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Q4 metric&lt;/th&gt;
 &lt;th&gt;Judgment holds&lt;/th&gt;
 &lt;th&gt;Judgment changes&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta minus Google ad growth gap&lt;/td&gt;
 &lt;td&gt;10 points or more&lt;/td&gt;
 &lt;td&gt;Below 8 points: we lower the edge over Google&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Google paid-click growth&lt;/td&gt;
 &lt;td&gt;Below query growth&lt;/td&gt;
 &lt;td&gt;Catches up with query growth: AI Mode gap closed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta price-per-ad growth&lt;/td&gt;
 &lt;td&gt;At or above impression growth&lt;/td&gt;
 &lt;td&gt;Falls below impression growth: performance edge weakening&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta minus Amazon ad growth gap&lt;/td&gt;
 &lt;td&gt;5-7 points&lt;/td&gt;
 &lt;td&gt;2 points or less: no edge over Amazon&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta revenue growth&lt;/td&gt;
 &lt;td&gt;21.9-26.9% (guidance)&lt;/td&gt;
 &lt;td&gt;Below 21.9%: read as weaker total demand, not share shift&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta Asia-Pacific advertiser revenue growth&lt;/td&gt;
 &lt;td&gt;Recovers&lt;/td&gt;
 &lt;td&gt;Stays below 18%: the EU duty shock is lasting&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The three companies report Q4 in late January or early February 2027. The dates are not yet set.&lt;/p&gt;
&lt;h2 id="limits"&gt;Limits
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Revenue basis:&lt;/strong&gt; Google&amp;rsquo;s ad revenue includes traffic acquisition costs paid to partners ($16.4 billion in Q3). These costs grew 10.0%, close to ad revenue, so they change the growth gap little.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Scope:&lt;/strong&gt; we did not compare TikTok, Microsoft, AppLovin, Reddit, Walmart or ChatGPT ads.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Inference:&lt;/strong&gt; the budget shift toward Meta is our inference from public results. We did not see advertisers&amp;rsquo; actual spending split by platform.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Call remarks:&lt;/strong&gt; Meta remarks come from the company&amp;rsquo;s official transcript; Google and Amazon remarks come from third-party transcripts. Jassy&amp;rsquo;s 850 basis points is ambiguously worded, so we estimated Amazon&amp;rsquo;s underlying growth from combined Q2 and Q3.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="main-sources"&gt;Main sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Meta: &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1326801/000162828026066700/meta-09302026xexhibit991.htm" target="_blank" rel="noopener"
 &gt;Q3 release&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001326801/000162828026066773/meta-20260930.htm" target="_blank" rel="noopener"
 &gt;10-Q&lt;/a&gt;, &lt;a class="link" href="https://s21.q4cdn.com/399680738/files/content_files/META-Q3-2026-Earnings-Call-Transcript.pdf" target="_blank" rel="noopener"
 &gt;official call transcript&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1326801/000119312526436410/d94886d424b5.htm" target="_blank" rel="noopener"
 &gt;common stock prospectus&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Alphabet: &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000080/googexhibit991q32026.htm" target="_blank" rel="noopener"
 &gt;Q3 release&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000084/goog-20260930.htm" target="_blank" rel="noopener"
 &gt;10-Q&lt;/a&gt;, &lt;a class="link" href="https://www.fool.com/earnings/call-transcripts/2026/10/29/alphabet-googl-q3-2026-earnings-call-transcript/" target="_blank" rel="noopener"
 &gt;call transcript (Motley Fool)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Amazon: &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/0001018724/000101872426000038/amzn-20260930xex991.htm" target="_blank" rel="noopener"
 &gt;Q3 release&lt;/a&gt;, &lt;a class="link" href="https://www.investing.com/news/transcripts/earnings-call-transcript-amazon-q3-2026-misses-eps-as-shares-fall-after-hours-93CH-4985489" target="_blank" rel="noopener"
 &gt;call transcript (Investing.com)&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Prior quarters: the three companies&amp;rsquo; earnings releases for Q2 2024 through Q2 2026 (SEC EDGAR)&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Is Meta&amp;rsquo;s 14% price-per-ad gain proof of a better ad machine, or a peak lifted by a macro tailwind? Tell us which number you expect to see in Q4.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;For research and information only; not investment advice. Judgments and test thresholds are analytical assumptions, not company guidance, trading instructions or individual portfolio decisions. Figures are dated. AI tools were used for research and editing. Amazon and Alphabet are investors in Anthropic, which makes one of those tools.&lt;/p&gt;</description></item><item><title>Dukovany's Builder Talks Broke Over Who Bears First-of-a-Kind Risk. The Market Just Priced Team Korea's Hidden Cost.</title><link>https://koreainvestinsights.com/post/dukovany-construction-talks-collapse-risk-allocation-team-korea-2026-10-29/</link><pubDate>Thu, 29 Oct 2026 12:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/dukovany-construction-talks-collapse-risk-allocation-team-korea-2026-10-29/</guid><description>&lt;img src="https://koreainvestinsights.com/" alt="Featured image of post Dukovany's Builder Talks Broke Over Who Bears First-of-a-Kind Risk. The Market Just Priced Team Korea's Hidden Cost." /&gt;&lt;p&gt;On October 29, Korea Hydro &amp;amp; Nuclear Power (KHNP) ended talks on the construction contract for Dukovany units 5 and 6 in the Czech Republic. Its counterparty was a consortium led by Daewoo E&amp;amp;C, with Doosan Enerbility as a member. KHNP will now pick a new builder by competitive tender. Its tentative plan is a bidder briefing in January 2027, a tender notice in April and a contract in July, and it keeps January 2028 as the start of site preparation. The detailed procedure and dates are not final. &lt;a class="link" href="https://www.fnnews.com/news/202610291003332769" target="_blank" rel="noopener"
 &gt;Financial News&lt;/a&gt;, &lt;a class="link" href="https://www.ajunews.com/view/20261029093930495" target="_blank" rel="noopener"
 &gt;Aju Business Daily&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The main contract between KHNP and the Czech owner, EDU II, stands. So do Doosan Enerbility&amp;rsquo;s main-equipment contract, worth about 5.6 trillion won, and KEPCO E&amp;amp;C&amp;rsquo;s design contract. What broke is one contract: the construction of the plant buildings and the installation of its equipment. Even so, by midday &lt;strong&gt;Daewoo E&amp;amp;C was down 20% and Doosan Enerbility was down 12%.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Our judgment: the variable that broke the talks was who carries the quantity risk on the first-ever APR1000, more than the price level. The roughly 5.2 trillion won of market value that four stocks lost by the regular-session close is 9 to 17 times the profit this construction work could earn. The stocks priced a discount on the Team Korea story, that Korea builds reactors cheaply and on time, more than the contract&amp;rsquo;s cash flows.&lt;/strong&gt; That story rested on affiliates and partners absorbing cost risk together. A private builder refused fixed-price risk on a first-of-a-kind reactor, and the premise became visible.&lt;/p&gt;
&lt;p&gt;In July we wrote that the Czech project&amp;rsquo;s contract risk had fallen but its margin risk remained. &lt;a class="link" href="https://koreainvestinsights.com/post/dukovany-westinghouse-ip-settlement-korea-nuclear-margin-risk-2026-07-04/" &gt;KII, July 4&lt;/a&gt; Today that margin risk broke the construction contract.&lt;/p&gt;
&lt;p&gt;This article is dated October 29, 2026. Korean market figures in the tables and charts are intraday values at 12:14 p.m. KST. Regular-session closing values are stated separately in the text (updated 6 p.m. KST, October 29).&lt;/p&gt;
&lt;h2 id="what-stands-and-what-broke"&gt;What stands and what broke
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;What stands:&lt;/strong&gt; KHNP&amp;rsquo;s main contract with EDU II (signed June 4, 2025), Doosan Enerbility&amp;rsquo;s main equipment at about 5.6 trillion won, and KEPCO E&amp;amp;C&amp;rsquo;s design work.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;What broke:&lt;/strong&gt; construction talks with the Daewoo E&amp;amp;C–Doosan Enerbility consortium, after about 15 months, 78 working meetings, 11 dispute-adjustment committee sessions and two meetings between the chief executives.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The dispute:&lt;/strong&gt; KHNP wanted a lump-sum fixed price. The consortium wanted to settle quantity increases after detailed design.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The stocks:&lt;/strong&gt; four companies lost about 7.5 trillion won of market value intraday and about 5.2 trillion won by the regular-session close. We estimate the consortium&amp;rsquo;s total construction profit at 0.3–0.6 trillion won.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The ecosystem:&lt;/strong&gt; a cost fight that erupted after construction at Barakah has erupted before construction in the Czech Republic. US–Korea nuclear cooperation has given private builders an alternative.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The next test:&lt;/strong&gt; who attends the January 2027 briefing, and what risk clauses the April tender carries.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="fifteen-months-of-talks"&gt;Fifteen months of talks
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Date&lt;/th&gt;
 &lt;th&gt;Event&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2018&lt;/td&gt;
 &lt;td&gt;KHNP forms Team Korea; Daewoo E&amp;amp;C is lead builder&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2022&lt;/td&gt;
 &lt;td&gt;First bid; by KHNP&amp;rsquo;s account, the parties agree price and lump-sum terms for one unit&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2023&lt;/td&gt;
 &lt;td&gt;Revised bid; by KHNP&amp;rsquo;s account, they agree again on a one-unit basis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;January 2024&lt;/td&gt;
 &lt;td&gt;The consortium submits construction unit prices&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;April 2024&lt;/td&gt;
 &lt;td&gt;Final two-unit bid; by KHNP&amp;rsquo;s account, with no agreement, KHNP bids its own estimated price&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 2024&lt;/td&gt;
 &lt;td&gt;The Czech government names KHNP preferred bidder&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;January 2025&lt;/td&gt;
 &lt;td&gt;KHNP and KEPCO settle the intellectual-property dispute with Westinghouse&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;June 4, 2025&lt;/td&gt;
 &lt;td&gt;KHNP and EDU II sign the main contract&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;December 2025&lt;/td&gt;
 &lt;td&gt;KEPCO E&amp;amp;C signs the main design contract (about 1.25 trillion won) and Doosan Enerbility the main-equipment contract (about 5.64 trillion won); Doosan&amp;rsquo;s has no advance payment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;September 30, 2026&lt;/td&gt;
 &lt;td&gt;US–Korea nuclear framework: up to $120 billion of Korean investment to help finance eight large US reactors (six AP1000s, two APR1400s), non-binding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;End of September 2026&lt;/td&gt;
 &lt;td&gt;By KHNP&amp;rsquo;s account, EDU II&amp;rsquo;s deadline for a construction contract passes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Early October 2026&lt;/td&gt;
 &lt;td&gt;KHNP tells a Korea–Czech implementation council it will decide by the end of October&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;October 7, 2026&lt;/td&gt;
 &lt;td&gt;KHNP presents changed contract terms (the consortium&amp;rsquo;s account)&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;October 29, 2026&lt;/td&gt;
 &lt;td&gt;The consortium submits a final offer (its account); KHNP ends the talks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Sources: Korea Economic Daily, Financial News, Aju Business Daily, Etoday, MTN, DART filings and Brookfield, compiled by KII.&lt;/p&gt;
&lt;h2 id="the-talks-broke-on-risk-allocation"&gt;The talks broke on risk allocation
&lt;/h2&gt;&lt;p&gt;&lt;strong&gt;Fixed-price risk flows down.&lt;/strong&gt; As main contractor, KHNP says it will manage any extra cost within its main contract price with EDU II. A KHNP official said Doosan Enerbility (main equipment) and KEPCO E&amp;amp;C (design) both signed lump-sum contracts, and that settling only construction on quantities would leave KHNP carrying the whole risk of cost growth. &lt;a class="link" href="https://hankyung.com/article/202610292607i" target="_blank" rel="noopener"
 &gt;Korea Economic Daily&lt;/a&gt; KHNP&amp;rsquo;s own risk closes only if construction is bound the same way. Daewoo E&amp;amp;C calls the construction contract a subcontract under KHNP. &lt;a class="link" href="https://www.etoday.co.kr/news/view/2641662" target="_blank" rel="noopener"
 &gt;Etoday&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Quantity risk collects in construction.&lt;/strong&gt; The APR1000 has never been built anywhere. KHNP&amp;rsquo;s compromise set the point for fixing the total at approval of the basic design, so the basic design is not yet approved. EBN, citing industry sources, reported that Czech licensing could change design and construction conditions. &lt;a class="link" href="https://www.ebn.co.kr/news/articleView.html?idxno=1729540" target="_blank" rel="noopener"
 &gt;EBN&lt;/a&gt; Design changes feed into quantities, schedules, quality inspections and construction time. The consortium argued that a builder without design authority cannot take that risk without price adjustment.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The price gap narrowed; the risk gap did not.&lt;/strong&gt; KHNP&amp;rsquo;s proposal fixes the total construction price on quantities at the basic-design stage. The builders&amp;rsquo; proposal settles increases against quantities at the detailed-design stage. &lt;a class="link" href="https://www.ajunews.com/view/20261029093930495" target="_blank" rel="noopener"
 &gt;Aju Business Daily&lt;/a&gt; KHNP says it offered an increase that assumed higher quantities, and a compromise that would move the point at which the total is fixed to approval of the basic design. The consortium says that on October 29 it submitted a final revised offer that accepted January 2024 unit prices and absorbed the losses. &lt;a class="link" href="https://www.etoday.co.kr/news/view/2641662" target="_blank" rel="noopener"
 &gt;Etoday&lt;/a&gt; If both accounts are right, the two sides moved toward each other on price. They still did not agree, so the remaining gap is who pays for design changes and what work the builder must do.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Claim&lt;/th&gt;
 &lt;th&gt;Who says so&lt;/th&gt;
 &lt;th&gt;Status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;The dispute is lump-sum (KHNP) versus post-design quantity settlement (consortium)&lt;/td&gt;
 &lt;td&gt;Both sides&lt;/td&gt;
 &lt;td&gt;Accounts match&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KHNP offered an increase and a later point for fixing the total&lt;/td&gt;
 &lt;td&gt;KHNP&lt;/td&gt;
 &lt;td&gt;KHNP&amp;rsquo;s account&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The consortium offered to accept January 2024 unit prices&lt;/td&gt;
 &lt;td&gt;Consortium&lt;/td&gt;
 &lt;td&gt;Consortium&amp;rsquo;s account&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KHNP changed the contract terms on October 7 and rejected a formal request to extend the deadline to December 31&lt;/td&gt;
 &lt;td&gt;Consortium&lt;/td&gt;
 &lt;td&gt;Consortium&amp;rsquo;s account&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The consortium refused to install reactor internals and the turbine generator&lt;/td&gt;
 &lt;td&gt;KHNP&lt;/td&gt;
 &lt;td&gt;KHNP&amp;rsquo;s account&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KHNP asked for a detailed basis for direct and indirect costs, and the builders did not provide it&lt;/td&gt;
 &lt;td&gt;KHNP&lt;/td&gt;
 &lt;td&gt;KHNP&amp;rsquo;s account&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The parties agreed one-unit terms in 2022 and 2023, and the builders changed position in the 2024 two-unit bid&lt;/td&gt;
 &lt;td&gt;KHNP&lt;/td&gt;
 &lt;td&gt;KHNP&amp;rsquo;s account&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KHNP sought a cut of at least 10%, over 1 trillion won, from the bid&lt;/td&gt;
 &lt;td&gt;JoongAng Ilbo (March) and EBN (October 28), citing industry sources&lt;/td&gt;
 &lt;td&gt;Not confirmed by KHNP; KHNP says unilateral cuts did not cause the breakdown&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Westinghouse settlement costs are funded by reallocating within the existing budget, so they do not pressure construction costs&lt;/td&gt;
 &lt;td&gt;KHNP&lt;/td&gt;
 &lt;td&gt;KHNP&amp;rsquo;s explanation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The Dukovany contract is 2.62 times Barakah&amp;rsquo;s two units (1.68 times after inflation) and 2.4 times Shin Hanul 3 and 4&lt;/td&gt;
 &lt;td&gt;KHNP&lt;/td&gt;
 &lt;td&gt;KHNP&amp;rsquo;s explanation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;strong&gt;The gap goes back to the 2024 bid.&lt;/strong&gt; By KHNP&amp;rsquo;s account, the parties agreed one-unit terms in 2022 and 2023. In the final two-unit bid in 2024, the builders changed their position on unit prices and quantity settlement, so there was no agreement, and KHNP bid a price it estimated itself from domestic plants, Barakah and Czech conditions. &lt;a class="link" href="https://news.mtn.co.kr/article/2026102909393462147" target="_blank" rel="noopener"
 &gt;MTN&lt;/a&gt;, &lt;a class="link" href="https://www.fnnews.com/news/202610291003332769" target="_blank" rel="noopener"
 &gt;Financial News&lt;/a&gt; We have not seen the consortium&amp;rsquo;s account of this history. On either account, fifteen months of talks after the award were an attempt to close a gap left at the bid.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;New costs arrived inside the same contract price.&lt;/strong&gt; KHNP says it funds its Westinghouse settlement payments by reallocating within the existing project budget, and that these payments do not pressure construction costs. It also argues that construction pay is higher than on earlier projects, because the Dukovany contract is 2.62 times Barakah&amp;rsquo;s two units (1.68 times after inflation) and 2.4 times Shin Hanul 3 and 4. &lt;a class="link" href="https://hankyung.com/article/202610292607i" target="_blank" rel="noopener"
 &gt;Korea Economic Daily&lt;/a&gt; Still, when a fixed-price contract absorbs a new cost, the slack in other lines shrinks. The JoongAng Ilbo in March and EBN on October 28 reported, citing industry sources, that KHNP had asked for at least 10% off the bid price. KHNP has not confirmed that figure, and it says unilateral cuts did not cause the breakdown. &lt;a class="link" href="https://www.ebn.co.kr/news/articleView.html?idxno=1729540" target="_blank" rel="noopener"
 &gt;EBN&lt;/a&gt;, &lt;a class="link" href="https://www.businessplus.kr/news/articleView.html?idxno=118919" target="_blank" rel="noopener"
 &gt;Business Plus&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;KHNP has already started to carry part of the risk itself.&lt;/strong&gt; Until a new builder is chosen, KHNP will do essential work such as design and licensing itself, and it will tender ancillary-facility and heavy-haul route design separately. It says any extra cost will be managed within the main contract price. &lt;a class="link" href="https://www.fnnews.com/news/202610291003332769" target="_blank" rel="noopener"
 &gt;Financial News&lt;/a&gt; There are also concerns that changing builders could delay the project by more than a year and raise costs through re-tendering, start-up spending and inflation. MTN reported that KHNP&amp;rsquo;s briefing material mentioned these concerns; Financial News attributed the delay concern to the existing builders. KHNP says it will run design and licensing in parallel to keep its targets: site preparation in January 2028 and main construction in November 2029. &lt;a class="link" href="https://news.mtn.co.kr/article/2026102909393462147" target="_blank" rel="noopener"
 &gt;MTN&lt;/a&gt; Daewoo E&amp;amp;C and Doosan Enerbility may bid again if they meet the qualifications.&lt;/p&gt;
&lt;p&gt;The re-tender will settle what kind of dispute this was. If large builders bid on lump-sum terms, the problem was the Daewoo consortium&amp;rsquo;s risk aversion. If bidders refuse the same terms, the problem is the structure of a first-of-a-kind build, and KHNP must change its terms.&lt;/p&gt;
&lt;h2 id="five-trillion-won-lost-against-half-a-trillion-at-stake"&gt;Five trillion won lost against half a trillion at stake
&lt;/h2&gt;&lt;p&gt;&lt;img alt="Market value lost intraday on October 29, compared with our estimate of the consortium’s total construction profit. Doosan Enerbility lost about 4.9 trillion won, Daewoo E&amp;C about 1.2 trillion, Hyundai E&amp;C about 1.0 trillion and KEPCO E&amp;C about 0.4 trillion. The estimated construction profit range for the whole consortium is 0.3 to 0.6 trillion won." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/dukovany-construction-talks-collapse-risk-allocation-team-korea-2026-10-29/value-lost-vs-profit.en.png"&gt;
&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Price (won)&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Change&lt;/th&gt;
 &lt;th style="text-align: right"&gt;vs KOSPI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Market value lost&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;11,410&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-20.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-17.5pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 1.18tn won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Doosan Enerbility&lt;/td&gt;
 &lt;td style="text-align: right"&gt;53,700&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-10.0pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 4.93tn won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;86,700&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.6pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 0.98tn won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;86,700&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7.3pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 0.36tn won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO KPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40,800&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.8pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;small&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO&lt;/td&gt;
 &lt;td style="text-align: right"&gt;28,100&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.7pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;small&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung C&amp;amp;T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;320,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.4pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;small&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Source: Naver Finance, intraday at 12:14 p.m. KST on October 29. The KOSPI was down 2.59% at the same time. Market value lost is the fall from the previous close times shares outstanding.&lt;/p&gt;
&lt;p&gt;Together the four stocks lost about 7.5 trillion won of market value intraday. Net of the KOSPI&amp;rsquo;s fall, the loss is still about 5.9 trillion won.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Close update:&lt;/strong&gt; the losses narrowed in the afternoon. At the regular-session close, Daewoo E&amp;amp;C was down 20.0%, Doosan Enerbility 8.1%, KEPCO E&amp;amp;C 7.6% and Hyundai E&amp;amp;C 5.1%, and the KOSPI was down 1.24%. The four stocks lost about 5.2 trillion won of market value, or about 4.5 trillion won net of the KOSPI&amp;rsquo;s fall. By company, Doosan Enerbility lost about 3.2 trillion won, Daewoo E&amp;amp;C about 1.2 trillion won, Hyundai E&amp;amp;C about 0.5 trillion won and KEPCO E&amp;amp;C about 0.3 trillion won.&lt;/p&gt;
&lt;p&gt;The profit at stake in this contract is far smaller. At signing, the Czech side put the cost of the two Dukovany units at about CZK 407 billion in current prices. &lt;a class="link" href="https://oenergetice.cz/en/nuclear-power/south-koreas-khnp-to-build-at-czechias-dukovany-plant-contract-signed-after-court-ruling" target="_blank" rel="noopener"
 &gt;oEnergetice and ČTK&lt;/a&gt; At the October 29 exchange rate of about 61 won per koruna, that is about 25 trillion won; Korean media cite 25–28 trillion won, depending on when they convert. The construction value is not public. The JoongAng Ilbo estimated it in March at 10-plus trillion won, and the reports that a 10% cut equals more than 1 trillion won point to the same scale. &lt;a class="link" href="https://v.daum.net/v/20260324050154461" target="_blank" rel="noopener"
 &gt;JoongAng Ilbo via Daum&lt;/a&gt; At about 10 trillion won of construction and a 3–6% operating margin, the whole consortium would earn 0.3–0.6 trillion won over about a decade. The margin and the consortium shares are not public, so this range is our assumption.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Daewoo E&amp;amp;C&amp;rsquo;s one-day loss alone is about twice the top of that range, and Daewoo would earn only part of it.&lt;/strong&gt; Doosan Enerbility kept its main-equipment contract and still lost about 4.9 trillion won intraday and about 3.2 trillion won by the close. The stocks therefore repriced an option and a story more than this contract&amp;rsquo;s cash flows. Daewoo E&amp;amp;C lost its option on an overseas nuclear EPC reference. Doosan Enerbility lost some of the story that Team Korea moves as one.&lt;/p&gt;
&lt;p&gt;Hyundai E&amp;amp;C fell 9% intraday and closed 5.1% lower even though the market is watching whether it and Samsung C&amp;amp;T will bid. &lt;a class="link" href="https://news.mtn.co.kr/article/2026102909393462147" target="_blank" rel="noopener"
 &gt;MTN&lt;/a&gt; The market did not read this as one company&amp;rsquo;s problem. The risk of building a first-of-a-kind reactor at a fixed price is the same whoever builds it, and Hyundai E&amp;amp;C faces the same question on US AP1000 work. &lt;a class="link" href="https://koreainvestinsights.com/post/us-nuclear-ap1000-apr1400-korea-epc-scenario-2026-08-30/" &gt;KII, August 30&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="todays-drop-hit-stocks-that-had-already-halved"&gt;Today&amp;rsquo;s drop hit stocks that had already halved
&lt;/h2&gt;&lt;p&gt;&lt;img alt="Share price paths of Daewoo E&amp;C, Doosan Enerbility, KEPCO E&amp;C, Hyundai E&amp;C and the KOSPI, rebased to 100 on January 2, 2026 (log scale). Daewoo E&amp;C rose nearly tenfold by late April and stood at 304 on October 29. Doosan Enerbility is at 72, KEPCO E&amp;C at 97, Hyundai E&amp;C at 126 and the KOSPI at 136." loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/post-assets/dukovany-construction-talks-collapse-risk-allocation-team-korea-2026-10-29/nuclear-price-paths.en.png"&gt;
&lt;/p&gt;
&lt;p&gt;The nuclear rerating peaked in the spring. At the October 28 close, Daewoo E&amp;amp;C was 62% below its peak, Doosan Enerbility 55%, KEPCO E&amp;amp;C 50% and Hyundai E&amp;amp;C 49%. The KOSPI was 34% below its June 22 peak. From September 21 to October 28 the four stocks fell 23–28%, while the KOSPI fell 14%. After the US–Korea nuclear framework appeared, nuclear stocks were weaker than the market. &lt;a class="link" href="https://koreainvestinsights.com/post/korea-us-nuclear-westinghouse-cameco-doosan-2026-09-22/" &gt;KII, September 22&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Even after today, Daewoo E&amp;amp;C trades at about 2.3 times its January 30 close of 4,970 won. If the nuclear EPC option drove a large part of that rise, its discount may not have finished in one day. Doosan Enerbility trades at about 200 times earnings at the close on Naver Finance data. A stock whose price rests more on a story than on earnings reacts hard to a flaw in the story.&lt;/p&gt;
&lt;h2 id="what-team-koreas-cost-structure-revealed"&gt;What Team Korea&amp;rsquo;s cost structure revealed
&lt;/h2&gt;&lt;p&gt;&lt;strong&gt;Low prices relied on shared risk.&lt;/strong&gt; Team Korea&amp;rsquo;s strengths were low build costs and on-time delivery. Those results relied on state companies, affiliates and partners absorbing cost risk among themselves. When a private builder refuses fixed-price risk on a first-of-a-kind reactor, that risk returns to KHNP. From the next overseas tender, a risk premium is likely to enter the price. This is our interpretation.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The Barakah cost fight moved to before construction.&lt;/strong&gt; In May 2025 KHNP filed for arbitration at the London Court of International Arbitration against KEPCO, claiming about $1.1 billion under the operating support services contract for the Barakah plant in the United Arab Emirates. In June 2026 the two companies agreed to move the case to the Korean Commercial Arbitration Board, where it continues. &lt;a class="link" href="https://dart.fss.or.kr/dsaf001/main.do?rcpNo=20250507800786" target="_blank" rel="noopener"
 &gt;KEPCO filing, May 2025&lt;/a&gt;, &lt;a class="link" href="https://dart.fss.or.kr/dsaf001/main.do?rcpNo=20260623800004" target="_blank" rel="noopener"
 &gt;KEPCO filing, June 2026&lt;/a&gt; Korean media describe the dispute as a settlement of additional construction costs that grew during the Barakah build. &lt;a class="link" href="https://www.etoday.co.kr/news/view/2545376" target="_blank" rel="noopener"
 &gt;Etoday, January 2026&lt;/a&gt; At Barakah, the fight over who pays for overruns came after the build. In the Czech Republic, the same fight came before it. Finding the risk early is better than a dispute later, but it costs schedule and the owner&amp;rsquo;s trust.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Bargaining power is moving to private builders.&lt;/strong&gt; The US–Korea nuclear framework is not yet binding, but it has opened Westinghouse projects as an alternative for Korean builders and equipment makers. KHNP&amp;rsquo;s new entry into Europe, by contrast, is reportedly limited by its Westinghouse settlement. The reported terms leave the European Union outside the Czech Republic, North America, the United Kingdom, Japan and Ukraine to Westinghouse. &lt;a class="link" href="https://www.hankookilbo.com/News/Read/A2025081917210002321" target="_blank" rel="noopener"
 &gt;Hankook Ilbo&lt;/a&gt; KHNP has declined to confirm the terms, citing confidentiality. If the reported terms are right, the Czech Republic is KHNP&amp;rsquo;s only open European reference. In our reading, that asymmetry gives KHNP a reason to absorb more cost in the Czech project. It points the same way as the bargaining-power case we made for Hyundai E&amp;amp;C in June. &lt;a class="link" href="https://koreainvestinsights.com/post/hyundai-ec-woojin-us-nuclear-expansion-bargaining-power-2026-06-13/" &gt;KII, June 13&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The Czech owner&amp;rsquo;s trust is the next variable.&lt;/strong&gt; By KHNP&amp;rsquo;s account, EDU II wanted a construction contract by the end of September. A new contract will come in July 2027 at the earliest. MTN reported that the Czech government and the owner had voiced concern about the delay. As of midday Korean time, we have not found an official response from EDU II, ČEZ or the Czech government; it is still early morning in Prague. The main contract includes an option for two units at Temelin. In January, Czech Industry and Trade Minister Karel Havlíček said the government will decide on that construction in 2027. He said the option does not automatically mean Korea will build Temelin, and that French or American suppliers can also offer a price. &lt;a class="link" href="https://oenergetice.cz/en/nuclear-power/government-to-decide-next-year-on-new-units-at-czech-temelin-havlicek-says" target="_blank" rel="noopener"
 &gt;oEnergetice&lt;/a&gt; How KHNP delivers Dukovany will be its scorecard in that contest.&lt;/p&gt;
&lt;h2 id="what-changes-for-each-company"&gt;What changes for each company
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Contract&lt;/th&gt;
 &lt;th&gt;Schedule and cash flow&lt;/th&gt;
 &lt;th&gt;What moves the stock&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Loses the construction award; settlement on ending the consortium agreement and possible legal action&lt;/td&gt;
 &lt;td&gt;Recovery of costs spent during the talks&lt;/td&gt;
 &lt;td&gt;Loss of the overseas nuclear EPC reference option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Doosan Enerbility&lt;/td&gt;
 &lt;td&gt;Keeps about 5.64tn won of main equipment (4.93tn reactor equipment, 0.71tn turbine generator); loses its construction share (size not public)&lt;/td&gt;
 &lt;td&gt;No advance payment; paid as work progresses, so revenue recognition may slip if construction is late&lt;/td&gt;
 &lt;td&gt;High earnings multiple and a public dispute with its customer&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Keeps main design (about 1.25tn won, from KHNP) and reactor-system design (about 0.37tn won, from Doosan)&lt;/td&gt;
 &lt;td&gt;Must reset the design–construction interface with a new builder&lt;/td&gt;
 &lt;td&gt;Pressure for design-change costs to land on a lump-sum design contract&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai E&amp;amp;C, Samsung C&amp;amp;T&lt;/td&gt;
 &lt;td&gt;Market watching whether they bid (unconfirmed)&lt;/td&gt;
 &lt;td&gt;A winner takes the same first-of-a-kind fixed-price risk&lt;/td&gt;
 &lt;td&gt;Tender terms; competition for staff and capital with US AP1000 work&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO KPS, KEPCO Nuclear Fuel&lt;/td&gt;
 &lt;td&gt;Negotiating contracts targeted for 2027&lt;/td&gt;
 &lt;td&gt;Contract and revenue timing may slip&lt;/td&gt;
 &lt;td&gt;Small&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO (KHNP&amp;rsquo;s parent)&lt;/td&gt;
 &lt;td&gt;KHNP may carry more construction risk&lt;/td&gt;
 &lt;td&gt;Czech project margin may shrink&lt;/td&gt;
 &lt;td&gt;Small relative to group size&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Smaller equipment makers&lt;/td&gt;
 &lt;td&gt;Limited direct Czech exposure&lt;/td&gt;
 &lt;td&gt;Indirect effect of delays&lt;/td&gt;
 &lt;td&gt;BHI, Woojin and Woori Technology fell 6–7% today&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="scenarios-and-the-signals-that-decide-them"&gt;Scenarios and the signals that decide them
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Path&lt;/th&gt;
 &lt;th&gt;What happens&lt;/th&gt;
 &lt;th&gt;Signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Replacement on plan&lt;/td&gt;
 &lt;td&gt;New construction contract in July 2027, lump-sum terms kept, site preparation from January 2028&lt;/td&gt;
 &lt;td&gt;April tender has no quantity-settlement clause and large builders bid&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Replacement on softer terms&lt;/td&gt;
 &lt;td&gt;A new builder comes in, but KHNP takes more risk through quantity settlement or contingency, and the schedule slips by months&lt;/td&gt;
 &lt;td&gt;Tender adds design-change adjustment clauses, or the contract slips to the second half of 2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Return of the old consortium&lt;/td&gt;
 &lt;td&gt;Government mediation produces a deal with the Daewoo consortium&lt;/td&gt;
 &lt;td&gt;Ministry of Trade, Industry and Energy announces mediation; termination of the consortium agreement stops&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Long delay&lt;/td&gt;
 &lt;td&gt;No valid bids or legal disputes delay the project by more than a year; friction with EDU II&lt;/td&gt;
 &lt;td&gt;Injunction filing, failed tender, official concern from EDU II or ČEZ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;strong&gt;We see replacement on softer terms as the most likely path.&lt;/strong&gt; Large builders face the same first-of-a-kind risk. Korea&amp;rsquo;s large builders also have an alternative in US AP1000 work. KHNP must protect the Czech schedule and its European reference, so it has strong reasons to soften its terms in the end. In that case KHNP and KEPCO give up part of the Czech project&amp;rsquo;s margin, and the new builder gets better risk terms than the Daewoo consortium had.&lt;/p&gt;
&lt;p&gt;We have set the evidence that would change this view. If two or more large builders bid on an April tender without a quantity-settlement clause, we move to replacement on plan. If EDU II or ČEZ states schedule concerns officially, or an injunction is filed, we move toward a long delay.&lt;/p&gt;
&lt;h2 id="what-to-check-next"&gt;What to check next
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;When&lt;/th&gt;
 &lt;th&gt;What&lt;/th&gt;
 &lt;th&gt;How to read it&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Late October–November&lt;/td&gt;
 &lt;td&gt;Official responses from EDU II, ČEZ and the Czech government&lt;/td&gt;
 &lt;td&gt;Mention of schedule concerns or contractual remedies raises the risk of long delay&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;November–December&lt;/td&gt;
 &lt;td&gt;Settlement on ending the consortium agreement; any injunction&lt;/td&gt;
 &lt;td&gt;An injunction raises the risk of long delay&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;November–December&lt;/td&gt;
 &lt;td&gt;Any mediation by the industry ministry&lt;/td&gt;
 &lt;td&gt;Mediation raises the chance of the old consortium&amp;rsquo;s return&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;January 2027&lt;/td&gt;
 &lt;td&gt;Who attends the bidder briefing&lt;/td&gt;
 &lt;td&gt;Absent large builders point to softer terms or delay&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;April 2027&lt;/td&gt;
 &lt;td&gt;Contract terms in the tender notice&lt;/td&gt;
 &lt;td&gt;A quantity-settlement clause means softer terms; none means on plan&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 2027&lt;/td&gt;
 &lt;td&gt;Construction contract and its value&lt;/td&gt;
 &lt;td&gt;Compare value and risk clauses with the consortium&amp;rsquo;s final offer&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;January 2028&lt;/td&gt;
 &lt;td&gt;Start of site preparation&lt;/td&gt;
 &lt;td&gt;A delay means softer terms or long delay&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Any time&lt;/td&gt;
 &lt;td&gt;Outcome of the Barakah arbitration at the Korean Commercial Arbitration Board&lt;/td&gt;
 &lt;td&gt;Precedent for cost allocation inside Team Korea&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;strong&gt;Should the owner carry quantity risk on a first-of-a-kind reactor, or should the builder price it in? Tell us which clause in the April tender would change your mind.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="main-sources"&gt;Main sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sedaily.com/article/20107533" target="_blank" rel="noopener"
 &gt;Seoul Economic Daily, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.fnnews.com/news/202610291003332769" target="_blank" rel="noopener"
 &gt;Financial News, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://hankyung.com/article/202610292607i" target="_blank" rel="noopener"
 &gt;Korea Economic Daily, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.etoday.co.kr/news/view/2641662" target="_blank" rel="noopener"
 &gt;Etoday, consortium statement, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.ebn.co.kr/news/articleView.html?idxno=1729540" target="_blank" rel="noopener"
 &gt;EBN, October 28&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.koreatimes.co.kr/business/20261029/khnp-ends-talks-with-consortium-on-czech-nuclear-plant-construction-contract" target="_blank" rel="noopener"
 &gt;The Korea Times, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://mpo.gov.cz/en/guidepost/for-the-media/press-releases/epc-contract-for-completion-of-dukovany-has-been-signed--minister-of-industry-and-trade--lukas-vlcek-will-visit-korea-again-in-july--287912/" target="_blank" rel="noopener"
 &gt;Czech Ministry of Industry and Trade, main contract signed, June 4, 2025&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.hankookilbo.com/News/Read/A2025081917210002321" target="_blank" rel="noopener"
 &gt;Hankook Ilbo, Westinghouse settlement terms, August 19, 2025&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.ajunews.com/view/20261029093930495" target="_blank" rel="noopener"
 &gt;Aju Business Daily, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.businessplus.kr/news/articleView.html?idxno=118919" target="_blank" rel="noopener"
 &gt;Business Plus, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://v.daum.net/v/20260324050154461" target="_blank" rel="noopener"
 &gt;JoongAng Ilbo via Daum, construction cut report, March 24&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.etoday.co.kr/news/view/2545376" target="_blank" rel="noopener"
 &gt;Etoday, Barakah additional-cost arbitration, January 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://news.mtn.co.kr/article/2026102909393462147" target="_blank" rel="noopener"
 &gt;MTN, October 29&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://oenergetice.cz/en/nuclear-power/south-koreas-khnp-to-build-at-czechias-dukovany-plant-contract-signed-after-court-ruling" target="_blank" rel="noopener"
 &gt;oEnergetice and ČTK, contract signing and cost, June 4, 2025&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://oenergetice.cz/en/nuclear-power/government-to-decide-next-year-on-new-units-at-czech-temelin-havlicek-says" target="_blank" rel="noopener"
 &gt;oEnergetice, Temelin decision timing, January 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://info.westinghousenuclear.com/news/u.s.-korea-framework-advances-deployment-of-westinghouse-nuclear-technology-in-the-united-states" target="_blank" rel="noopener"
 &gt;Brookfield release on Westinghouse&amp;rsquo;s site, US–Korea framework, September 30, 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://dart.fss.or.kr/dsaf001/main.do?rcpNo=20251216801147" target="_blank" rel="noopener"
 &gt;Doosan Enerbility filing, reactor equipment contract, December 16, 2025&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://dart.fss.or.kr/dsaf001/main.do?rcpNo=20250507800786" target="_blank" rel="noopener"
 &gt;KEPCO filing, Barakah arbitration, May 7, 2025&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://dart.fss.or.kr/dsaf001/main.do?rcpNo=20260623800004" target="_blank" rel="noopener"
 &gt;KEPCO filing, change of arbitration body, June 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Naver Finance prices, market values and exchange rate, intraday on October 29&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Research and information only; not investment advice. Estimates and scenarios are analytical assumptions, not company guidance, trade instructions or personalized portfolio decisions. Figures are date-bound. AI tools assisted research and editing.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>The U.S.-Japan Yen Intervention and FIMA: Did Lower Oil Unlock the August Macro Regime?</title><link>https://koreainvestinsights.com/post/us-japan-yen-intervention-fima-repo-oil-yields-korea-market-2026-08-03/</link><pubDate>Mon, 03 Aug 2026 22:55:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/us-japan-yen-intervention-fima-repo-oil-yields-korea-market-2026-08-03/</guid><description>&lt;p&gt;Our August 1 report argued that oil, U.S. long yields, and the yen carry trade were the three remaining macro gates for the equity rebound. All three moved at once over the weekend and into Monday.&lt;/p&gt;
&lt;p&gt;The United States and Japan jointly bought yen on July 31, an operation officially confirmed by Japan&amp;rsquo;s Ministry of Finance on August 3. President Trump then ordered U.S. forces to hold off on new strikes against Iran. WTI and Brent fell 7% to 8% intraday, while the U.S. 10-year and 30-year yields declined by roughly 6bp and 5bp.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;One distinction is decisive. &lt;strong&gt;The joint intervention is confirmed, but there is no evidence yet that Japan funded the July 31 operation through the Fed&amp;rsquo;s FIMA repo facility.&lt;/strong&gt; Japan&amp;rsquo;s official statement says it plans to use FIMA in the future. The latest H.4.1 release is dated July 29, before the intervention, and showed zero foreign-official repo outstanding.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;The joint yen-buying intervention is official.&lt;/strong&gt; Japan&amp;rsquo;s Ministry of Finance said it bought yen in coordination with the U.S. Treasury and would not hesitate to act again. The operation relied on the September 2025 bilateral principle that intervention may be used against excessive volatility and disorderly moves.&lt;sup id="fnref1:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;The estimated size, about JPY8.45 trillion or roughly $53 billion, is not final.&lt;/strong&gt; It is a market estimate based on Bank of Japan account projections. Japan&amp;rsquo;s Ministry of Finance is scheduled to disclose the monthly total for July 30 through August 26 on August 28.&lt;sup id="fnref:7"&gt;&lt;a href="#fn:7" class="footnote-ref" role="doc-noteref"&gt;7&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:8"&gt;&lt;a href="#fn:8" class="footnote-ref" role="doc-noteref"&gt;8&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Actual FIMA usage remains unverified.&lt;/strong&gt; FIMA is a collateralized repo that lets approved foreign monetary authorities temporarily exchange Treasuries for dollars. Published terms allow overnight or seven-day maturities and a $60 billion daily cap per counterparty.&lt;sup id="fnref:9"&gt;&lt;a href="#fn:9" class="footnote-ref" role="doc-noteref"&gt;9&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:10"&gt;&lt;a href="#fn:10" class="footnote-ref" role="doc-noteref"&gt;10&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Calling FIMA money printing or QE is too strong.&lt;/strong&gt; The transaction can temporarily expand reserves and the Fed&amp;rsquo;s assets, but it is fully collateralized, reversible, and priced as a backstop. It can reduce the tail risk of forced Treasury sales without becoming a conventional bond-purchase program.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Trump paused new strikes; he did not end the war.&lt;/strong&gt; At about 9:40 a.m. ET on August 3, WTI was $78.74 and Brent $83.03, down 7.0% and 7.9% from July 31. A breakdown in diplomacy could quickly restore the risk premium.&lt;sup id="fnref1:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref1:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Rate stabilization has started, not finished.&lt;/strong&gt; The 10-year was 4.688% and the 30-year 5.228% at the same snapshot. Lower oil compresses inflation risk, while AI-related bond issuance, fiscal deficits, and Japanese repatriation continue to support a higher yield floor.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Korea&amp;rsquo;s market split is consistent with carry friction, but it is not causal proof.&lt;/strong&gt; On August 3, the KOSPI fell 5.12%, Samsung Electronics 8.76%, and SK hynix 8.79%, while the KOSDAQ rose 2.44%. The pattern points to liquid mega-cap pressure, but profit-taking, ADR flows, program trading, and the reversal of Friday&amp;rsquo;s surge may also have mattered.&lt;sup id="fnref:11"&gt;&lt;a href="#fn:11" class="footnote-ref" role="doc-noteref"&gt;11&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The joint intervention does not eliminate the yen carry trade; it attempts to manage the speed of adjustment. FIMA is a short-term liquidity bridge that can avoid forced Treasury sales. Oil's sharp decline clearly improved the August macro direction, but the bridge reaches solid ground only if gradual BOJ normalization, softer U.S. inflation, and a narrower rate differential follow. The locks are turning, but they are not yet fully open.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-verification-table"&gt;1. Verification table
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Claim&lt;/th&gt;
 &lt;th&gt;Verdict&lt;/th&gt;
 &lt;th&gt;Supported conclusion&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Joint U.S.-Japan yen buying on July 31&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Confirmed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Official Japanese statement identifies coordination with the U.S. Treasury&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First coordinated operation since 2011&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Broadly supported&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;AP cites 2011 as the last major coordinated case, though that operation sold yen rather than bought it&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First U.S. yen purchase since 1998&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Reported&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Reported by Reuters and other major outlets&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Japan spent $59 billion&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Needs correction&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Market estimate is about JPY8.45tn, roughly $52.8bn at the prevailing rate; official total is due August 28&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FIMA funded the full intervention&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Unverified&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Japan announced prospective use, not the July 31 amount&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The FIMA limit has already been raised&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Unverified&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Current public limit remains $60bn per counterparty per day&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FIMA is backdoor QE&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Debatable interpretation&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;It is collateralized and temporary, but it can reduce forced Treasury selling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Trump ended the Iran war&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Overstated&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;He ordered a pause in new strikes and said a deal was near&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Lower oil removed rate risk&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Partly true&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Yields fell, but structural term-premium pressures remain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea&amp;rsquo;s memory selloff proves a yen carry unwind&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Plausible, not proven&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Price action fits the hypothesis, but funding-currency data are unavailable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-the-three-day-sequence"&gt;2. The three-day sequence
&lt;/h2&gt;&lt;h3 id="july-31-coordinated-fx-intervention"&gt;July 31: coordinated FX intervention
&lt;/h3&gt;&lt;p&gt;Japan&amp;rsquo;s Ministry of Finance bought yen in coordination with the U.S. Treasury to counter excessive volatility and disorderly moves. The policy basis was already written into the September 2025 U.S.-Japan finance ministers&amp;rsquo; statement.&lt;sup id="fnref2:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref1:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;USD/JPY moved from 163.30 on July 30 to 160.18 on July 31 and 156.38 intraday on August 3. That is roughly a 4.4% appreciation of the yen.&lt;sup id="fnref2:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The historical comparison needs care. The 2011 coordinated action was designed to weaken an excessively strong yen after the Tohoku earthquake. The 2026 operation bought yen to counter excessive weakness. The comparison demonstrates rarity, not identical policy direction.&lt;sup id="fnref:12"&gt;&lt;a href="#fn:12" class="footnote-ref" role="doc-noteref"&gt;12&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h3 id="august-2-3-pause-in-iran-strikes-and-an-oil-selloff"&gt;August 2-3: pause in Iran strikes and an oil selloff
&lt;/h3&gt;&lt;p&gt;Trump ordered U.S. forces to hold off on new attacks against Iran and said an agreement to end the fighting was near. Oil immediately gave back part of its war premium.&lt;sup id="fnref2:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Asset&lt;/th&gt;
 &lt;th style="text-align: right"&gt;July 31&lt;/th&gt;
 &lt;th style="text-align: right"&gt;August 3 intraday&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Change&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;WTI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$84.67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$78.74&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Brent&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$90.12&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$83.03&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. 5-year&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.460%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.403%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.7bp&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. 10-year&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.745%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.688%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.7bp&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. 30-year&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.275%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.228%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.7bp&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Dollar index&lt;/td&gt;
 &lt;td style="text-align: right"&gt;99.80&lt;/td&gt;
 &lt;td style="text-align: right"&gt;99.77&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.03%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The August 3 data are intraday Yahoo Finance snapshots around 9:40 a.m. ET and may differ from the closing values.&lt;sup id="fnref3:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The transmission is straightforward: lower oil reduces near-term inflation expectations and the probability of another Fed hike. It also reduces the second-round risk from freight, insurance, and petrochemical costs. But a pause in strikes is reversible. Renewed fighting or shipping disruption would quickly reverse part of the move.&lt;/p&gt;
&lt;h2 id="3-how-fima-actually-works"&gt;3. How FIMA actually works
&lt;/h2&gt;&lt;p&gt;FIMA stands for the Foreign and International Monetary Authorities Repo Facility. Approved foreign central banks and monetary authorities temporarily exchange Treasuries held at the Federal Reserve Bank of New York for dollars, then repay dollars plus interest and receive the Treasuries back.&lt;sup id="fnref1:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref1:9"&gt;&lt;a href="#fn:9" class="footnote-ref" role="doc-noteref"&gt;9&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Traditional intervention funding
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Japan sells Treasuries in the market → receives dollars → sells dollars and buys yen
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ↓
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; upward pressure on Treasury yields
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;FIMA-backed funding
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Japan posts Treasuries to the Fed in repo → receives dollars → buys yen
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ↓
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; avoids a forced market sale
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Term&lt;/th&gt;
 &lt;th&gt;Current official setting&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Eligible users&lt;/td&gt;
 &lt;td&gt;Approved foreign central banks and monetary authorities&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Collateral&lt;/td&gt;
 &lt;td&gt;U.S. Treasuries held at the New York Fed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Maturity&lt;/td&gt;
 &lt;td&gt;Overnight or seven days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Limit&lt;/td&gt;
 &lt;td&gt;$60bn per counterparty per day&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pricing&lt;/td&gt;
 &lt;td&gt;Backstop rate generally above private repo in normal markets&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Risk control&lt;/td&gt;
 &lt;td&gt;Fully collateralized and margined; no FX risk for the Fed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disclosure&lt;/td&gt;
 &lt;td&gt;Weekly H.4.1 foreign-official repo balance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The estimated intervention size of about $52.8 billion is close to the $60 billion daily counterparty cap. That numerical proximity is notable, but it is not evidence that FIMA financed the operation. Japan can combine dollar cash, deposits, outright Treasury sales, and repo funding.&lt;/p&gt;
&lt;h3 id="what-the-latest-h41-can-and-cannot-tell-us"&gt;What the latest H.4.1 can and cannot tell us
&lt;/h3&gt;&lt;p&gt;The latest H.4.1 observation is July 29, two days before the intervention. Foreign-official repo was zero. This does not prove that FIMA was unused on July 31. It means the post-event data have not yet been published.&lt;sup id="fnref1:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The verification sequence is therefore:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Check the August 6 H.4.1 for an increase in foreign-official repo.&lt;/li&gt;
&lt;li&gt;Track whether the balance is repaid quickly or rolled for several weeks.&lt;/li&gt;
&lt;li&gt;Compare it with Japan&amp;rsquo;s official monthly intervention amount on August 28.&lt;/li&gt;
&lt;li&gt;Cross-check Japan&amp;rsquo;s reserves and the Fed&amp;rsquo;s custody holdings for foreign official accounts.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The large “foreign official” reverse-repo balance in H.4.1 is a different item with the opposite direction. It should not be mistaken for FIMA borrowing.&lt;/p&gt;
&lt;h2 id="4-is-fima-qe-or-a-bridge"&gt;4. Is FIMA QE or a bridge?
&lt;/h2&gt;&lt;h3 id="why-the-backdoor-qe-criticism-exists"&gt;Why the backdoor-QE criticism exists
&lt;/h3&gt;&lt;p&gt;The criticism has a real foundation:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;While outstanding, the Fed receives Treasuries and supplies dollar liquidity.&lt;/li&gt;
&lt;li&gt;Japan can avoid selling Treasuries into the market, reducing a potential yield spike.&lt;/li&gt;
&lt;li&gt;If the rate differential remains wide, intervention may fade and the repo may need to be rolled.&lt;/li&gt;
&lt;li&gt;An official pledge to stop disorderly moves may encourage some investors to rebuild carry positions.&lt;/li&gt;
&lt;li&gt;The adjustment cost can migrate from the Treasury market to the central-bank balance sheet.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="why-it-is-not-conventional-qe"&gt;Why it is not conventional QE
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;FIMA repo&lt;/th&gt;
 &lt;th&gt;Conventional QE&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Overnight or seven-day transaction&lt;/td&gt;
 &lt;td&gt;Persistent purchase of longer-duration assets&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Counterparty repays dollars plus interest&lt;/td&gt;
 &lt;td&gt;Central bank holds assets for an extended period&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Fully collateralized by Treasuries&lt;/td&gt;
 &lt;td&gt;Duration is structurally removed from the market&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Backstop pricing limits normal-time use&lt;/td&gt;
 &lt;td&gt;Easier financial conditions are the policy objective&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Dollar liquidity and market functioning&lt;/td&gt;
 &lt;td&gt;Long-yield and credit-easing objective&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The best description is therefore a &lt;strong&gt;priced, collateralized loan that lets a foreign authority obtain dollars without a forced Treasury sale&lt;/strong&gt;. It is not free funding. It is also not economically neutral, because avoiding a forced sale is supportive for Treasury market functioning.&lt;/p&gt;
&lt;h3 id="the-bridge-test"&gt;The bridge test
&lt;/h3&gt;&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Bridge reaches solid ground
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;lower oil → softer U.S. inflation pressure → less need for Fed hikes
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; + gradual BOJ normalization
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → narrower U.S.-Japan rate gap → less yen-selling pressure
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → FIMA repayment and fewer interventions
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Bridge fails
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;oil reaccelerates and U.S. inflation stays firm → high Fed rates
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; + limits on further BOJ hikes
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → wide rate gap persists → yen shorts rebuild
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → repeat intervention, FIMA rollover, or outright Treasury sales
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The fastest way to settle the debate is to watch FIMA balances, not rhetoric. A temporary increase followed by repayment would validate the bridge. A balance that grows or rolls for weeks would look more like deferred adjustment.&lt;/p&gt;
&lt;h2 id="5-how-far-can-lower-oil-pull-yields-down"&gt;5. How far can lower oil pull yields down?
&lt;/h2&gt;&lt;p&gt;Oil is not the only driver of U.S. yields, but it became the fastest upstream variable in late July.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;lower oil
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ less pressure on gasoline, freight, and petrochemicals
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ lower near-term inflation expectations
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ less need for a September hike
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ lower front-end yields
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ some compression in long-end inflation premium
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Three structural forces still push the other way:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;AI data-center and big-tech bond issuance&lt;/li&gt;
&lt;li&gt;U.S. fiscal deficits and net Treasury supply&lt;/li&gt;
&lt;li&gt;Potential Japanese repatriation as BOJ policy normalizes&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The realistic form of stabilization is therefore not an immediate return to a low-4% 10-year. It is lower volatility inside a 4.5% to 4.8% range.&lt;/p&gt;
&lt;h3 id="one-month-yield-scenarios"&gt;One-month yield scenarios
&lt;/h3&gt;&lt;p&gt;These are analytical probabilities as of August 3, not official forecasts.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Path&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability&lt;/th&gt;
 &lt;th style="text-align: right"&gt;U.S. 10-year&lt;/th&gt;
 &lt;th&gt;Conditions&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Orderly stabilization&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.45-4.70%&lt;/td&gt;
 &lt;td&gt;Brent below $85, softer inflation, contained refunding burden&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High range&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.60-4.90%&lt;/td&gt;
 &lt;td&gt;Oil at $80-90, firm labor data, continued long issuance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Retest of 5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.90-5.10%&lt;/td&gt;
 &lt;td&gt;Iran re-escalation, upside CPI shock, larger issuance or weak auctions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="6-august-5-refunding-the-real-test-of-the-bessent-put"&gt;6. August 5 refunding: the real test of the “Bessent put”
&lt;/h2&gt;&lt;p&gt;In May, Treasury said it expected to keep nominal coupon and FRN auction sizes unchanged for at least several quarters and use bills and cash-management bills for near-term financing variation. The next quarterly refunding announcement is scheduled for August 5.&lt;sup id="fnref:13"&gt;&lt;a href="#fn:13" class="footnote-ref" role="doc-noteref"&gt;13&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;Markets should watch four items:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Changes in 10-, 20-, and 30-year auction sizes&lt;/li&gt;
&lt;li&gt;Continued reliance on bills&lt;/li&gt;
&lt;li&gt;The scale of liquidity-support buybacks&lt;/li&gt;
&lt;li&gt;Treasury&amp;rsquo;s treatment of potential Japanese Treasury sales&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Bill-heavy funding can reduce pressure on long yields, but it increases the speed at which fiscal interest costs reprice to the Fed&amp;rsquo;s policy rate. Lower oil and an eventual decline in policy rates therefore matter to Treasury as well.&lt;/p&gt;
&lt;p&gt;It would still be speculative to claim that the Iran decision was taken to engineer disinflation. Lower oil clearly helps fiscal and rate objectives, but it is not the only possible motive for military policy.&lt;/p&gt;
&lt;h2 id="7-the-objective-is-orderly-adjustment-not-a-yen-target"&gt;7. The objective is orderly adjustment, not a yen target
&lt;/h2&gt;&lt;p&gt;The operative policy words are “excess volatility” and “disorderly movements.” The United States and Japan did not announce a fixed USD/JPY target.&lt;/p&gt;
&lt;p&gt;The policy has two opposing effects on the carry trade.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Short-term cushion&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;An official buyer exists in the FX market.&lt;/li&gt;
&lt;li&gt;FIMA can provide an alternative to forced Treasury sales.&lt;/li&gt;
&lt;li&gt;Dollar liquidity stress can be contained during intervention.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Medium-term pressure&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The policy tolerance for one-way yen weakness has narrowed.&lt;/li&gt;
&lt;li&gt;The probability of BOJ normalization rises.&lt;/li&gt;
&lt;li&gt;A larger risk of yen appreciation reduces expected carry returns.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;This is not a simple put for carry traders. It cuts the liquidation tail while also making the trade&amp;rsquo;s return less attractive.&lt;/p&gt;
&lt;h2 id="8-why-koreas-kospi-and-kosdaq-diverged"&gt;8. Why Korea&amp;rsquo;s KOSPI and KOSDAQ diverged
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Asset&lt;/th&gt;
 &lt;th style="text-align: right"&gt;August 3 level&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Daily move&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6,257.45&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.12%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;737.35&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.44%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW239,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.76%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW1,567,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.79%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;USD/KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,427.08&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Intraday, +0.46% from July 31&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;USD/JPY&lt;/td&gt;
 &lt;td style="text-align: right"&gt;156.38&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Intraday, -2.37% from July 31&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW per JPY100&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 912.5&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Yen strengthening faster than won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Two observations are defensible. Selling was concentrated in index-heavy, liquid memory mega-caps, and a rising KOSDAQ argues against a broad domestic funding seizure.&lt;/p&gt;
&lt;p&gt;It is not a controlled experiment. Samsung and SK hynix were also natural profit-taking targets after Friday&amp;rsquo;s surge. U.S. memory moves, ADR and program flows, and index hedging may have contributed. Real-time data on investors&amp;rsquo; funding currencies do not exist.&lt;/p&gt;
&lt;p&gt;The careful conclusion is that the price pattern is &lt;strong&gt;consistent with short-term friction from a rapid yen appreciation&lt;/strong&gt;, but one day of returns cannot prove forced liquidation of yen-funded Korean equity positions.&lt;/p&gt;
&lt;h3 id="medium-term-korean-channels"&gt;Medium-term Korean channels
&lt;/h3&gt;&lt;p&gt;If yen appreciation becomes orderly, four channels could turn supportive:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;A softer dollar and more stable Asian currencies reduce FX risk for foreign investors.&lt;/li&gt;
&lt;li&gt;Faster yen appreciation can improve Korean exporters&amp;rsquo; competitive position against Japan in autos, machinery, steel, and shipbuilding.&lt;/li&gt;
&lt;li&gt;A stronger won plus lower oil reduces import inflation and expands the Bank of Korea&amp;rsquo;s policy room.&lt;/li&gt;
&lt;li&gt;Memory has limited direct Japanese competition outside NAND, so global AI demand and currency translation matter more than bilateral export competition.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="9-big-tech-and-semiconductor-implications"&gt;9. Big tech and semiconductor implications
&lt;/h2&gt;&lt;h3 id="us-big-tech"&gt;U.S. big tech
&lt;/h3&gt;&lt;p&gt;Lower oil and yields help in two ways: they reduce the discount rate on future cash flow and ease funding costs for data-center bonds and project finance. Yet a 10-year yield in the high-4% range is not a low-discount-rate regime. Treasury refunding and big-tech issuance can still cap valuation expansion.&lt;/p&gt;
&lt;h3 id="memory-semiconductors"&gt;Memory semiconductors
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Horizon&lt;/th&gt;
 &lt;th&gt;Channel&lt;/th&gt;
 &lt;th&gt;Impact&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Short term&lt;/td&gt;
 &lt;td&gt;Yen spike, carry reduction, mega-cap hedging&lt;/td&gt;
 &lt;td&gt;More volatility in Samsung and SK hynix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Medium term&lt;/td&gt;
 &lt;td&gt;Lower oil and yields, easier AI CAPEX financing&lt;/td&gt;
 &lt;td&gt;Supportive for HBM and server DRAM demand duration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Translation&lt;/td&gt;
 &lt;td&gt;Stronger won&lt;/td&gt;
 &lt;td&gt;Mild headwind to KRW-reported dollar revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Competition&lt;/td&gt;
 &lt;td&gt;Stronger yen&lt;/td&gt;
 &lt;td&gt;Limited direct impact outside Kioxia/NAND&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The central memory variables remain 2027-2028 volume and pricing, hyperscaler CAPEX, customer prepayments, and purchase contracts. Macro stability does not create the earnings. It determines the discount rate applied to those earnings.&lt;/p&gt;
&lt;h2 id="10-three-scenarios"&gt;10. Three scenarios
&lt;/h2&gt;&lt;h3 id="a-orderly-policy-bridge-45"&gt;A. Orderly policy bridge, 45%
&lt;/h3&gt;&lt;p&gt;USD/JPY stabilizes at 153-158, Brent remains below $85, any FIMA balance is repaid within weeks, and the U.S. 10-year settles at 4.45-4.70%. Earnings regain control of big-tech and memory pricing.&lt;/p&gt;
&lt;h3 id="b-temporary-patch-and-repeat-intervention-35"&gt;B. Temporary patch and repeat intervention, 35%
&lt;/h3&gt;&lt;p&gt;The yen strengthens initially but weakens back toward 160, FIMA rolls or foreign-official Treasury custody falls, and the 10-year trades at 4.6-4.9%. Equities remain caught between earnings rallies and rate/FX pullbacks.&lt;/p&gt;
&lt;h3 id="c-bridge-failure-and-disorderly-adjustment-20"&gt;C. Bridge failure and disorderly adjustment, 20%
&lt;/h3&gt;&lt;p&gt;Either Iran re-escalates and lifts oil and yields, or USD/JPY falls through 150 quickly enough to accelerate carry liquidation. Leveraged and highly liquid growth exposures would face the most volatility.&lt;/p&gt;
&lt;h2 id="11-dates-and-dashboards"&gt;11. Dates and dashboards
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Date&lt;/th&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;August 5&lt;/td&gt;
 &lt;td&gt;U.S. Treasury quarterly refunding&lt;/td&gt;
 &lt;td&gt;Long-end supply, bill reliance, buybacks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;August 6&lt;/td&gt;
 &lt;td&gt;Federal Reserve H.4.1&lt;/td&gt;
 &lt;td&gt;First post-intervention clue on FIMA use&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;August 7&lt;/td&gt;
 &lt;td&gt;U.S. employment report&lt;/td&gt;
 &lt;td&gt;Hike expectations and the rate differential&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mid-August&lt;/td&gt;
 &lt;td&gt;U.S. CPI&lt;/td&gt;
 &lt;td&gt;Lagged effect of July&amp;rsquo;s oil rise and August&amp;rsquo;s decline&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;August 28&lt;/td&gt;
 &lt;td&gt;Japan&amp;rsquo;s monthly intervention total&lt;/td&gt;
 &lt;td&gt;Official test of the JPY8.45tn estimate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;September 17-18&lt;/td&gt;
 &lt;td&gt;BOJ meeting&lt;/td&gt;
 &lt;td&gt;Whether gradual normalization can land the bridge&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The useful daily dashboard is compact: daily USD/JPY volatility, Brent at $85, the 10-year at 4.70% and 5.00%, H.4.1 foreign-official repo, and the relative return of Korea&amp;rsquo;s mega-caps versus the KOSDAQ.&lt;/p&gt;
&lt;h2 id="12-red-team"&gt;12. Red team
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Intervention may create more confidence than expected.&lt;/strong&gt; Repeated coordination plus a BOJ hike could narrow the rate gap quickly, making FIMA unnecessary.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Lower oil may not be enough.&lt;/strong&gt; Fiscal deficits and corporate issuance could keep the 10-year near 4.8-5.0% even with WTI below $80.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A stronger yen may remain negative for Korea.&lt;/strong&gt; If yen-funded exposure is large, capital withdrawal could precede any export-competitiveness benefit.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Iran diplomacy may fail again.&lt;/strong&gt; The strike pause is reversible, and renewed shipping risk could reprice oil, inflation, and yields together.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="13-final-assessment"&gt;13. Final assessment
&lt;/h2&gt;&lt;p&gt;August 3 was not the day the macro locks fully opened. It was the first day policy and prices began pointing in the same direction.&lt;/p&gt;
&lt;p&gt;The United States and Japan jointly resisted disorderly yen weakness. Japan disclosed a plan to use FIMA as an alternative to forced Treasury sales. The United States paused new strikes against Iran. Oil and long yields fell together.&lt;/p&gt;
&lt;p&gt;Those developments reduce the late-July macro tail. They do not eliminate the interest differential, confirm actual FIMA usage, end the war, or remove structural long-bond supply.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;The joint intervention and oil decline improved the macro direction. Policy has bought time for an orderly adjustment, not removed the adjustment itself. Whether that time becomes durable stability will be tested by the August 5 refunding, the August 6 FIMA data, Japan&amp;rsquo;s August 28 intervention total, and the September BOJ meeting.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="14-evidence-and-data-health"&gt;14. Evidence and data health
&lt;/h2&gt;&lt;p&gt;&lt;strong&gt;Fact:&lt;/strong&gt; Official Japanese, U.S. Treasury, and Federal Reserve documents support the intervention framework and FIMA mechanics. Korean market levels are August 3 closes from Naver Finance. Oil, FX, and U.S. yields are August 3 intraday Yahoo Finance snapshots.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Inference:&lt;/strong&gt; The speed-management interpretation, the oil-to-yield transmission, the Korean carry-friction hypothesis, and the scenario probabilities are analytical judgments.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Blocked:&lt;/strong&gt; The official intervention amount and U.S.-Japan split, actual FIMA use, the yen-funded share of foreign Korean equity exposure, final Iran terms, and the August 5 issuance mix are not yet public.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Coverage health:&lt;/strong&gt; Policy-document confidence is high. Market data are intraday and FIMA remains the central missing variable, so overall confidence is medium-high. The next decisive checks are the August 6 H.4.1 and Japan&amp;rsquo;s August 28 monthly release.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;For research and information purposes only. This is not investment advice, and the scenario probabilities are analytical judgments rather than official forecasts.&lt;/em&gt;&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.&lt;/em&gt;&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.mof.go.jp/english/public_relations/statement/others/20260803073000.html" target="_blank" rel="noopener"
 &gt;Japan Ministry of Finance, Statement on Joint Foreign Exchange Intervention, August 3, 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref2:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://apnews.com/article/194ab3d3130bb44de445cf3d05954beb" target="_blank" rel="noopener"
 &gt;AP, Oil prices drop after Trump orders U.S. forces to hold off on new strikes against Iran&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref2:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://finance.yahoo.com/quote/CL%3DF/" target="_blank" rel="noopener"
 &gt;Yahoo Finance WTI&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/quote/BZ%3DF/" target="_blank" rel="noopener"
 &gt;Brent&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/quote/JPY%3DX/" target="_blank" rel="noopener"
 &gt;USD/JPY&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/quote/KRW%3DX/" target="_blank" rel="noopener"
 &gt;USD/KRW&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/quote/%5ETNX/" target="_blank" rel="noopener"
 &gt;U.S. 10-year&lt;/a&gt;, and &lt;a class="link" href="https://finance.yahoo.com/quote/%5ETYX/" target="_blank" rel="noopener"
 &gt;U.S. 30-year&lt;/a&gt;, August 3 intraday snapshot around 9:40 a.m. ET.&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref2:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref3:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.federalreserve.gov/monetarypolicy/fima-repo-facility.htm" target="_blank" rel="noopener"
 &gt;Federal Reserve, Foreign and International Monetary Authorities Repo Facility&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.federalreserve.gov/releases/h41/current/" target="_blank" rel="noopener"
 &gt;Federal Reserve, H.4.1 Factors Affecting Reserve Balances&lt;/a&gt;. The latest observation is July 29, 2026.&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;&lt;a class="link" href="https://home.treasury.gov/news/press-releases/sb0245" target="_blank" rel="noopener"
 &gt;U.S. Treasury, U.S.-Japan Finance Ministers&amp;rsquo; Joint Statement, September 11, 2025&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:7"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.bloomberg.com/news/articles/2026-07-31/boj-data-suggest-japan-intervention-of-about-8-45-trillion-yen" target="_blank" rel="noopener"
 &gt;Bloomberg, BOJ data suggest Japan intervention of about JPY8.45 trillion&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:7" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:8"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.mof.go.jp/policy/international_policy/reference/feio/index.html" target="_blank" rel="noopener"
 &gt;Japan Ministry of Finance, Foreign Exchange Intervention Operations release schedule&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:8" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:9"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.federalreserve.gov/monetarypolicy/fima-repo-facility-faqs.htm" target="_blank" rel="noopener"
 &gt;Federal Reserve, FIMA Repo Facility FAQs&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:9" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:9" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:10"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.federalreserve.gov/monetarypolicy/files/FOMC_StandingFIMARepoResolution.pdf" target="_blank" rel="noopener"
 &gt;Federal Reserve, Standing FIMA Repurchase Agreement Resolution&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:10" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:11"&gt;
&lt;p&gt;&lt;a class="link" href="https://finance.naver.com/sise/sise_index.naver?code=KOSPI" target="_blank" rel="noopener"
 &gt;Naver Finance KOSPI&lt;/a&gt;, &lt;a class="link" href="https://finance.naver.com/sise/sise_index.naver?code=KOSDAQ" target="_blank" rel="noopener"
 &gt;KOSDAQ&lt;/a&gt;, &lt;a class="link" href="https://finance.naver.com/item/main.naver?code=005930" target="_blank" rel="noopener"
 &gt;Samsung Electronics&lt;/a&gt;, and &lt;a class="link" href="https://finance.naver.com/item/main.naver?code=000660" target="_blank" rel="noopener"
 &gt;SK hynix&lt;/a&gt;, August 3, 2026 close.&amp;#160;&lt;a href="#fnref:11" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:12"&gt;
&lt;p&gt;&lt;a class="link" href="https://apnews.com/article/7316599afed35629a27ae23a35f569fd" target="_blank" rel="noopener"
 &gt;AP, U.S. dollar weakens sharply against the Japanese yen after market interventions&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:12" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:13"&gt;
&lt;p&gt;&lt;a class="link" href="https://home.treasury.gov/news/press-releases/sb0489" target="_blank" rel="noopener"
 &gt;U.S. Treasury, May 2026 Quarterly Refunding Statement&lt;/a&gt;. The next announcement is scheduled for August 5, 2026.&amp;#160;&lt;a href="#fnref:13" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>The Last Gate Is Macro Stability: Oil, Long Yields, and the Yen Carry Trade</title><link>https://koreainvestinsights.com/post/macro-stability-oil-us-yields-boj-yen-carry-august-2026/</link><pubDate>Sat, 01 Aug 2026 09:52:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/macro-stability-oil-us-yields-boj-yen-carry-august-2026/</guid><description>&lt;p&gt;Corporate and market-structure evidence improved in the final week of July. Big tech continued to validate AI demand and capital spending, memory earnings estimates held up, and the transfer of a large public-equity book suggested that forced selling in crowded AI infrastructure trades had passed its peak.&lt;/p&gt;
&lt;p&gt;The next leg of the market, however, will not be opened by earnings alone. Friday&amp;rsquo;s WTI close was $86.80 per barrel and Brent closed at $90.12. The U.S. 10-year Treasury yield ended at 4.75%, while the 30-year yield was 5.27%. The Bank of Japan held its policy rate near 1.0%, but one member preferred 1.25%. USD/JPY then fell from 163.86 on July 29 to 157.40 on August 1.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The remaining question is straightforward:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Can validated micro fundamentals overcome shocks from oil, long yields, and a potential yen carry unwind?&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;p&gt;This report first verifies the claims circulating after Friday&amp;rsquo;s close, then maps the August 3-7 calendar and translates the outcomes into scenarios for big tech and memory stocks.&lt;/p&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Micro improved, but macro did not stabilize.&lt;/strong&gt; Earnings, order books, and the end of forced selling improved the market floor. Oil around $90 and long yields near 5%, however, can still cap valuation multiples.&lt;/li&gt;
&lt;li&gt;Re-escalation in oil and Middle East risk is real. Claims that diplomacy is completely dead or that Saudi Arabia&amp;rsquo;s East-West pipeline was definitively destroyed are stronger than the available evidence. Formal ceasefire conditions have not been restored, but mediation channels remain open.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;The IEA&amp;rsquo;s 400 million-barrel release is real, but it was decided on March 11, not in July. Calling it a failure is not identifiable from current prices. A more defensible conclusion is that the release did not eliminate the risk premium, while the counterfactual price without the release is unknowable.&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;The Fed held the target range at 3.50-3.75% on July 29. The vote was 9-3, with three members preferring a 25bp increase. The hold was not a dovish all-clear.&lt;sup id="fnref:7"&gt;&lt;a href="#fn:7" class="footnote-ref" role="doc-noteref"&gt;7&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;On July 31, the official Treasury curve showed 4.75% for the 10-year, 5.28% for the 20-year, and 5.27% for the 30-year. The widely cited 5.28% belonged to the 20-year point.&lt;sup id="fnref1:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;A 63% probability of a September hike was a CME FedWatch market snapshot, not Fed guidance. BofA projected 75bp of hikes in 2026, but that was not the broad brokerage consensus.&lt;sup id="fnref:8"&gt;&lt;a href="#fn:8" class="footnote-ref" role="doc-noteref"&gt;8&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:9"&gt;&lt;a href="#fn:9" class="footnote-ref" role="doc-noteref"&gt;9&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;The BOJ held at 1.0% by an 8-1 vote and identified oil, semiconductor prices, and yen depreciation as upside inflation risks. Its next meeting is September 17-18, followed by October 29-30 and December 17-18.&lt;sup id="fnref:10"&gt;&lt;a href="#fn:10" class="footnote-ref" role="doc-noteref"&gt;10&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:11"&gt;&lt;a href="#fn:11" class="footnote-ref" role="doc-noteref"&gt;11&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;The key events are the August 5 Treasury quarterly refunding and ISM Services, followed by the August 7 U.S. employment report. Strong labor data plus high oil would reopen the path to a 5% 10-year yield.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 August is a contest between validated micro fundamentals and three unresolved macro gates. The base case is Brent in an $80-95 range and the U.S. 10-year at 4.60-4.90%. The dangerous tail is a simultaneous break above $100 oil and 5% long yields while the yen strengthens abruptly. The July rebound becomes a durable trend only when volatility in oil, rates, and the yen declines together.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-verification-before-interpretation"&gt;1. Verification before interpretation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Claim&lt;/th&gt;
 &lt;th&gt;Verdict&lt;/th&gt;
 &lt;th&gt;What can be supported&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Friday WTI at $86.80&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Confirmed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;July 31 futures close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Friday Brent above $90&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Confirmed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;$90.12 close; $90.74 on July 29&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The ceasefire track is officially dead&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Overstated&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Fighting resumed and no formal ceasefire was restored, but mediation continued&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Houthis definitively hit the Saudi East-West pipeline&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Not independently confirmed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Threats and attacks involving shipping and oil infrastructure are documented&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The IEA released 400 million barrels&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Confirmed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Decision by 32 members on March 11&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The IEA release failed&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Not identifiable&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;The counterfactual oil price is unobservable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The Fed hold caused a credibility shock&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Market interpretation&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Yields rose, but oil, inflation, issuance, and the BOJ also mattered&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;September hike probability was 63%&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Market snapshot&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;CME-based pricing, not official guidance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. 30-year was 5.28%&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Corrected by 1bp&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;20-year 5.28%, 30-year 5.27%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;BOJ held at 1%, one member wanted 1.25%&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Confirmed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;8-1 decision&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The yen remains at a 40-year low&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Time-sensitive&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;It approached 164 per dollar, then strengthened to 157.40 by August 1&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. yen intervention is confirmed&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Reported, not official&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Discussions were reported; no official intervention announcement was found&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The broad conclusion survives these corrections. Oil and long yields remain high, while yen carry risk has increased. The corrections simply prevent unsupported events from being used as false certainty.&lt;/p&gt;
&lt;h2 id="2-oil-the-first-macro-lock"&gt;2. Oil: the first macro lock
&lt;/h2&gt;&lt;p&gt;WTI moved from $82.61 on July 27 to $86.80 on July 31, after briefly falling to $79.26 on July 28. Brent moved from $88.36 to $90.12 and printed $90.74 on July 29.&lt;sup id="fnref1:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The path reflects three observable pressures:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;U.S.-Iran military tensions re-escalated.&lt;/li&gt;
&lt;li&gt;Security risks remained around Hormuz and Red Sea shipping.&lt;/li&gt;
&lt;li&gt;A formal ceasefire was not restored, leaving a fat tail around physical supply.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The correct dashboard is not rhetoric alone. It is physical cargo flow, weekly attack frequency, war-risk insurance, tanker freight, and verified mediation outcomes.&lt;/p&gt;
&lt;h3 id="the-400-million-barrel-release"&gt;The 400 million-barrel release
&lt;/h3&gt;&lt;p&gt;On March 11, IEA members agreed to release 400 million barrels, including 172 million from the United States and 80 million from Japan. The IEA said exports through Hormuz had fallen below 10% of their pre-conflict level at the time.&lt;sup id="fnref1:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The release could bridge a temporary physical gap and limit an immediate price spike. It could not remove military risk, create permanent production, or normalize insurance and shipping. Current $90 Brent therefore shows that the risk premium survived, not that the release had no effect.&lt;/p&gt;
&lt;h3 id="one-month-oil-scenarios"&gt;One-month oil scenarios
&lt;/h3&gt;&lt;p&gt;These are judgments as of August 1, not official forecasts.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Brent range&lt;/th&gt;
 &lt;th&gt;Conditions&lt;/th&gt;
 &lt;th&gt;Market effect&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Diplomatic repair&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$75-85&lt;/td&gt;
 &lt;td&gt;Verifiable maritime agreement and fewer attacks&lt;/td&gt;
 &lt;td&gt;Lower breakevens and long yields&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Managed conflict&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$80-95&lt;/td&gt;
 &lt;td&gt;Limited fighting, partial disruption, OPEC and stock-release buffers&lt;/td&gt;
 &lt;td&gt;Persistent inflation pressure and capped growth multiples&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Wider supply shock&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$100-120&lt;/td&gt;
 &lt;td&gt;Hormuz blockade attempt or material damage to bypass infrastructure&lt;/td&gt;
 &lt;td&gt;Higher odds of 5% yields and cross-asset deleveraging&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The base case remains a high $80-95 range. Diplomacy is not closed, but it needs observable outputs rather than statements.&lt;/p&gt;
&lt;h2 id="3-the-fed-and-long-yields-confidence-plus-supply"&gt;3. The Fed and long yields: confidence plus supply
&lt;/h2&gt;&lt;p&gt;The Federal Reserve held the target range at 3.50-3.75% on July 29. Beth Hammack, Neel Kashkari, and Lorie Logan preferred a 25bp increase. The statement described inflation as elevated and highlighted uncertainty from the energy supply shock.&lt;sup id="fnref1:7"&gt;&lt;a href="#fn:7" class="footnote-ref" role="doc-noteref"&gt;7&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The three dissents made another hike a live policy option. Markets at one point assigned roughly 63% odds to a September hike through CME FedWatch.&lt;sup id="fnref1:8"&gt;&lt;a href="#fn:8" class="footnote-ref" role="doc-noteref"&gt;8&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;But the hold alone did not mechanically cause the rise in long yields. The 10-year embeds the expected policy path, inflation expectations, real growth, Treasury supply, term premium, and foreign demand. Oil, persistent inflation, coupon supply, and changing Japanese demand all contributed.&lt;/p&gt;
&lt;p&gt;June data were also mixed. Headline PCE fell 0.1% month over month but was 3.7% year over year. Core PCE eased to 3.3% from 3.4%, so it was not at a 12-month high. Q2 GDP rose at a 1.5% annualized rate, while consumption grew 3.2% and private domestic demand 3.9%.&lt;sup id="fnref:12"&gt;&lt;a href="#fn:12" class="footnote-ref" role="doc-noteref"&gt;12&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h3 id="one-month-10-year-scenarios"&gt;One-month 10-year scenarios
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability&lt;/th&gt;
 &lt;th style="text-align: right"&gt;10-year range&lt;/th&gt;
 &lt;th&gt;Required mix&lt;/th&gt;
 &lt;th&gt;Equity effect&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Confidence repair&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.45-4.60%&lt;/td&gt;
 &lt;td&gt;Lower oil, softer labor, benign refunding&lt;/td&gt;
 &lt;td&gt;Multiple recovery in AI and memory&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High range&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.60-4.90%&lt;/td&gt;
 &lt;td&gt;Oil at $80-95, resilient growth, open September debate&lt;/td&gt;
 &lt;td&gt;Earnings stocks hold; narrative stocks weaken&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Break above 5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Above 5.00%&lt;/td&gt;
 &lt;td&gt;$100 oil, strong labor, supply shock, rapid yen reversal&lt;/td&gt;
 &lt;td&gt;Growth and leveraged assets deleverage together&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A September hike could paradoxically mark the peak in long yields if it restores inflation credibility and is paired with a signal that policy is sufficiently restrictive. That path requires oil stabilization as well as the hike itself.&lt;/p&gt;
&lt;h2 id="4-boj-and-the-yen-the-second-mechanical-risk"&gt;4. BOJ and the yen: the second mechanical risk
&lt;/h2&gt;&lt;p&gt;The BOJ held the overnight call rate near 1.0% on July 31. Hajime Takata proposed 1.25%, but the motion was defeated 8-1.&lt;sup id="fnref1:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The July Outlook said Japan would be supported by policy measures, accommodative financial conditions, and global AI demand. It also warned that CPI could remain clearly above 2% in the second half of fiscal 2026 because of crude oil, semiconductor prices, and yen depreciation. Inflation risks were skewed to the upside.&lt;sup id="fnref1:10"&gt;&lt;a href="#fn:10" class="footnote-ref" role="doc-noteref"&gt;10&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;This creates an uncomfortable feedback loop for memory investors:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;AI demand
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ higher semiconductor and memory prices
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ higher Japanese import inflation
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ more pressure on the BOJ to tighten
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ stronger yen and carry reduction
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ a liquidity and discount-rate shock to AI and Korean memory equities
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The danger is speed, not yen appreciation itself. USD/JPY fell from 163.86 on July 29 to 160.18 on July 31 and 157.40 on August 1.&lt;sup id="fnref:13"&gt;&lt;a href="#fn:13" class="footnote-ref" role="doc-noteref"&gt;13&lt;/a&gt;&lt;/sup&gt; A gradual move allows hedges to adjust. A 2-3% daily move can force leveraged positions to close.&lt;/p&gt;
&lt;p&gt;The real-time warning list is:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;a 2-3% one-day fall in USD/JPY,&lt;/li&gt;
&lt;li&gt;a jump in yen volatility,&lt;/li&gt;
&lt;li&gt;a sharp rise in the JGB 10-year yield,&lt;/li&gt;
&lt;li&gt;simultaneous selling in high-beta tech and emerging markets,&lt;/li&gt;
&lt;li&gt;deterioration in cross-currency funding conditions.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The next BOJ meeting is September 17-18, not October. Later meetings are October 29-30 and December 17-18.&lt;sup id="fnref1:11"&gt;&lt;a href="#fn:11" class="footnote-ref" role="doc-noteref"&gt;11&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h2 id="5-treasury-and-the-limits-of-a-bessent-put"&gt;5. Treasury and the limits of a “Bessent put”
&lt;/h2&gt;&lt;p&gt;The idea that Treasury Secretary Scott Bessent focuses on the 10-year yield is a policy and political framing, not a transfer of monetary policy from the Fed to the Treasury.&lt;/p&gt;
&lt;p&gt;Treasury&amp;rsquo;s actual tools are maturity composition, predictable issuance, buybacks, cash-management bills, and the demand effects of financial regulation. In May, Treasury said nominal coupon and FRN auction sizes were expected to remain steady for at least several quarters, with bills and cash-management bills absorbing near-term financing variation. It also planned up to $38 billion in liquidity-support buybacks and up to $25 billion in cash-management buybacks for the May-July quarter.&lt;sup id="fnref:14"&gt;&lt;a href="#fn:14" class="footnote-ref" role="doc-noteref"&gt;14&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;Markets sometimes call this a “Bessent put.” It is not a guaranteed ceiling on yields. A greater bill share can postpone long-duration supply pressure, but it raises refinancing sensitivity and cannot erase the fiscal deficit.&lt;/p&gt;
&lt;p&gt;The August 5 quarterly refunding should be read through three questions:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Are 10-, 20-, and 30-year auction sizes maintained?&lt;/li&gt;
&lt;li&gt;How much financing variation is shifted into bills?&lt;/li&gt;
&lt;li&gt;How are buyback sizes and maturity buckets changed?&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Reports also said U.S. officials discussed support for the yen and viewed it as undervalued.&lt;sup id="fnref:15"&gt;&lt;a href="#fn:15" class="footnote-ref" role="doc-noteref"&gt;15&lt;/a&gt;&lt;/sup&gt; No official U.S. intervention announcement was identified. Treasury&amp;rsquo;s July 23 FX report placed Japan on the Monitoring List and said its currency practices would be scrutinized.&lt;sup id="fnref:16"&gt;&lt;a href="#fn:16" class="footnote-ref" role="doc-noteref"&gt;16&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h2 id="6-august-3-7-calendar"&gt;6. August 3-7 calendar
&lt;/h2&gt;&lt;p&gt;Korea is 13 hours ahead of U.S. Eastern Daylight Time.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;KST&lt;/th&gt;
 &lt;th&gt;Event&lt;/th&gt;
 &lt;th&gt;Key data&lt;/th&gt;
 &lt;th&gt;Market question&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 3, 23:00&lt;/td&gt;
 &lt;td&gt;July ISM Manufacturing&lt;/td&gt;
 &lt;td&gt;New orders, employment, prices&lt;/td&gt;
 &lt;td&gt;Is growth weakening while costs rise?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 4, 21:30&lt;/td&gt;
 &lt;td&gt;June international trade&lt;/td&gt;
 &lt;td&gt;Real imports and exports&lt;/td&gt;
 &lt;td&gt;How resilient is domestic demand?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 4, 23:00&lt;/td&gt;
 &lt;td&gt;June JOLTS&lt;/td&gt;
 &lt;td&gt;Openings, quits, layoffs&lt;/td&gt;
 &lt;td&gt;Is labor demand still strong enough to absorb hikes?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 5&lt;/td&gt;
 &lt;td&gt;Treasury quarterly refunding&lt;/td&gt;
 &lt;td&gt;Maturity mix and buybacks&lt;/td&gt;
 &lt;td&gt;Does long-end supply lift term premium again?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 5, 23:00&lt;/td&gt;
 &lt;td&gt;July ISM Services&lt;/td&gt;
 &lt;td&gt;Employment, prices, new orders&lt;/td&gt;
 &lt;td&gt;Are service inflation and demand cooling?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 6, 21:30&lt;/td&gt;
 &lt;td&gt;Q2 productivity and costs&lt;/td&gt;
 &lt;td&gt;Productivity, unit labor costs&lt;/td&gt;
 &lt;td&gt;Can productivity absorb wage pressure?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Aug. 7, 21:30&lt;/td&gt;
 &lt;td&gt;July Employment Situation&lt;/td&gt;
 &lt;td&gt;Payrolls, unemployment, wages&lt;/td&gt;
 &lt;td&gt;Does September hike pricing harden or reverse?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The schedule uses official BLS, BEA, ISM, and Treasury calendars.&lt;sup id="fnref:17"&gt;&lt;a href="#fn:17" class="footnote-ref" role="doc-noteref"&gt;17&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:18"&gt;&lt;a href="#fn:18" class="footnote-ref" role="doc-noteref"&gt;18&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:19"&gt;&lt;a href="#fn:19" class="footnote-ref" role="doc-noteref"&gt;19&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref1:14"&gt;&lt;a href="#fn:14" class="footnote-ref" role="doc-noteref"&gt;14&lt;/a&gt;&lt;/sup&gt; July CPI follows on August 12 and will be the first major test of how quickly the oil shock reaches inflation data.&lt;/p&gt;
&lt;h2 id="7-cross-asset-matrix"&gt;7. Cross-asset matrix
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Oil&lt;/th&gt;
 &lt;th&gt;10-year&lt;/th&gt;
 &lt;th&gt;Yen&lt;/th&gt;
 &lt;th&gt;Likely market response&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Down&lt;/td&gt;
 &lt;td&gt;Down&lt;/td&gt;
 &lt;td&gt;Gradual strength&lt;/td&gt;
 &lt;td&gt;Best combination for big tech, memory multiples, and foreign flows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High range&lt;/td&gt;
 &lt;td&gt;4.60-4.90%&lt;/td&gt;
 &lt;td&gt;Stable&lt;/td&gt;
 &lt;td&gt;Earnings stocks hold while unproven narratives separate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Up&lt;/td&gt;
 &lt;td&gt;Above 5%&lt;/td&gt;
 &lt;td&gt;Stable&lt;/td&gt;
 &lt;td&gt;Discount-rate shock caps AI CAPEX and long-duration earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stable&lt;/td&gt;
 &lt;td&gt;Down&lt;/td&gt;
 &lt;td&gt;2-3% daily surge&lt;/td&gt;
 &lt;td&gt;Carry unwind can overwhelm better fundamentals&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Above $100&lt;/td&gt;
 &lt;td&gt;Above 5%&lt;/td&gt;
 &lt;td&gt;Abrupt strength&lt;/td&gt;
 &lt;td&gt;Worst tail: simultaneous deleveraging in growth, EM, and crowded trades&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The underappreciated case is stable oil and lower U.S. yields combined with a violent yen rally. It is positive for inflation but negative for leveraged positioning.&lt;/p&gt;
&lt;h2 id="8-implications-for-big-tech-and-memory"&gt;8. Implications for big tech and memory
&lt;/h2&gt;&lt;h3 id="big-tech"&gt;Big tech
&lt;/h3&gt;&lt;p&gt;AI revenue and higher discount rates can rise at the same time.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Higher AI revenue and backlog
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ higher future cash flow
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Higher long and credit yields
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ higher discount rate
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ lower present value of the same cash flow
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;August should therefore be judged by whether AI revenue growth exceeds CAPEX growth, backlog converts to cash, power constraints allow utilization, financing costs remain contained, and existing capacity earns enough pricing power to absorb depreciation.&lt;/p&gt;
&lt;h3 id="memory"&gt;Memory
&lt;/h3&gt;&lt;p&gt;Oil and rates do not immediately erase HBM, server DRAM, or NAND orders. They hit equities first through discount rates and positioning.&lt;/p&gt;
&lt;p&gt;The pullback is more likely an opportunity when HBM and server-memory volume assumptions hold, hyperscaler CAPEX remains intact, foreign and program selling slows, currencies move in an orderly manner, and earnings estimates hold or rise.&lt;/p&gt;
&lt;p&gt;It becomes a warning when $100 oil and a 5% 10-year persist together, hyperscalers cancel server deliveries for power or financing reasons, 2027-2028 memory volume estimates fall, or a yen surge triggers broad deleveraging.&lt;/p&gt;
&lt;h3 id="korea-specific-transmission"&gt;Korea-specific transmission
&lt;/h3&gt;&lt;p&gt;Korea combines heavy semiconductor weight with high foreign-flow sensitivity.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;High oil worsens import prices and terms of trade.&lt;/li&gt;
&lt;li&gt;High U.S. yields weaken the won and raise foreign required returns.&lt;/li&gt;
&lt;li&gt;A rapid yen rally can unwind yen-funded positions.&lt;/li&gt;
&lt;li&gt;A gradual yen rally can improve pricing competition for autos and machinery.&lt;/li&gt;
&lt;li&gt;Continued AI demand supports semiconductor exports and earnings.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;One-direction labels such as “yen strength is good for Korea” are therefore insufficient. Speed and sector-level earnings attribution matter.&lt;/p&gt;
&lt;h2 id="9-operational-checklist"&gt;9. Operational checklist
&lt;/h2&gt;&lt;p&gt;Before Monday&amp;rsquo;s open:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Check whether Brent gaps above $90.&lt;/li&gt;
&lt;li&gt;Verify weekend attacks and actual Hormuz cargo flow.&lt;/li&gt;
&lt;li&gt;Watch for USD/JPY breaking below 157 abruptly.&lt;/li&gt;
&lt;li&gt;Check whether Treasury futures reprice the 10-year above 4.75%.&lt;/li&gt;
&lt;li&gt;Monitor whether Korean foreign futures and cash selling resume together.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Macro stabilization begins when at least three of four conditions hold:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Brent stays below $85 for three sessions.&lt;/li&gt;
&lt;li&gt;The U.S. 10-year returns below 4.60%.&lt;/li&gt;
&lt;li&gt;Daily USD/JPY volatility falls below 1%.&lt;/li&gt;
&lt;li&gt;Big-tech and high-yield credit spreads narrow.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;If two of Brent above $100, the 10-year above 5%, and a 3% daily USD/JPY decline occur together, deleveraging should be treated as the immediate risk even when micro fundamentals remain constructive.&lt;/p&gt;
&lt;h2 id="10-red-team"&gt;10. Red team
&lt;/h2&gt;&lt;p&gt;This framework can fail in four ways.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Oil can rise while long yields fall if the shock destroys demand quickly and recession risk dominates inflation.&lt;/li&gt;
&lt;li&gt;The yen can strengthen in an orderly, well-signaled process without a 2024-style carry unwind.&lt;/li&gt;
&lt;li&gt;Treasury maturity management and buybacks can improve liquidity without controlling the structural level of long yields.&lt;/li&gt;
&lt;li&gt;AI revenue and memory earnings can grow fast enough to overwhelm a 4.8-5.0% discount rate, making macro stability a condition for multiple expansion rather than a condition for price gains.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="11-final-judgment"&gt;11. Final judgment
&lt;/h2&gt;&lt;p&gt;The end-July rebound had better quality than a simple oversold bounce. Earnings estimates held, and a major forced seller appears to have exited its public-equity book. The micro floor improved.&lt;/p&gt;
&lt;p&gt;The macro ceiling, however, became heavier. Oil returned to around $90, U.S. long yields approached 5%, three Fed voters preferred a hike, the BOJ explicitly identified upside inflation risks, the yen began a rapid reversal, and Treasury&amp;rsquo;s long-end supply strategy faces an August 5 test.&lt;/p&gt;
&lt;p&gt;The correct early-August question is not whether the earnings evidence was real. It was. The question is whether discount rates and funding markets stabilize enough to let that evidence be capitalized into prices.&lt;/p&gt;
&lt;p&gt;Four gauges will answer it:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Brent crude
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;U.S. 10-year yield
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;USD/JPY volatility
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;big-tech and high-yield credit spreads
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;When all four ease together, the July rebound can become a trend. If one accelerates again, the rebound may remain a high-volatility repair rather than a durable rerating.&lt;/p&gt;
&lt;h2 id="12-evidence-status"&gt;12. Evidence status
&lt;/h2&gt;&lt;p&gt;&lt;strong&gt;Facts:&lt;/strong&gt; Oil prices use Yahoo Finance daily futures data through July 31. Treasury yields use the official U.S. Treasury curve. Fed and BOJ decisions use official statements. The coming calendar uses BLS, BEA, ISM, and Treasury schedules.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Judgments:&lt;/strong&gt; One-month probabilities and ranges are scenario estimates as of August 1. “Macro locks” and “the second mechanical risk” are analytical transmission frameworks.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Blocked:&lt;/strong&gt; The exact size of global yen carry positions, official U.S. intervention activity, the final Middle East mediation outcome, Monday&amp;rsquo;s September Fed pricing, and the counterfactual oil price without the IEA release cannot be observed from public data.&lt;/p&gt;
&lt;p&gt;This report is market analysis, not personalized investment advice. Conflict, oil, and rate pricing can change quickly and should be rechecked before Monday&amp;rsquo;s open.&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://finance.yahoo.com/quote/CL%3DF/history/" target="_blank" rel="noopener"
 &gt;Yahoo Finance WTI futures history&lt;/a&gt; and &lt;a class="link" href="https://finance.yahoo.com/quote/BZ%3DF/history/" target="_blank" rel="noopener"
 &gt;Brent futures history&lt;/a&gt;, July 31, 2026.&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://home.treasury.gov/resource-center/data-chart-center/interest-rates/TextView?type=daily_treasury_yield_curve&amp;amp;field_tdr_date_value=2026" target="_blank" rel="noopener"
 &gt;U.S. Treasury Daily Treasury Par Yield Curve Rates&lt;/a&gt;, July 31, 2026.&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.boj.or.jp/en/mopo/mpmdeci/mpr_2026/k260731a.pdf" target="_blank" rel="noopener"
 &gt;Bank of Japan, Statement on Monetary Policy, July 31, 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://apnews.com/article/ad9fa27d5b1b5fd51e30d923ee738238" target="_blank" rel="noopener"
 &gt;AP, U.S.-Iran ceasefire mediation&lt;/a&gt; and &lt;a class="link" href="https://apnews.com/article/d57e675a7be6dbdd34561909ced240d0" target="_blank" rel="noopener"
 &gt;renewed fighting&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://apnews.com/article/9304893b474a48d18913d6f70fae4a8e" target="_blank" rel="noopener"
 &gt;AP, July 31, 2026, renewed U.S.-Iran tensions&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.iea.org/news/iea-member-countries-to-carry-out-largest-ever-oil-stock-release-amid-market-disruptions-from-middle-east-conflict" target="_blank" rel="noopener"
 &gt;IEA, Largest-ever oil stock release, March 11, 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:7"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.federalreserve.gov/newsevents/pressreleases/monetary20260729a.htm" target="_blank" rel="noopener"
 &gt;Federal Reserve, FOMC statement, July 29, 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:7" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:7" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:8"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.cmegroup.com/markets/interest-rates/cme-fedwatch-tool.html" target="_blank" rel="noopener"
 &gt;CME FedWatch Tool&lt;/a&gt; and &lt;a class="link" href="https://www.kiplinger.com/investing/live/fed-meeting-updates-and-commentary-july-2026" target="_blank" rel="noopener"
 &gt;Kiplinger July FOMC live updates&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:8" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:8" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:9"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.investing.com/news/economy-news/bofa-forecasts-75-bps-of-rate-hikes-in-2026-on-labour-market-resilience-new-fed-chair-4752311" target="_blank" rel="noopener"
 &gt;Reuters republished, BofA 75bp hike forecast for 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:9" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:10"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.boj.or.jp/en/mopo/outlook/gor2607a.pdf" target="_blank" rel="noopener"
 &gt;Bank of Japan, Outlook for Economic Activity and Prices, July 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:10" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:10" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:11"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.boj.or.jp/en/mopo/mpmsche_minu/m_ref/mref250731a.pdf" target="_blank" rel="noopener"
 &gt;Bank of Japan, Scheduled Dates of Monetary Policy Meetings in 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:11" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:11" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:12"&gt;
&lt;p&gt;&lt;a class="link" href="https://apnews.com/article/caf3b24d92688568f9c4f95725c87e55" target="_blank" rel="noopener"
 &gt;AP, U.S. Q2 GDP and June PCE&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:12" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:13"&gt;
&lt;p&gt;&lt;a class="link" href="https://finance.yahoo.com/quote/JPY%3DX/history/" target="_blank" rel="noopener"
 &gt;Yahoo Finance USD/JPY history&lt;/a&gt;, July 29-August 1, 2026.&amp;#160;&lt;a href="#fnref:13" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:14"&gt;
&lt;p&gt;&lt;a class="link" href="https://home.treasury.gov/news/press-releases/sb0489" target="_blank" rel="noopener"
 &gt;U.S. Treasury, May 2026 Quarterly Refunding Statement&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:14" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:14" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:15"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.axios.com/2026/07/31/bank-of-japan-rates-yen" target="_blank" rel="noopener"
 &gt;Axios, U.S. Treasury and yen intervention discussions, July 31, 2026&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:15" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:16"&gt;
&lt;p&gt;&lt;a class="link" href="https://home.treasury.gov/news/press-releases/sb0574" target="_blank" rel="noopener"
 &gt;U.S. Treasury, July 2026 Macroeconomic and Foreign Exchange Policies Report&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:16" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:17"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.bls.gov/schedule/2026/home.htm" target="_blank" rel="noopener"
 &gt;U.S. Bureau of Labor Statistics, 2026 release calendar&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:17" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:18"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.bea.gov/news/schedule/" target="_blank" rel="noopener"
 &gt;U.S. Bureau of Economic Analysis, release schedule&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:18" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:19"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.ismworld.org/supply-management-news-and-reports/reports/rob-report-calendar/" target="_blank" rel="noopener"
 &gt;Institute for Supply Management, Report On Business calendar&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:19" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>Who Was Buying on Non-Arbitrage Inflow Days? A 16-Stock KOSPI 200 Flow Screen</title><link>https://koreainvestinsights.com/post/kospi200-non-arbitrage-spot-real-money-cross-screen-2026-07-31/</link><pubDate>Fri, 31 Jul 2026 12:40:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/kospi200-non-arbitrage-spot-real-money-cross-screen-2026-07-31/</guid><description>&lt;p&gt;Futures are useful for reading broad market direction, but they are noisy tools for finding individual stocks. Index futures mix directional trades, cash-market hedges, basis arbitrage, options delta adjustments, and cross-market relative value. Foreign cash buying and stock-level purchases by investment trusts and pensions provide a clearer view of actual security selection.&lt;/p&gt;
&lt;p&gt;This screen starts from that distinction. It first identifies the ten trading days between July 1 and July 30, 2026 when KOSPI non-arbitrage program flow was positive. It then retains KOSPI 200 stocks for which foreign cash flow, single-stock program flow, investment-trust flow, and pension flow were all positive both across those ten days in aggregate and across the full July window.&lt;/p&gt;
&lt;p&gt;Sixteen stocks passed. We then compared FY2026 operating-profit consensus on June 30 and July 30 to distinguish confirmed earnings candidates from cyclical revisions, early turnaround buying, and value or shareholder-return trades.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Context: This is a follow-up to our &lt;a class="link" href="https://koreainvestinsights.com/en/post/real-money-flow-framework-korea-institution-quality-2026-06-03/" &gt;real-money flow framework&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/en/post/kr-flow-accumulation-absorption-screen-2026-05-15/" &gt;accumulation and absorption screen&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/en/post/korea-nonsemiconductor-ytd-leaders-kospi-kosdaq-breadth-2026-07-10/" &gt;Korean non-semiconductor breadth study&lt;/a&gt;. Related research is available in the &lt;a class="link" href="https://koreainvestinsights.com/en/page/korea-daily-market-hub/" &gt;Korea Daily Market hub&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/en/page/exclusive-analysis-hub/" &gt;Exclusive Analysis hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;KOSPI non-arbitrage flow was positive on ten days from July 1 to July 30. Aggregate inflow on those days was approximately KRW 10.83 trillion.&lt;/li&gt;
&lt;li&gt;Sixteen KOSPI 200 stocks had positive foreign cash, stock-program, investment-trust, and pension flow both on those ten days in aggregate and across the full month.&lt;/li&gt;
&lt;li&gt;The most interesting entry asymmetries are HMM, Kakao, Korea Gas Corporation, and Nongshim. HMM had the strongest earnings revision, but it remains a freight-rate-sensitive cyclical trade.&lt;/li&gt;
&lt;li&gt;Celltrion, Samyang Foods, LG Electronics, Amorepacific, and Nongshim offer the best alignment between flows and fundamentals. Celltrion and Samyang Foods are strong leaders, making pullbacks preferable to chasing.&lt;/li&gt;
&lt;li&gt;NAVER and LG Electronics attracted substantial cumulative buying despite sharp price declines. This may be absorption, but it may also mean that existing buyers are underwater. Recent three-to-five-day re-accumulation is required.&lt;/li&gt;
&lt;li&gt;Single-stock program data are not separated into arbitrage and non-arbitrage components. The market-level non-arbitrage series is a gate; stock-level flow is total program flow.&lt;/li&gt;
&lt;li&gt;A date-format duplication was found in Kia&amp;rsquo;s stock-program records. After keeping the latest observation per trading day, July program buying was approximately KRW 192.8 billion rather than KRW 177.0 billion.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 HMM is the strongest cyclical revision candidate, Kakao is an expectations-led setup, Korea Gas Corporation is a policy and dividend option, and Nongshim is an early earnings transition. Celltrion and Samyang Foods are higher-quality businesses, but their entry points require more patience.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-screening-methodology"&gt;1. Screening methodology
&lt;/h2&gt;&lt;h3 id="why-exclude-futures"&gt;Why exclude futures?
&lt;/h3&gt;&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Futures flow
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;├─ market-direction exposure
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;├─ cash-portfolio hedging
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;├─ cash-futures arbitrage
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;├─ options delta and gamma hedging
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;└─ country and sector relative value
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;A foreign futures purchase can represent a bullish view, a hedge against a short cash book, or an options adjustment. Stock-level cash and institutional data are better suited to security selection.&lt;/p&gt;
&lt;h3 id="four-filters"&gt;Four filters
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Step&lt;/th&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th&gt;Purpose&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1&lt;/td&gt;
 &lt;td&gt;Identify positive KOSPI non-arbitrage program-flow days from July 1 to July 30&lt;/td&gt;
 &lt;td&gt;Find sessions with basket and passive inflow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2&lt;/td&gt;
 &lt;td&gt;Require positive aggregate foreign cash, stock-program, trust, and pension flow across those ten sessions&lt;/td&gt;
 &lt;td&gt;Find securities selected during inflow days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3&lt;/td&gt;
 &lt;td&gt;Require all four flows to remain positive for the full July window&lt;/td&gt;
 &lt;td&gt;Exclude one-day buying&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;4&lt;/td&gt;
 &lt;td&gt;Limit the universe to the July 30 KODEX 200 constituent snapshot&lt;/td&gt;
 &lt;td&gt;Improve liquidity and comparability&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;“Positive on the ten sessions” means positive in aggregate across the ten dates. It does not mean that all four buyers were positive on every single date. Actual four-way positive-session counts ranged from one to seven.&lt;/p&gt;
&lt;h3 id="positive-non-arbitrage-sessions"&gt;Positive non-arbitrage sessions
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Date&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KOSPI non-arbitrage flow&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;July 8&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 800.2B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 9&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1,247.8B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 14&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 609.8B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 15&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1,818.1B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 680.0B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 21&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 847.9B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 22&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1,524.3B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 23&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1,800.1B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 142.7B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1,354.6B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Total&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+KRW 10,825.5B&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-data-validation-and-corrections"&gt;2. Data validation and corrections
&lt;/h2&gt;&lt;p&gt;The local database reproduced exactly 16 passing securities. Foreign, investment-trust, and pension data came from Kiwoom&amp;rsquo;s stock-level investor series. Stock-program and market non-arbitrage data came from Kiwoom REST. The KOSPI 200 universe used the July 30 Naver KODEX 200 constituent snapshot.&lt;/p&gt;
&lt;p&gt;Three adjustments matter:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Single-stock program flow is total program flow, not official stock-level non-arbitrage flow.&lt;/li&gt;
&lt;li&gt;Kia had duplicated records for several early-July dates stored in two timestamp formats. Deduplication raises its July program total from KRW 177.0 billion to KRW 192.8 billion.&lt;/li&gt;
&lt;li&gt;Returns were recalculated consistently from the July 1 close to the July 30 close. Five-day returns use July 23 to July 30. This changes Samyang Foods to +7.0%, Kia to -14.6%, and Nongshim to +13.5%.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;July 31 intraday prices are excluded because flow and consensus data end on July 30.&lt;/p&gt;
&lt;h2 id="3-sixteen-passing-stocks"&gt;3. Sixteen passing stocks
&lt;/h2&gt;&lt;p&gt;Flow values are in KRW 100 million.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Program&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Trust&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Pension&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Jul. 1-30&lt;/th&gt;
 &lt;th style="text-align: right"&gt;5D&lt;/th&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Innovation&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+420&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+796&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2,670&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+19.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.8%&lt;/td&gt;
 &lt;td&gt;Pension-led, needs stabilization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,386&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,872&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+93&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+405&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-23.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-21.1%&lt;/td&gt;
 &lt;td&gt;Absorption or damaged trend&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samyang Foods&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+943&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,506&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+740&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+431&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.3%&lt;/td&gt;
 &lt;td&gt;Strong leader&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Celltrion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+105&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+443&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+725&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2,116&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+8.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.3%&lt;/td&gt;
 &lt;td&gt;Strong leader&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAVER&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+208&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+968&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+659&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,367&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.5%&lt;/td&gt;
 &lt;td&gt;Flow not reflected in price&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kia&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,324&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+1,928&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+61&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-14.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-19.4%&lt;/td&gt;
 &lt;td&gt;Trend repair required&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kakao&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+980&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,253&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+75&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+78&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.7%&lt;/td&gt;
 &lt;td&gt;Early accumulation candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HMM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+789&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+859&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+148&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+511&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.7%&lt;/td&gt;
 &lt;td&gt;Pullback candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amorepacific&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+656&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+491&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+266&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+618&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.6%&lt;/td&gt;
 &lt;td&gt;Strong but extended&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG H&amp;amp;H&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+584&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+688&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+62&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+401&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.7%&lt;/td&gt;
 &lt;td&gt;Short-term overextension&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai Glovis&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+722&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+553&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+391&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.3%&lt;/td&gt;
 &lt;td&gt;Needs stabilization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;IBK&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+257&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+418&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+102&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.9%&lt;/td&gt;
 &lt;td&gt;Quiet value accumulation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nongshim&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+194&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+194&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+106&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+132&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+13.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.3%&lt;/td&gt;
 &lt;td&gt;Trend-transition candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Gas Corp.&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+184&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+191&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.1%&lt;/td&gt;
 &lt;td&gt;Low-heat accumulation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pan Ocean&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+85&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+168&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+66&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.9%&lt;/td&gt;
 &lt;td&gt;Needs trend recovery&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Poongsan&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+80&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+93&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.1%&lt;/td&gt;
 &lt;td&gt;Still in a downtrend&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="4-buying-on-the-ten-inflow-days"&gt;4. Buying on the ten inflow days
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Program&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Trust&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Pension&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Four-way positive days&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Innovation&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+291&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+443&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+396&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,344&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,137&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,421&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+537&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samyang Foods&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+167&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+456&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+414&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+79&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Celltrion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+236&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+572&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+305&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+692&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAVER&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+467&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,072&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+262&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+808&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kia&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+567&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,771&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+90&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kakao&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+496&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+808&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+35&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+74&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HMM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+564&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+273&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;7&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amorepacific&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+600&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+523&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+58&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG H&amp;amp;H&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+405&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+456&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+49&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+260&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai Glovis&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+334&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+262&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+196&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;IBK&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+115&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+41&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nongshim&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+57&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+60&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;5&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Gas Corp.&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+84&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+87&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+13&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pan Ocean&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+72&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+123&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+8&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Poongsan&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;HMM had seven four-way positive sessions, the highest in the group. Nongshim followed with five, while Korea Gas Corporation and Poongsan had four. This gives HMM the strongest combination of aggregate buying and repetition.&lt;/p&gt;
&lt;h2 id="5-fy2026-operating-profit-revisions"&gt;5. FY2026 operating-profit revisions
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;June 30&lt;/th&gt;
 &lt;th style="text-align: right"&gt;July 30&lt;/th&gt;
 &lt;th style="text-align: right"&gt;One-month change&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Innovation&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.922T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 5.460T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+10.9%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Revision aligns with pension buying&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 3.888T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.238T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+9.0%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Earnings and foreign flow align&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samyang Foods&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 726.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 733.8B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.0%&lt;/td&gt;
 &lt;td&gt;High estimates maintained&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Celltrion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.756T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.847T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+5.2%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Confirmed earnings improvement&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAVER&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.419T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.387T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.3%&lt;/td&gt;
 &lt;td&gt;Flow leads earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kia&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 10.129T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 10.234T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.0%&lt;/td&gt;
 &lt;td&gt;Value and shareholder return&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kakao&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 939.8B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 947.4B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.8%&lt;/td&gt;
 &lt;td&gt;Expectations lead&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HMM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.038T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.368T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+31.8%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Strongest cyclical revision&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amorepacific&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 460.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 465.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.1%&lt;/td&gt;
 &lt;td&gt;Global-channel growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG H&amp;amp;H&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 333.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 344.4B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.2%&lt;/td&gt;
 &lt;td&gt;Early bottoming&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai Glovis&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.182T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.149T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.5%&lt;/td&gt;
 &lt;td&gt;Buying ahead of recovery&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;IBK&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 3.703T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 3.686T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.5%&lt;/td&gt;
 &lt;td&gt;Dividend and value&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nongshim&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 213.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 216.8B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.6%&lt;/td&gt;
 &lt;td&gt;Early overseas growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Gas Corp.&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.265T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.265T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.0%&lt;/td&gt;
 &lt;td&gt;Policy and dividend option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pan Ocean&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 568.9B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 594.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+4.5%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Freight recovery&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Poongsan&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 360.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 359.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.3%&lt;/td&gt;
 &lt;td&gt;Copper and defense expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Consensus values are Naver aggregates stored in the local database. Analyst counts were unavailable, so revision breadth may differ materially by stock.&lt;/p&gt;
&lt;h2 id="6-five-fundamental-groups"&gt;6. Five fundamental groups
&lt;/h2&gt;&lt;h3 id="structural-earnings"&gt;Structural earnings
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Samyang Foods&lt;/li&gt;
&lt;li&gt;Celltrion&lt;/li&gt;
&lt;li&gt;LG Electronics&lt;/li&gt;
&lt;li&gt;Amorepacific&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Celltrion reported Q2 revenue of KRW 1.394 trillion and operating profit of KRW 451.8 billion, supported by high-margin new products and cost improvement. &lt;a class="link" href="https://www.celltrion.com/ko-kr/company/media-center/press-release/4862" target="_blank" rel="noopener"
 &gt;Celltrion&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;LG Electronics reported Q2 revenue growth of 14.9% and operating-profit growth of 146.9%, while continuing investment in AI data-center cooling. &lt;a class="link" href="https://www.lge.co.kr/story/newsroom/236115" target="_blank" rel="noopener"
 &gt;LG Electronics&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Amorepacific&amp;rsquo;s Amazon Prime Day revenue rose 20% in the U.S. and 22% in Europe. &lt;a class="link" href="https://www.apgroup.com/int/en/news/2026-07-03-1.html" target="_blank" rel="noopener"
 &gt;Amorepacific&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="rapid-cyclical-revisions"&gt;Rapid cyclical revisions
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;HMM&lt;/li&gt;
&lt;li&gt;SK Innovation&lt;/li&gt;
&lt;li&gt;Pan Ocean&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;HMM is the clearest case. Its FY2026 operating-profit estimate increased 31.8% as SCFI and CCFI rose. The same operating leverage can reverse when freight rates normalize. &lt;a class="link" href="https://www.truefriend.com/main/research/research/StrategyDetail.jsp?id=157282&amp;amp;jkGubun=6" target="_blank" rel="noopener"
 &gt;Korea Investment &amp;amp; Securities&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="turnaround-accumulation"&gt;Turnaround accumulation
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;LG H&amp;amp;H&lt;/li&gt;
&lt;li&gt;Nongshim&lt;/li&gt;
&lt;li&gt;NAVER&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Nongshim offers the clearest near-term earnings path. NAVER offers the largest optionality but also the largest gap between usage metrics and current profit. NAVER said AI Tab daily queries rose sevenfold from beta and queries per user rose 1.7 times. &lt;a class="link" href="https://www.navercorp.com/media/pressReleasesDetail?seq=10034477" target="_blank" rel="noopener"
 &gt;NAVER&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="value-and-shareholder-return"&gt;Value and shareholder return
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Kia&lt;/li&gt;
&lt;li&gt;IBK&lt;/li&gt;
&lt;li&gt;Korea Gas Corporation&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The central variables are valuation, dividends, and policy normalization rather than aggressive revenue growth.&lt;/p&gt;
&lt;h3 id="expectations-ahead-of-numbers"&gt;Expectations ahead of numbers
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Kakao&lt;/li&gt;
&lt;li&gt;Hyundai Glovis&lt;/li&gt;
&lt;li&gt;Poongsan&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;These stocks require future earnings to catch up with current positioning. Flow can reverse quickly if AI monetization, fuel-cost pass-through, copper, or defense revenue disappoints.&lt;/p&gt;
&lt;h2 id="7-two-separate-rankings"&gt;7. Two separate rankings
&lt;/h2&gt;&lt;h3 id="best-alignment-between-flow-and-fundamentals"&gt;Best alignment between flow and fundamentals
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Celltrion&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Samyang Foods&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;LG Electronics&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Amorepacific&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Nongshim&lt;/strong&gt;&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;This is a quality ranking, not an entry-price ranking.&lt;/p&gt;
&lt;h3 id="best-current-asymmetry"&gt;Best current asymmetry
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;HMM&lt;/strong&gt;: strongest repetition and revision, but cyclical.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Kakao&lt;/strong&gt;: large flow with limited price reflection, but profit confirmation is weak.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Korea Gas Corporation&lt;/strong&gt;: low-heat accumulation and policy optionality.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Nongshim&lt;/strong&gt;: repeated buying with an improving overseas earnings path.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Samyang Foods and Celltrion&lt;/strong&gt;: strongest leaders, better bought on pullbacks.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="8-key-candidates"&gt;8. Key candidates
&lt;/h2&gt;&lt;h3 id="hmm"&gt;HMM
&lt;/h3&gt;&lt;p&gt;HMM combined positive foreign, program, trust, and pension flow with seven four-way positive sessions. The 31.8% operating-profit revision is the strongest in the screen.&lt;/p&gt;
&lt;p&gt;The risk is duration. Freight earnings can peak quickly, and HMM should be treated as a revision trade rather than a permanent compounder.&lt;/p&gt;
&lt;h3 id="kakao"&gt;Kakao
&lt;/h3&gt;&lt;p&gt;Kakao attracted KRW 98.0 billion of foreign buying and KRW 125.3 billion of program buying while rising only 3.9% over the period. Its operating-profit revision was only +0.8%, and trust and pension buying were modest.&lt;/p&gt;
&lt;p&gt;The position is buying monetization before it appears in earnings. Evidence must emerge through advertising, commerce, subscriptions, or cost efficiency.&lt;/p&gt;
&lt;h3 id="korea-gas-corporation"&gt;Korea Gas Corporation
&lt;/h3&gt;&lt;p&gt;All four flows were positive without price overextension. Earnings estimates were unchanged, making this a policy, receivables, and dividend normalization trade rather than an earnings-growth trade.&lt;/p&gt;
&lt;h3 id="nongshim"&gt;Nongshim
&lt;/h3&gt;&lt;p&gt;Nongshim recorded five four-way positive sessions and balanced monthly buying. Overseas growth and cost control offer a clearer earnings bridge than most turnaround candidates, although the July return already reached 13.5%.&lt;/p&gt;
&lt;h3 id="samyang-foods-and-celltrion"&gt;Samyang Foods and Celltrion
&lt;/h3&gt;&lt;p&gt;These are the strongest businesses in the screen, but both gained 10.3% over the final five sessions. Pullback behavior and renewed institutional buying matter more than chasing.&lt;/p&gt;
&lt;h2 id="9-absorption-or-trapped-buyers"&gt;9. Absorption or trapped buyers?
&lt;/h2&gt;&lt;p&gt;LG Electronics recorded strong foreign and program buying and a 9.0% earnings revision, yet fell 23.4% from July 1 to July 30. NAVER had strong pension and trust buying while earnings estimates fell 1.3% and the stock lost 11.5% over five sessions.&lt;/p&gt;
&lt;p&gt;There are two possible interpretations:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Bull case: institutions absorbed a large seller
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Bear case: existing buyers are underwater and may sell into rebounds
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Cumulative monthly flow cannot distinguish them. A higher low and renewed three-to-five-day foreign and real-money buying are required.&lt;/p&gt;
&lt;h2 id="10-execution-framework"&gt;10. Execution framework
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;Confirmation&lt;/th&gt;
 &lt;th&gt;Invalidation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HMM&lt;/td&gt;
 &lt;td&gt;Strong SCFI/CCFI and renewed foreign-pension buying&lt;/td&gt;
 &lt;td&gt;Freight collapse and downward revisions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kakao&lt;/td&gt;
 &lt;td&gt;Renewed foreign-trust buying and AI revenue metrics&lt;/td&gt;
 &lt;td&gt;Program flow remains while real money exits&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Gas Corp.&lt;/td&gt;
 &lt;td&gt;Receivables decline, tariff and dividend normalization&lt;/td&gt;
 &lt;td&gt;Policy delays and pension selling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nongshim&lt;/td&gt;
 &lt;td&gt;Buying persists on a pullback&lt;/td&gt;
 &lt;td&gt;Overseas slowdown and downward revisions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samyang Foods and Celltrion&lt;/td&gt;
 &lt;td&gt;Overextension cools and institutions re-accumulate&lt;/td&gt;
 &lt;td&gt;Post-earnings distribution&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;LG Electronics, NAVER, and Kia need trend repair. Amorepacific and LG H&amp;amp;H need to digest strong July gains. Poongsan&amp;rsquo;s positive flow is too small relative to the downtrend to qualify as more than an early watch signal.&lt;/p&gt;
&lt;h2 id="11-red-team"&gt;11. Red Team
&lt;/h2&gt;&lt;h3 id="non-arbitrage-days-may-be-basket-days-not-active-selection"&gt;Non-arbitrage days may be basket days, not active selection
&lt;/h3&gt;&lt;p&gt;KOSPI 200 stocks can be bought mechanically through index and ETF baskets. Foreign and program buying does not prove that an active manager selected the business.&lt;/p&gt;
&lt;h3 id="trust-and-pension-flow-are-not-always-permanent-capital"&gt;Trust and pension flow are not always permanent capital
&lt;/h3&gt;&lt;p&gt;Investment trusts include index and ETF assets, while pensions rebalance and use external managers. This screen uses a narrow real-money definition of trusts plus pensions and excludes insurance and private funds.&lt;/p&gt;
&lt;h3 id="monthly-totals-can-hide-late-month-exits"&gt;Monthly totals can hide late-month exits
&lt;/h3&gt;&lt;p&gt;LG Electronics, Kia, NAVER, and SK Innovation suffered substantial late-month drawdowns. Positive monthly flow does not guarantee additional future demand.&lt;/p&gt;
&lt;h3 id="consensus-revisions-have-different-quality"&gt;Consensus revisions have different quality
&lt;/h3&gt;&lt;p&gt;HMM&amp;rsquo;s 31.8% cyclical revision is not directly comparable with Celltrion&amp;rsquo;s 5.2% post-earnings revision. Analyst counts and forecast dispersion were not available.&lt;/p&gt;
&lt;h3 id="july-31-is-excluded"&gt;July 31 is excluded
&lt;/h3&gt;&lt;p&gt;The data cutoff is July 30. A large July 31 intraday move does not validate the screen before final investor flows are collected.&lt;/p&gt;
&lt;h2 id="12-data-coverage"&gt;12. Data coverage
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Data&lt;/th&gt;
 &lt;th&gt;Provider&lt;/th&gt;
 &lt;th&gt;Date&lt;/th&gt;
 &lt;th&gt;Limitation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign, trust, pension flow&lt;/td&gt;
 &lt;td&gt;Kiwoom local DB&lt;/td&gt;
 &lt;td&gt;July 30, 2026&lt;/td&gt;
 &lt;td&gt;Provider-verified, separate KRX check not available&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stock-program flow&lt;/td&gt;
 &lt;td&gt;Kiwoom REST &lt;code&gt;ka90013&lt;/code&gt;&lt;/td&gt;
 &lt;td&gt;July 30&lt;/td&gt;
 &lt;td&gt;Total program, not stock-level non-arbitrage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Market arbitrage and non-arbitrage&lt;/td&gt;
 &lt;td&gt;Kiwoom REST &lt;code&gt;ka90007&lt;/code&gt;&lt;/td&gt;
 &lt;td&gt;July 30&lt;/td&gt;
 &lt;td&gt;Used only as the market gate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI 200 universe&lt;/td&gt;
 &lt;td&gt;Naver KODEX 200 snapshot&lt;/td&gt;
 &lt;td&gt;July 30&lt;/td&gt;
 &lt;td&gt;ETF-constituent proxy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Prices and returns&lt;/td&gt;
 &lt;td&gt;Naver domestic chart&lt;/td&gt;
 &lt;td&gt;July 30&lt;/td&gt;
 &lt;td&gt;July 1 and July 23 close comparisons&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2026 operating-profit consensus&lt;/td&gt;
 &lt;td&gt;Naver consensus local DB&lt;/td&gt;
 &lt;td&gt;June 30 and July 30&lt;/td&gt;
 &lt;td&gt;Analyst count unavailable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;This cross-screen identifies stocks where foreign cash, program, trust, and pension buying overlapped during market-level non-arbitrage inflow days. It is better suited to security selection than a screen that adds futures, but it does not convert program flow into stock-level non-arbitrage evidence.&lt;/p&gt;
&lt;p&gt;The 16 names are not equivalent. Celltrion, Samyang Foods, LG Electronics, and Amorepacific are earnings-backed. HMM, SK Innovation, and Pan Ocean are cyclical revision trades. NAVER, Kakao, and Hyundai Glovis are expectations-led. Kia, IBK, and Korea Gas Corporation are value or policy trades.&lt;/p&gt;
&lt;p&gt;HMM, Kakao, Korea Gas Corporation, and Nongshim offer the clearest current asymmetry. Celltrion and Samyang Foods offer higher business quality but less comfortable entry points. The main lesson is to separate the best company from the best current price.&lt;/p&gt;
&lt;p&gt;This report is market-flow research based on public and local data. It is not a recommendation to buy or sell any security.&lt;/p&gt;</description></item><item><title>Did One Hedge Fund Cause the July AI Rout? Fact-Checking Situational Awareness and the SK hynix ADR</title><link>https://koreainvestinsights.com/post/situational-awareness-liquidation-sk-hynix-adr-july-ai-rout-fact-check-2026-07-31/</link><pubDate>Fri, 31 Jul 2026 12:15:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/situational-awareness-liquidation-sk-hynix-adr-july-ai-rout-fact-check-2026-07-31/</guid><description>&lt;p&gt;A compelling explanation spread after AI infrastructure stocks staged an extraordinary rebound at the end of July 2026. The story was that Situational Awareness, the hedge fund led by former OpenAI researcher Leopold Aschenbrenner, had taken a $7 billion cornerstone position in the SK hynix ADR, received margin calls, and transferred a $16 billion public-equity book to Citadel. The fund&amp;rsquo;s liquidation supposedly explained both the July selloff and the July 31 rebound.&lt;/p&gt;
&lt;p&gt;That narrative mixes confirmed facts, credible reporting, a material misreading of the SEC prospectus, and causal claims that cannot yet be proved. The best-supported conclusion is narrower: a large transfer of Situational Awareness&amp;rsquo;s public-equity book appears to have occurred, and forced deleveraging likely amplified moves in crowded AI infrastructure stocks. But the fund did not publicly commit to a $7 billion SK hynix allocation by itself, and no public dataset can attribute 40% to 50% of the KOSPI decline to one fund.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Data cutoff: U.S. returns use the July 30, 2026 close. Korean prices are intraday snapshots at 12:12 p.m. KST on July 31. Private-fund AUM, leverage, prime-broker relationships, and margin calls have limited public disclosure, so this report separates SEC filings from media reports and inference. It follows our &lt;a class="link" href="https://koreainvestinsights.com/en/post/china-duv-steelman-verdict-black-tuesday-korea-memory-2026-07-28/" &gt;July 28 Korean memory selloff analysis&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/en/post/ai-capex-eight-doubts-rebound-or-terminal-48hours-2026-07-29/" &gt;eight questions behind the AI CAPEX correction&lt;/a&gt;. More research is available in the &lt;a class="link" href="https://koreainvestinsights.com/en/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI and HBM hub&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/en/page/exclusive-analysis-hub/" &gt;Exclusive Analysis hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;The Wall Street Journal and Axios reported that Situational Awareness sold most or all of its public-equity portfolio to Citadel after large losses. The direction of the transaction is supported by credible reporting, although the exact value and terms are not publicly filed.&lt;/li&gt;
&lt;li&gt;The SEC prospectus does not say Situational Awareness alone received a $7 billion SK hynix allocation. Baillie Gifford, Coatue-managed funds, and Situational Awareness collectively expressed non-binding interest in up to $7 billion. Individual allocations were not disclosed.&lt;/li&gt;
&lt;li&gt;Q1 2026 filings confirm meaningful exposure to Bloom Energy, CoreWeave, and SanDisk. A May Schedule 13G confirms a 5.6% Nebius stake. Micron and Nvidia exposure, however, included substantial options. The portfolio was not a simple long-only AI basket.&lt;/li&gt;
&lt;li&gt;On July 30, NBIS rose 27.13%, BE 26.49%, SNDK 25.99%, CRWV 21.51%, MU 18.36%, and SKHY 17.52%, versus 2.65% for NVDA. This is consistent with forced-seller removal and short covering, but it is not a controlled experiment proving a single-fund cause.&lt;/li&gt;
&lt;li&gt;A separate claim that Korea suffered KRW 37.7 trillion of forced margin liquidation is also incorrect. That figure appears to refer to outstanding margin-credit balances, not executed forced sales.&lt;/li&gt;
&lt;li&gt;The most defensible verdict is that Situational Awareness was likely an important marginal seller and amplifier inside a leveraged, crowded factor. Its book transfer can support short-term normalization, but it does not answer the fundamental question of AI and memory earnings after 2028.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Verdict&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The public-equity transfer is well reported. The alleged standalone $7 billion SK hynix allocation and a 40% to 50% attribution of the July decline are not established. This is best understood as a market-structure event that amplified price moves, not proof that fundamentals caused or ended the rout.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-claim-by-claim-verdict"&gt;1. Claim-by-claim verdict
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Claim&lt;/th&gt;
 &lt;th&gt;Verdict&lt;/th&gt;
 &lt;th&gt;What the evidence says&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Situational Awareness transferred its public-equity book to Citadel&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Largely supported&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;WSJ reported most of the book; Axios reported all of it&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AUM fell from $20 billion to $10 billion&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Reported, not independently filed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Bloomberg and follow-on reports cited the decline&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The fund used up to four times leverage&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Reported, not filed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Media reports cite the figure; 13F cannot establish leverage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Goldman Sachs, JPMorgan, and BofA issued margin calls&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Reported, not filed&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Prime-broker terms and margin-call records are private&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The fund alone bought $7 billion of SKHY&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;False&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;$7 billion was the aggregate, non-binding indication from three investors&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SKHY triggered the margin call&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Plausible but unproved&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Timing fits, but allocation, entry price, and collateral terms are unknown&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The fund was concentrated in SKHY, MU, CRWV, NBIS, BE, and SNDK&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Partly supported&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Q1 13F and May 13G support several names; SKHY was not yet listed; MU exposure was mixed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The fund did not own NVDA&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;False&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Q1 13F reported a small common position and large put exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The rebound was a controlled experiment proving causality&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Overstated&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Betas, options, shorts, earnings, and macro variables also moved&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea saw KRW 37.7 trillion of forced liquidation&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;False&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;The figure appears to be outstanding credit balances&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;One fund caused 40% to 50% of the July decline&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Not measurable&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Daily fund sales, swaps, hedges, and a counterfactual are unavailable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-the-central-correction-7-billion-was-not-one-funds-allocation"&gt;2. The central correction: $7 billion was not one fund&amp;rsquo;s allocation
&lt;/h2&gt;&lt;p&gt;SK hynix&amp;rsquo;s final SEC prospectus covered 177.9 million ADSs at $149 each, for a gross offering of roughly $26.5 billion and estimated net proceeds of about $26.2 billion.&lt;/p&gt;
&lt;p&gt;The prospectus named three investors:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Baillie Gifford Overseas Limited&lt;/li&gt;
&lt;li&gt;funds managed by Coatue&lt;/li&gt;
&lt;li&gt;Situational Awareness Partners LP&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;They had, severally and not jointly, indicated interest in purchasing up to an aggregate of $7 billion of ADSs. The indications were explicitly non-binding, and each investor could buy more, less, or none. &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119312526299963/d32785d424b4.htm" target="_blank" rel="noopener"
 &gt;SK hynix final SEC prospectus&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The public record therefore establishes only:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Aggregate indication ceiling from three investors: $7 billion
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Individual allocations: undisclosed
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Situational Awareness allocation: undisclosed
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Binding commitment: no
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Situational Awareness may have received a large allocation. It may also have received far less than $7 billion. Treating the aggregate indication as a single-fund position inflated every downstream conclusion about leverage, the margin call, and market impact.&lt;/p&gt;
&lt;h2 id="3-what-is-actually-known-about-the-fund"&gt;3. What is actually known about the fund
&lt;/h2&gt;&lt;h3 id="31-the-public-equity-transfer"&gt;3.1 The public-equity transfer
&lt;/h3&gt;&lt;p&gt;The &lt;a class="link" href="https://www.wsj.com/finance/citadel-buys-situational-awarenesss-stock-portfolio-after-big-losses-in-ai-5117159b" target="_blank" rel="noopener"
 &gt;Wall Street Journal&lt;/a&gt; reported that Situational Awareness sold the bulk of its stock portfolio to Citadel after deep AI-related losses. &lt;a class="link" href="https://www.axios.com/2026/07/30/ai-hedge-fund-situational-awareness-citadel" target="_blank" rel="noopener"
 &gt;Axios&lt;/a&gt; cited one source saying it sold the entire public-equity portfolio.&lt;/p&gt;
&lt;p&gt;The source discrepancy matters. The safest wording is that most or all of the public-equity book was transferred. That does not establish that the fund itself was liquidated. Private holdings and separate vehicles may remain, and no public announcement says the manager has ceased operations.&lt;/p&gt;
&lt;p&gt;Some secondary reports put the transferred book at $16 billion. There is no public transaction filing showing whether this figure was market value, gross exposure, or some other measure. The trade&amp;rsquo;s existence is more reliable than the exact amount.&lt;/p&gt;
&lt;h3 id="32-aum-returns-leverage-and-margin-calls"&gt;3.2 AUM, returns, leverage, and margin calls
&lt;/h3&gt;&lt;p&gt;Media reports described assets rising above $20 billion and then falling to about $10 billion. Other reports cited a 439% net return through June, more than 1,000% since inception, up to four times leverage, and margin calls from multiple prime brokers.&lt;/p&gt;
&lt;p&gt;These are reported figures, not values that can be reconstructed from public SEC filings. Form 13F does not disclose:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;net asset value&lt;/li&gt;
&lt;li&gt;cash and borrowings&lt;/li&gt;
&lt;li&gt;gross and net leverage&lt;/li&gt;
&lt;li&gt;OTC derivatives and swaps&lt;/li&gt;
&lt;li&gt;prime-broker collateral&lt;/li&gt;
&lt;li&gt;monthly P&amp;amp;L and redemptions&lt;/li&gt;
&lt;li&gt;private-asset valuations&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The media reporting may be accurate, but it should not be presented as audited public fact.&lt;/p&gt;
&lt;h2 id="4-what-the-13f-and-13g-filings-show"&gt;4. What the 13F and 13G filings show
&lt;/h2&gt;&lt;p&gt;Situational Awareness&amp;rsquo;s Q1 2026 13F contained 42 reportable entries with aggregate reported value of about $13.68 billion. Option entries can reflect the market value of underlying securities rather than option premium or delta exposure. The total is therefore neither AUM nor a measure of net risk. &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2045724/000204572426000008/salp13fq1xml.xml" target="_blank" rel="noopener"
 &gt;Situational Awareness Q1 2026 13F&lt;/a&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Security&lt;/th&gt;
 &lt;th style="text-align: right"&gt;March 31 filing&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bloom Energy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$878.7M common plus $55.3M call underlying value&lt;/td&gt;
 &lt;td&gt;Large directional exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CoreWeave&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$556.1M common plus $140.6M calls&lt;/td&gt;
 &lt;td&gt;Material AI infrastructure long&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SanDisk&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$724.4M common plus $388.8M calls&lt;/td&gt;
 &lt;td&gt;Material storage exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5.9M common, $422.3M calls, $583.7M puts&lt;/td&gt;
 &lt;td&gt;Mixed options book, not a clean long&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nvidia&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Small common position plus $1.568B of put underlying value&lt;/td&gt;
 &lt;td&gt;Material bearish option exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A more recent Schedule 13G reported 12,410,060 Nebius shares, or 5.6%, as of May 19. &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2045724/000093583626000303/primary_doc.xml" target="_blank" rel="noopener"
 &gt;Nebius Schedule 13G&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;These filings support the view that the fund had large AI infrastructure, power, and storage exposure. They also show why a simple held-versus-not-held table is insufficient. Micron included both calls and puts, and Nvidia was associated with substantial put exposure.&lt;/p&gt;
&lt;p&gt;SKHY cannot be verified from the Q1 filing because the ADR listed in July. A broad Q3 holdings check will not arrive until mid-November. A Schedule 13D or 13G could appear earlier if an investor crossed the 5% threshold, but current public filings do not disclose Situational Awareness&amp;rsquo;s allocation.&lt;/p&gt;
&lt;h2 id="5-the-rebound-strong-evidence-not-a-controlled-experiment"&gt;5. The rebound: strong evidence, not a controlled experiment
&lt;/h2&gt;&lt;h3 id="us-close-on-july-30"&gt;U.S. close on July 30
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Security&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Return&lt;/th&gt;
 &lt;th&gt;Public filing connection&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Nebius&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27.13%&lt;/td&gt;
 &lt;td&gt;5.6% May 13G stake&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bloom Energy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26.49%&lt;/td&gt;
 &lt;td&gt;Q1 common and calls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SanDisk&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+25.99%&lt;/td&gt;
 &lt;td&gt;Q1 common and calls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CoreWeave&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.51%&lt;/td&gt;
 &lt;td&gt;Q1 common and calls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.36%&lt;/td&gt;
 &lt;td&gt;Q1 common, calls, and puts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SKHY ADR&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.52%&lt;/td&gt;
 &lt;td&gt;Fund allocation undisclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SOXX&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+8.55%&lt;/td&gt;
 &lt;td&gt;Semiconductor benchmark&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nvidia&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.65%&lt;/td&gt;
 &lt;td&gt;Q1 common and substantial puts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="korea-intraday-at-1212-pm-kst-on-july-31"&gt;Korea intraday at 12:12 p.m. KST on July 31
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Asset&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Price&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Return&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 250,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.01%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,645,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24.43%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6,389.04&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14.22%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The cross-section is consistent with the removal of a forced seller. It may also reflect short covering and options hedging in the names that had suffered the largest declines.&lt;/p&gt;
&lt;p&gt;It is not a controlled experiment for five reasons:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;NBIS, BE, and CRWV have much higher beta than Nvidia.&lt;/li&gt;
&lt;li&gt;A block transfer to Citadel is different from all shares being sold in the open market.&lt;/li&gt;
&lt;li&gt;Short covering and dealer gamma can magnify the rebound.&lt;/li&gt;
&lt;li&gt;The fund&amp;rsquo;s net exposure is unknown because strikes, maturities, deltas, and swaps are not disclosed.&lt;/li&gt;
&lt;li&gt;FOMC news, hyperscaler earnings, macro risk appetite, and Korean program flows moved at the same time.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The strongest defensible conclusion is that the rebound supports a forced-flow amplification hypothesis. It does not identify the original trigger or measure one fund&amp;rsquo;s share of the decline.&lt;/p&gt;
&lt;h2 id="6-the-krw-377-trillion-margin-liquidation-error"&gt;6. The KRW 37.7 trillion margin-liquidation error
&lt;/h2&gt;&lt;p&gt;Outstanding margin credit, collateral shortfalls, forced-sale orders, and executed forced sales are different statistics.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Margin-credit balance&lt;/td&gt;
 &lt;td&gt;Borrowed purchase funding still outstanding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Collateral shortfall&lt;/td&gt;
 &lt;td&gt;Amount below required maintenance margin&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Forced-sale order&lt;/td&gt;
 &lt;td&gt;Broker order submitted for repayment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Executed forced sale&lt;/td&gt;
 &lt;td&gt;Portion of orders actually filled&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The publicly cited KRW 37.7 trillion figure appears to refer to outstanding margin-credit balances, not executed forced sales over July 24 to 28. Treating the entire stock of credit as liquidation volume materially exaggerates market impact.&lt;/p&gt;
&lt;p&gt;The same caution applies to ELS knock-in hedging. Futures selling alone cannot identify ELS hedges without product-level delta and gamma data.&lt;/p&gt;
&lt;h2 id="7-a-more-defensible-causal-chain"&gt;7. A more defensible causal chain
&lt;/h2&gt;&lt;p&gt;The following mechanism fits the evidence:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;High-beta AI infrastructure stocks decline
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ collateral capacity deteriorates in leveraged portfolios
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ public-equity exposure is reduced or block-transferred
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ crowded-factor pressure intensifies
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ removal of the urgent seller becomes public
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ short covering, dealer hedging, and high-beta rebound
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;What remains unknown is crucial:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Situational Awareness&amp;rsquo;s actual SKHY allocation and purchase price&lt;/li&gt;
&lt;li&gt;the position that triggered the first margin call&lt;/li&gt;
&lt;li&gt;prime-broker liquidation timing and size&lt;/li&gt;
&lt;li&gt;security-level quantities and hedges transferred to Citadel&lt;/li&gt;
&lt;li&gt;daily July net sales by the fund&lt;/li&gt;
&lt;li&gt;actual Korean forced-sale and ELS hedge execution&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;SKHY may have contributed to the stress. But the public record does not prove that it was the trigger. Losses on multiple AI infrastructure longs, software shorts moving against the fund, and derivatives collateral could have combined.&lt;/p&gt;
&lt;h2 id="8-why-a-40-to-50-market-attribution-is-not-possible"&gt;8. Why a 40% to 50% market attribution is not possible
&lt;/h2&gt;&lt;p&gt;To assign a share of the KOSPI or AI-factor decline to one fund, an analyst would need:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;security-level sale quantities and execution prices&lt;/li&gt;
&lt;li&gt;SKHY ADR versus 000660 arbitrage flows&lt;/li&gt;
&lt;li&gt;futures, options, and swap hedges&lt;/li&gt;
&lt;li&gt;executed Korean margin-liquidation data&lt;/li&gt;
&lt;li&gt;issuer-level ELS hedge data&lt;/li&gt;
&lt;li&gt;high-frequency event windows&lt;/li&gt;
&lt;li&gt;a credible counterfactual group&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;None is publicly complete. Larger drawdowns and rebounds in fund-linked names show an amplification channel, not a market-wide contribution percentage.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Factor&lt;/th&gt;
 &lt;th&gt;Current assessment&lt;/th&gt;
 &lt;th&gt;Confidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Forced reduction of leveraged AI infrastructure exposure&lt;/td&gt;
 &lt;td&gt;Meaningful amplifier&lt;/td&gt;
 &lt;td&gt;Medium-high&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SKHY as the single initial trigger&lt;/td&gt;
 &lt;td&gt;Possible, unproved&lt;/td&gt;
 &lt;td&gt;Medium-low&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean credit and derivative amplification&lt;/td&gt;
 &lt;td&gt;Direction plausible, size unknown&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China DUV, CXMT, and vendor-financing concerns&lt;/td&gt;
 &lt;td&gt;Supplied a narrative for repricing&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Macro, rates, and earnings&lt;/td&gt;
 &lt;td&gt;Independent concurrent drivers&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;One fund explaining half the KOSPI decline&lt;/td&gt;
 &lt;td&gt;Unsupported&lt;/td&gt;
 &lt;td&gt;Low&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="9-investment-implications"&gt;9. Investment implications
&lt;/h2&gt;&lt;h3 id="91-what-improved"&gt;9.1 What improved
&lt;/h3&gt;&lt;p&gt;If the public-equity transfer occurred as reported, an urgently liquidity-constrained marginal seller is likely less active. That supports a reversal of the portion of the July decline unrelated to fundamentals. The extraordinary rebound suggests that normalization had begun.&lt;/p&gt;
&lt;h3 id="92-what-did-not-improve"&gt;9.2 What did not improve
&lt;/h3&gt;&lt;p&gt;The following questions remain open:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Can AI token demand absorb memory supply growth after 2028?&lt;/li&gt;
&lt;li&gt;Can HBM4 and HBM4E market share and pricing hold?&lt;/li&gt;
&lt;li&gt;Will commodity DRAM and NAND move into oversupply?&lt;/li&gt;
&lt;li&gt;Can hyperscaler cash flow fund CAPEX and long-term purchase commitments?&lt;/li&gt;
&lt;li&gt;How will Chinese memory expansion and U.S. controls split global pricing?&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The rebound is a flow signal, not an earnings validation.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Group&lt;/th&gt;
 &lt;th&gt;Meaning of this event&lt;/th&gt;
 &lt;th&gt;Next evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td&gt;Potential easing of ADR and leverage overhang; allocation unknown&lt;/td&gt;
 &lt;td&gt;Q3 13F/13G, ADR borrow and volume, HBM earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Technical normalization through the same memory factor&lt;/td&gt;
 &lt;td&gt;HBM qualification, commodity DRAM profit, program flows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td&gt;Mixed disclosed options make a simple forced-selling thesis difficult&lt;/td&gt;
 &lt;td&gt;Open interest, long-term contracts, guidance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NBIS, BE, CRWV, SNDK&lt;/td&gt;
 &lt;td&gt;Forced-flow and short-covering effects likely material&lt;/td&gt;
 &lt;td&gt;Volume normalization, block trades, earnings and funding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nvidia&lt;/td&gt;
 &lt;td&gt;Q1 filing showed significant fund put exposure&lt;/td&gt;
 &lt;td&gt;AI demand, dealer positioning, receivables and customer concentration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="10-citadel-is-not-permanent-unlevered-capital"&gt;10. Citadel is not permanent, unlevered capital
&lt;/h2&gt;&lt;p&gt;Describing Citadel as unlevered patient capital would be inaccurate. Citadel is a sophisticated hedge fund and trading ecosystem capable of rapidly pricing and hedging risk.&lt;/p&gt;
&lt;p&gt;The positive interpretation is narrower: positions moved from a seller that reportedly needed immediate liquidity to a much larger risk absorber. Citadel can still:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;hedge with single-stock, index, and options positions&lt;/li&gt;
&lt;li&gt;distribute shares as liquidity returns&lt;/li&gt;
&lt;li&gt;run relative-value trades&lt;/li&gt;
&lt;li&gt;retain some positions and sell others&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The overhang may have shifted from urgent to manageable. It did not necessarily disappear.&lt;/p&gt;
&lt;h2 id="11-what-to-monitor"&gt;11. What to monitor
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Schedules 13D and 13G&lt;/strong&gt; for any investor crossing 5% of SKHY.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Q3 2026 13F filings&lt;/strong&gt; in mid-November for broad SKHY and AI infrastructure changes.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;ADR versus local-share pricing&lt;/strong&gt;, borrow costs, volume, and conversion economics.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Options open interest and implied volatility&lt;/strong&gt; to assess whether dealer and short-covering pressure is fading.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Follow-on block trades&lt;/strong&gt; that could indicate redistribution of the Citadel book.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;2027-2028 earnings estimates&lt;/strong&gt;, the final test of whether price normalization is supported by cash flow.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="12-red-team"&gt;12. Red Team
&lt;/h2&gt;&lt;h3 id="the-fund-was-an-effect-not-a-cause"&gt;The fund was an effect, not a cause
&lt;/h3&gt;&lt;p&gt;AI infrastructure may have been repriced because of valuation, rates, power constraints, and vendor-financing concerns. Situational Awareness could have been a casualty of that move rather than its origin.&lt;/p&gt;
&lt;h3 id="forced-selling-may-not-be-over"&gt;Forced selling may not be over
&lt;/h3&gt;&lt;p&gt;Citadel may hedge or distribute the acquired positions, while other leveraged funds could face margin calls. One transfer does not cleanse an entire crowded factor.&lt;/p&gt;
&lt;h3 id="the-rebound-may-be-mostly-short-covering"&gt;The rebound may be mostly short covering
&lt;/h3&gt;&lt;p&gt;A one-day 18% to 27% rebound can be driven by short covering and options gamma. If volume fades and prices revisit prior lows, the normalization interpretation weakens.&lt;/p&gt;
&lt;h3 id="invalidation"&gt;Invalidation
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Future SEC filings show little or no SKHY position at Situational Awareness.&lt;/li&gt;
&lt;li&gt;The public-equity sale reports are denied by the parties.&lt;/li&gt;
&lt;li&gt;Large follow-on block sales occur after the Citadel transfer.&lt;/li&gt;
&lt;li&gt;AI infrastructure and memory earnings estimates decline structurally.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="13-evidence-classification"&gt;13. Evidence classification
&lt;/h2&gt;&lt;h3 id="fact"&gt;[Fact]
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;SK hynix offered 177.9 million ADSs at $149.&lt;/li&gt;
&lt;li&gt;Three investors collectively indicated non-binding interest in up to $7 billion.&lt;/li&gt;
&lt;li&gt;Q1 filings show BE, CRWV, SNDK, MU, and NVDA exposures.&lt;/li&gt;
&lt;li&gt;A May Schedule 13G reported a 5.6% NBIS stake.&lt;/li&gt;
&lt;li&gt;AI infrastructure and Korean memory stocks rebounded sharply.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="reported"&gt;[Reported]
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Most or all of the public-equity book was transferred to Citadel.&lt;/li&gt;
&lt;li&gt;Fund assets fell to roughly $10 billion.&lt;/li&gt;
&lt;li&gt;The fund used up to four times leverage and faced margin calls.&lt;/li&gt;
&lt;li&gt;The transferred book was approximately $16 billion.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="inference"&gt;[Inference]
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;The transfer likely reduced urgent forced-selling pressure.&lt;/li&gt;
&lt;li&gt;Leveraged AI infrastructure positioning likely amplified the July decline.&lt;/li&gt;
&lt;li&gt;Citadel may have converted an urgent overhang into a more manageable one.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="blocked"&gt;[Blocked]
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Actual SKHY allocation and entry price&lt;/li&gt;
&lt;li&gt;Prime-broker collateral and margin-call records&lt;/li&gt;
&lt;li&gt;Exact transaction value, security quantities, and Citadel hedges&lt;/li&gt;
&lt;li&gt;Daily July net sales and total fund net exposure&lt;/li&gt;
&lt;li&gt;Executed Korean margin liquidations and ELS hedges&lt;/li&gt;
&lt;li&gt;The fund&amp;rsquo;s exact contribution to the KOSPI and AI-factor decline&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;The Situational Awareness episode is an important piece of the July AI infrastructure selloff. The reported transfer of a large public-equity book and the explosive rebound in high-beta names support the view that forced deleveraging distorted prices.&lt;/p&gt;
&lt;p&gt;But three corrections are essential. The $7 billion was not Situational Awareness&amp;rsquo;s disclosed standalone SK hynix allocation. KRW 37.7 trillion was not verified forced-liquidation volume. And the rebound, while powerful evidence of flow normalization, was not a controlled experiment assigning 40% to 50% of the market decline to one fund.&lt;/p&gt;
&lt;p&gt;The most accurate summary is:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Situational Awareness has not been proved to be the sole cause of the July AI rout. It was likely an important marginal seller and amplifier inside a leveraged, crowded market structure. The public-equity transfer can help short-term normalization, but the durability of AI and memory earnings after 2028 still requires separate fundamental proof.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;p&gt;This report is market-structure research based on public information, not a recommendation to buy or sell any security.&lt;/p&gt;</description></item><item><title>Re-Ranking 19 Korean AI Data-Center Stocks: Power, Cooling and Backup Systems Before Operators</title><link>https://koreainvestinsights.com/post/korea-ai-datacenter-beneficiaries-power-cooling-operator-flow-ranking-2026-07-23/</link><pubDate>Thu, 23 Jul 2026 13:35:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/korea-ai-datacenter-beneficiaries-power-cooling-operator-flow-ranking-2026-07-23/</guid><description>&lt;p&gt;At 1:29 p.m. KST on July 23, GnCenergy was up 29.5%, SK Gas 11.4%, LS ELECTRIC 11.2%, LS Materials 10.9% and Seojin System 12.6%. Korea&amp;rsquo;s AI data-center power theme had returned in force.&lt;/p&gt;
&lt;p&gt;That price order is not the same as an investment ranking. GnCenergy has the cleanest backup-power exposure, but it had already risen 77.9% over the prior 20 sessions. LS ELECTRIC has a confirmed KRW 318.96 billion contract, yet trades near 61.6x 2026 earnings while foreign investors sold KRW 198.7 billion over 20 sessions. LG Electronics, by contrast, combines chillers, coolant distribution units and direct-to-chip cooling. Its shares fell 10.7% over the same window while foreign and real-money investors bought KRW 85.5 billion and KRW 105.3 billion.&lt;/p&gt;
&lt;p&gt;That distinction drives this report: &lt;strong&gt;business directness and a good entry price must be scored separately.&lt;/strong&gt;&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Related context
This is the market and flow update to the July 4 &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-power-bottleneck-korea-value-chain-screen-2026-07-04/" &gt;AI data-center power bottleneck map&lt;/a&gt;. Company-specific context is available in our work on &lt;a class="link" href="https://koreainvestinsights.com/post/gnc-energy-ai-datacenter-backup-generator-sustainable-rally-2026-07-04/" &gt;GnCenergy&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/vinatech-bloom-ai-datacenter-supercapacitor-power-buffer-2026-07-04/" &gt;VinaTech and Bloom Energy&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-gas-lng-cold-energy-ai-datacenter-cooling-option-2026-07-04/" &gt;SK Gas LNG cold energy&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/sk-gas-sk-oceanplant-ai-datacenter-power-second-line-2026-07-04/" &gt;SK Gas versus SK Oceanplant&lt;/a&gt;. See also the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/korean-ai-companies-hub/" &gt;Korean AI Companies Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;Korea&amp;rsquo;s data-center theme mixes operators, facility and power suppliers, and AI server component vendors. The first revenue capture often sits with &lt;strong&gt;power, cooling and backup-power suppliers&lt;/strong&gt;. Operators initially carry capex and only generate attractive returns after securing utilization, power economics and tenant contracts.&lt;/p&gt;
&lt;p&gt;We score 19 names across three dimensions.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Dimension&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Weight&lt;/th&gt;
 &lt;th&gt;What it measures&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fundamentals and direct exposure&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td&gt;Confirmed contracts, product and customer directness, repeat-order potential&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign and real-money flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td&gt;20-session foreign flow plus insurance, funds, private funds and pensions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Valuation and entry position&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td&gt;Forward P/E, recent return, intraday spike and price response to flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The current risk-adjusted order is &lt;strong&gt;LG Electronics, Iljin Electric, Samsung SDS, GST and SK Gas&lt;/strong&gt;.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Bucket&lt;/th&gt;
 &lt;th&gt;Names&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Begin watching or scaling&lt;/td&gt;
 &lt;td&gt;LG Electronics, GST&lt;/td&gt;
 &lt;td&gt;Lower price burden with accumulated flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flow-confirmation candidate&lt;/td&gt;
 &lt;td&gt;Iljin Electric&lt;/td&gt;
 &lt;td&gt;Direct orders and reasonable valuation, but foreign buying must persist&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Wait for a pullback&lt;/td&gt;
 &lt;td&gt;Samsung SDS, SK Gas&lt;/td&gt;
 &lt;td&gt;Strong theses, less attractive after the July 23 move&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High quality, early entry&lt;/td&gt;
 &lt;td&gt;SK Telecom, Sanil Electric, HD Hyundai Electric, Seojin System&lt;/td&gt;
 &lt;td&gt;Flow conflict, high expectations or balance-sheet questions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Do not chase&lt;/td&gt;
 &lt;td&gt;GnCenergy, LS ELECTRIC, Hyosung Heavy, VinaTech, LS Materials, Gaon Cable&lt;/td&gt;
 &lt;td&gt;Direct exposure is offset by price, valuation or weakening flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="1-method"&gt;1. Method
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Basis&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Live price&lt;/td&gt;
 &lt;td&gt;Naver Finance at 13:29 KST on July 23, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20-session return&lt;/td&gt;
 &lt;td&gt;June 24 close to July 22 close, adjusted Naver FChart prices&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investor flow&lt;/td&gt;
 &lt;td&gt;June 24 through July 22, Kiwoom REST &lt;code&gt;ka10059&lt;/code&gt;, amount basis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Real money&lt;/td&gt;
 &lt;td&gt;Insurance, investment trusts, private funds and pensions; financial-investment desks excluded&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Valuation&lt;/td&gt;
 &lt;td&gt;July 22 Naver/WiseReport consensus and July 23 intraday price&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;We recalculated returns from Naver&amp;rsquo;s adjusted-price series because the local &lt;code&gt;prices_daily&lt;/code&gt; table contained corporate-action adjustment errors in a few names. July 23 prices are intraday, not closing prices. Forward multiples for thin-consensus and loss-making companies are only rough reference points.&lt;/p&gt;
&lt;h2 id="2-three-different-data-center-trades"&gt;2. Three Different Data-Center Trades
&lt;/h2&gt;&lt;h3 id="operators"&gt;Operators
&lt;/h3&gt;&lt;p&gt;Samsung SDS, NHN, SK Telecom, LG Uplus, KT and NAVER own or rent data-center and GPU capacity to sell cloud, GPUaaS and enterprise AI services. Their earnings path runs from site and power procurement to construction, anchor customers, utilization and finally operating profit.&lt;/p&gt;
&lt;p&gt;SK Telecom&amp;rsquo;s up-to-15GW plan is a useful scale marker, but remains a plan that depends on sites, power, tenants and financing.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt; NHN Cloud is one step further along, having begun service on a 7,656-B200-GPU cluster in Yangpyeong.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h3 id="facility-and-power-suppliers"&gt;Facility and power suppliers
&lt;/h3&gt;&lt;p&gt;Switchgear, transformers, cables, backup generators, ESS and cooling systems can recognize sales from the order and construction phase, before the operator reaches mature utilization. This is the most direct early-cycle listed-equity layer. The risks are one-off contracts, customer concentration, input costs and working capital.&lt;/p&gt;
&lt;h3 id="ai-server-components"&gt;AI server components
&lt;/h3&gt;&lt;p&gt;Samsung Electronics and SK Hynix memory, FADU enterprise SSD controllers, Isu Petasys networking PCBs and Samsung Electro-Mechanics or Daeduck substrates respond more to server configuration, accelerator units and memory content than to the number of buildings.&lt;/p&gt;
&lt;p&gt;The key is timing. Facility suppliers capture the front end of the investment cycle. Operators capture recurring revenue and platform value later.&lt;/p&gt;
&lt;h2 id="3-direct-beneficiaries"&gt;3. Direct Beneficiaries
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Area&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Ticker&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Directness&lt;/th&gt;
 &lt;th&gt;Evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Distribution systems&lt;/td&gt;
 &lt;td&gt;LS ELECTRIC&lt;/td&gt;
 &lt;td style="text-align: right"&gt;010120&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A+&lt;/td&gt;
 &lt;td&gt;KRW 318.96bn of Bloom-linked North American data-center switchgear and distribution transformers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Backup generation&lt;/td&gt;
 &lt;td&gt;GnCenergy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;119850&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A+&lt;/td&gt;
 &lt;td&gt;Naver, KT and LG Uplus references; Paju AIDC cooperation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Special transformers&lt;/td&gt;
 &lt;td&gt;Sanil Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;062040&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A&lt;/td&gt;
 &lt;td&gt;KRW 50.3bn Bloom-related U.S. data-center transformer contract&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power buffer&lt;/td&gt;
 &lt;td&gt;VinaTech&lt;/td&gt;
 &lt;td style="text-align: right"&gt;126340&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A&lt;/td&gt;
 &lt;td&gt;KRW 41.2bn Bloom-related supercapacitor system contract&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High-voltage cable&lt;/td&gt;
 &lt;td&gt;Iljin Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;103590&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A-&lt;/td&gt;
 &lt;td&gt;Repeat 275kV cable orders for Malaysian data centers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MV cable and bus duct&lt;/td&gt;
 &lt;td&gt;Gaon Cable&lt;/td&gt;
 &lt;td style="text-align: right"&gt;000500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A-&lt;/td&gt;
 &lt;td&gt;U.S. AI data-center medium-voltage cable and local capacity expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Grid package&lt;/td&gt;
 &lt;td&gt;HD Hyundai Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;267260&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A-&lt;/td&gt;
 &lt;td&gt;North American data-center and grid transformers and distribution packages&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Transmission&lt;/td&gt;
 &lt;td&gt;Hyosung Heavy Industries&lt;/td&gt;
 &lt;td style="text-align: right"&gt;298040&lt;/td&gt;
 &lt;td style="text-align: right"&gt;B+&lt;/td&gt;
 &lt;td&gt;High-voltage transformers and reactors for grid expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cooling&lt;/td&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;066570&lt;/td&gt;
 &lt;td style="text-align: right"&gt;A-&lt;/td&gt;
 &lt;td&gt;Chillers, CDUs and direct liquid cooling as an integrated package&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;LS ELECTRIC&amp;rsquo;s contract equals 6.42% of 2025 revenue and runs from April 28, 2026 to March 29, 2027.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt; It is among the clearest disclosed contracts in the theme.&lt;/p&gt;
&lt;p&gt;LG Electronics is different. Cooling is not yet large enough to drive group earnings, but it is one of the few large listed Korean vendors able to offer chillers, a 1.4MW CDU, direct-to-chip cooling and air systems together.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt; In the LG group&amp;rsquo;s Paju AIDC design, LG Electronics provides cooling, LG Energy Solution provides power equipment and LG Uplus operates the site.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt; The company said chiller orders tripled in 2025 and expected full-scale order and revenue recognition from 2026.&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h2 id="4-ess-power-quality-and-fuel"&gt;4. ESS, Power Quality and Fuel
&lt;/h2&gt;&lt;p&gt;The second line includes Seojin System&amp;rsquo;s enclosures and racks, Samsung SDI and LG Energy Solution battery systems, LS Materials ultracapacitors, Samwha Capacitor power components and SK Eternix&amp;rsquo;s distributed-power and PPA optionality.&lt;/p&gt;
&lt;p&gt;VinaTech&amp;rsquo;s move from cells to system-level delivery can raise content value, but customer concentration and production margins remain decisive. Seojin System shows a sharp flow conflict: foreigners bought KRW 128.9 billion over 20 sessions, while real-money investors sold KRW 30.5 billion and the stock fell 34.7%. Its roughly 11.5x 2027 P/E should be treated as a turnaround estimate, not as proven cheapness.&lt;/p&gt;
&lt;p&gt;SK Gas offers an unusually broad option set across the 1.2GW Ulsan GPS plant, KET LNG terminal, LNG cold energy and SK&amp;rsquo;s AIDC buildout. The missing bridge is economic ownership: the power contract, cold-energy tariff, PPA economics and profit attributable to SK Gas are not public.&lt;/p&gt;
&lt;h2 id="5-operators"&gt;5. Operators
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Ticker&lt;/th&gt;
 &lt;th&gt;Core asset&lt;/th&gt;
 &lt;th&gt;What must be proven&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung SDS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;018260&lt;/td&gt;
 &lt;td&gt;Samsung Cloud and 60MW Gumi AI data center&lt;/td&gt;
 &lt;td&gt;Customer ramp and cloud-margin conversion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NHN&lt;/td&gt;
 &lt;td style="text-align: right"&gt;181710&lt;/td&gt;
 &lt;td&gt;Gwangju/Yangpyeong AI data centers and GPUaaS&lt;/td&gt;
 &lt;td&gt;Utilization, pricing and post-depreciation returns&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Telecom&lt;/td&gt;
 &lt;td style="text-align: right"&gt;017670&lt;/td&gt;
 &lt;td&gt;Ulsan and DSX large-scale AI Cloud&lt;/td&gt;
 &lt;td&gt;Financing, tenants, utilization and ROIC&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Uplus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;032640&lt;/td&gt;
 &lt;td&gt;Pyeongchon and Paju AIDC&lt;/td&gt;
 &lt;td&gt;Occupancy and long-term leases&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;030200&lt;/td&gt;
 &lt;td&gt;kt cloud and IDC&lt;/td&gt;
 &lt;td&gt;Indirect listed exposure because kt cloud is private&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAVER&lt;/td&gt;
 &lt;td style="text-align: right"&gt;035420&lt;/td&gt;
 &lt;td&gt;GAK Chuncheon/Sejong and own AI Cloud&lt;/td&gt;
 &lt;td&gt;AI-service monetization, not only infrastructure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung SDS has a 60MW Gumi AI data-center plan and said an anchor customer had already been secured.&lt;sup id="fnref:7"&gt;&lt;a href="#fn:7" class="footnote-ref" role="doc-noteref"&gt;7&lt;/a&gt;&lt;/sup&gt; Its balance sheet and captive demand make it the most comfortable operator exposure, but a 22x-plus 2026 P/E needs cloud and generative-AI revenue to lift margins.&lt;/p&gt;
&lt;p&gt;SK Telecom has excellent recent flow: foreign investors bought KRW 71.5 billion and real money KRW 80.1 billion. At around 17x 2026 earnings, however, the market is already assigning some AI-infrastructure value. The next rerating needs contracted customers and an explicit return framework, not only a 15GW ambition.&lt;/p&gt;
&lt;h2 id="6-the-current-top-ten"&gt;6. The Current Top Ten
&lt;/h2&gt;&lt;p&gt;This is an entry ranking, not a permanent business-quality ranking.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;20D return&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign / real money&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current valuation&lt;/th&gt;
 &lt;th style="text-align: right"&gt;July 23 intraday&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-10.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 85.5bn / +105.3bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 15.0x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 184,300, +0.9%&lt;/td&gt;
 &lt;td&gt;Accumulation candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;Iljin Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-17.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-24.3bn / +10.7bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 19.3x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 64,600, +6.1%&lt;/td&gt;
 &lt;td&gt;Flow-turn candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Samsung SDS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.9bn / +57.4bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 22.4x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 207,000, +5.6%&lt;/td&gt;
 &lt;td&gt;Quality core&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;GST&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-16.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.7bn / +4.7bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;TTM P/E about 14.4x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 43,050, +1.7%&lt;/td&gt;
 &lt;td&gt;Deep-alpha candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.4bn / +7.4bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 4.6x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 254,000, +11.4%&lt;/td&gt;
 &lt;td&gt;Wait for pullback&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;SK Telecom&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+71.5bn / +80.1bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E about 17.0x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 99,000, +5.2%&lt;/td&gt;
 &lt;td&gt;Great flow; verify ROIC&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;LG Uplus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.2bn / +31.7bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 8.7x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 14,730, +0.6%&lt;/td&gt;
 &lt;td&gt;Defensive AIDC&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;8&lt;/td&gt;
 &lt;td&gt;Sanil Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-25.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28.9bn / -33.3bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 26.1x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 177,200, +6.3%&lt;/td&gt;
 &lt;td&gt;Wait for institution turn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;9&lt;/td&gt;
 &lt;td&gt;HD Hyundai Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-15.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-148.1bn / +14.5bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026E P/E 31.8x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 857,000, +7.8%&lt;/td&gt;
 &lt;td&gt;Great company, early entry&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td&gt;Seojin System&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-34.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+128.9bn / -30.5bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2027E P/E about 11.5x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 40,800, +12.6%&lt;/td&gt;
 &lt;td&gt;High-risk turnaround&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="7-why-lg-electronics-ranks-first"&gt;7. Why LG Electronics Ranks First
&lt;/h2&gt;&lt;p&gt;LG Electronics has the best balance of product breadth, flow, valuation and unpriced optionality.&lt;/p&gt;
&lt;p&gt;It can combine chillers, CDUs, direct liquid cooling and air systems. Foreign and real-money investors bought together while the share price fell 10.7% over 20 sessions. At roughly 15x 2026 earnings, it is cheaper than the large electrical-equipment leaders. The July 23 move was also limited to about 1%, reducing chase risk.&lt;/p&gt;
&lt;p&gt;The cooling business still needs proof at group level. If global orders translate into disclosed revenue and margin, the multiple can expand. If cooling orders lag while core-business profitability weakens, the option should not be assigned a large value.&lt;/p&gt;
&lt;h2 id="8-iljin-electric-samsung-sds-gst-and-sk-gas"&gt;8. Iljin Electric, Samsung SDS, GST and SK Gas
&lt;/h2&gt;&lt;h3 id="iljin-electric"&gt;Iljin Electric
&lt;/h3&gt;&lt;p&gt;Iljin combines direct data-center cable orders with a lower multiple than the large electrical-equipment leaders. Real money bought KRW 10.7 billion over 20 sessions, and foreign flow turned positive by about KRW 5.8 billion over the latest five. A further three to five sessions of foreign buying would turn the case from oversold to flow reversal.&lt;/p&gt;
&lt;h3 id="samsung-sds"&gt;Samsung SDS
&lt;/h3&gt;&lt;p&gt;Real money bought KRW 57.4 billion over 20 sessions and was a net buyer on 16 of those days. The share price rose only 3.3%. The 60MW Gumi project, Samsung Cloud and enterprise generative AI make it the safest operator exposure, but an entry near KRW 200,000 would be more attractive after the July 23 rise.&lt;/p&gt;
&lt;h3 id="gst"&gt;GST
&lt;/h3&gt;&lt;p&gt;GST is not yet a proven data-center cooling supplier. The case is a profitable semiconductor scrubber business with low leverage plus a low-cost immersion-cooling option. Foreign and real-money flow were both positive while the stock fell 16%. A commercial cooling order, not another pilot, is the catalyst.&lt;/p&gt;
&lt;h3 id="sk-gas"&gt;SK Gas
&lt;/h3&gt;&lt;p&gt;SK Gas is the cheapest name in the group at about 4.6x 2026 earnings, but the July 23 11.4% rise reduced short-term margin of safety. A revisit to KRW 230,000-240,000 or a disclosed power/cold-energy contract would offer a cleaner entry.&lt;/p&gt;
&lt;h2 id="9-high-quality-names-to-wait-on"&gt;9. High-Quality Names to Wait On
&lt;/h2&gt;&lt;p&gt;SK Telecom has the best flow but needs financing, tenant and ROIC evidence. LG Uplus is a lower-multiple defensive AIDC exposure, though the profit contribution remains small. Sanil Electric has a strong Bloom contract and high margins, but real money sold KRW 33.3 billion. HD Hyundai Electric&amp;rsquo;s order quality is excellent, but 31.8x 2026 earnings and heavy foreign selling leave limited room for error. Seojin System needs institutional flow, working-capital and cash-flow confirmation.&lt;/p&gt;
&lt;h2 id="10-do-not-chase-direct-exposure"&gt;10. Do Not Chase Direct Exposure
&lt;/h2&gt;&lt;p&gt;GnCenergy is the cleanest backup-power beneficiary, but a 77.9% 20-session gain and a 29.5% intraday rise leave poor entry asymmetry. LS ELECTRIC&amp;rsquo;s contract is real, but so are a 61.6x 2026 P/E, KRW 198.7 billion of foreign selling and an 11.2% intraday rise. Hyosung Heavy combines strong earnings with a roughly 34x multiple and selling by both foreign and real-money accounts. VinaTech has a real Bloom contract but a trailing loss and real-money selling. LS Materials and Gaon Cable lack a comfortable combination of confirmed incremental earnings, valuation and flow.&lt;/p&gt;
&lt;h2 id="11-action-framework"&gt;11. Action Framework
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Approach&lt;/th&gt;
 &lt;th&gt;Confirmation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td&gt;Scale gradually&lt;/td&gt;
 &lt;td&gt;Support in the mid-KRW 180,000s, continued dual flow, cooling revenue conversion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GST&lt;/td&gt;
 &lt;td&gt;Small pilot position&lt;/td&gt;
 &lt;td&gt;KRW 41,000-43,000 support, stable core earnings, commercial cooling order&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Iljin Electric&lt;/td&gt;
 &lt;td&gt;Flow-confirmation entry&lt;/td&gt;
 &lt;td&gt;Three to five more sessions of foreign buying and acceptance near KRW 64,000&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung SDS&lt;/td&gt;
 &lt;td&gt;Wait for pullback&lt;/td&gt;
 &lt;td&gt;Re-test near KRW 200,000 with continued real-money buying&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td&gt;Wait for price or contract&lt;/td&gt;
 &lt;td&gt;KRW 230,000-240,000 or disclosed AIDC power/cold-energy economics&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The upside catalysts are additional LG Electronics cooling orders, repeat Iljin data-center cable contracts, Gumi utilization and cloud-margin evidence at Samsung SDS, a commercial immersion-cooling order at GST, and disclosed PPA or cold-energy economics at SK Gas.&lt;/p&gt;
&lt;p&gt;The thesis fails if data-center starts and equipment orders are delayed, large contracts prove one-off, input and financing costs erase backlog margins, operators cannot achieve utilization, or cooling and ESS products remain pilots rather than commercial revenue.&lt;/p&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;The Korean AI data-center theme needs three separate lenses: operators, power and cooling suppliers, and server components. Suppliers capture the front of the investment cycle through orders and backlog. Operators earn the long-duration value only after filling their centers and recovering power and depreciation.&lt;/p&gt;
&lt;p&gt;LS ELECTRIC and GnCenergy lead on directness. Once price and flow are included, the ranking changes.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The current order is LG Electronics, Iljin Electric, Samsung SDS, GST and SK Gas.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;The next returns are likely to depend less on whether a company can tell an AI data-center story and more on three questions: Is there a contract? Is durable capital buying? How much is already in the price?&lt;/p&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;p&gt;Price and flow data come from the Korea Invest Insights local database, Naver FChart and Kiwoom REST &lt;code&gt;ka10059&lt;/code&gt;. This report is for research and information only and is not investment advice.&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://news.sktelecom.com/227469" target="_blank" rel="noopener"
 &gt;SK Telecom newsroom, up-to-15GW AIDC plan&lt;/a&gt;&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://inside.nhn.com/news/939" target="_blank" rel="noopener"
 &gt;NHN, Yangpyeong 7,656-GPU infrastructure&lt;/a&gt;&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://kind.krx.co.kr/external/2026/04/29/000110/20260428001034/91370.htm" target="_blank" rel="noopener"
 &gt;KRX KIND, LS ELECTRIC supply contract&lt;/a&gt;&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.lg.com/global/newsroom/news/eco-solution/lg-electronics-showcases-ai-data-center-cooling-solutions-at-data-center-world-2026/" target="_blank" rel="noopener"
 &gt;LG Electronics at Data Center World 2026&lt;/a&gt;&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.lg.co.kr/media/release/29882" target="_blank" rel="noopener"
 &gt;LG group, Paju AI data-center infrastructure&lt;/a&gt;&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.lg.com/content/dam/channel/wcms/global/ir/05-ir-events/2026/2026_Q1_Earning_Release_of_LGE_Transcript.pdf" target="_blank" rel="noopener"
 &gt;LG Electronics Q1 2026 earnings transcript&lt;/a&gt;&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:7"&gt;
&lt;p&gt;&lt;a class="link" href="https://image.samsungsds.com/kr/investor/ir_events/ear_rele/__icsFiles/afieldfile/2026/01/23/Samsung%20SDS%202025Q4_Script_Kor_%EC%B5%9C%EC%A2%85_1.pdf" target="_blank" rel="noopener"
 &gt;Samsung SDS Q4 2025 earnings script&lt;/a&gt;&amp;#160;&lt;a href="#fnref:7" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>FADU's Q2 Surprise and Its Real Moat: Controller-Firmware Integration Matters More Than Gen6 Speed</title><link>https://koreainvestinsights.com/post/fadu-2q26-earnings-surprise-gen6-essd-controller-narrow-moat-2026-07-22/</link><pubDate>Wed, 22 Jul 2026 17:32:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/fadu-2q26-earnings-surprise-gen6-essd-controller-narrow-moat-2026-07-22/</guid><description>&lt;p&gt;FADU reported preliminary Q2 2026 revenue of KRW 73.155bn and operating profit of KRW 16.267bn. Revenue beat local consensus by 16.1%, while operating profit beat by 89.2%. The operating margin reached 22.2%, above the 20.5% upper bound of this site&amp;rsquo;s earlier bull case.&lt;/p&gt;
&lt;p&gt;The quarter removed much of the uncertainty around whether FADU can generate profit from its Gen5 controller business. The stock market immediately moved to a harder question. FADU closed down 9.3% at KRW 70,400 on July 22, while foreigners sold a net KRW 70.3bn over the latest ten sessions. Investors now want evidence for 2027 growth: Gen6 mass production, commercial QLC adoption and customer diversification.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Related research
This report reviews the outcome of the &lt;a class="link" href="https://koreainvestinsights.com/post/fadu-2q26-earnings-preview-essd-controller-moderate-beat-2026-07-04/" &gt;FADU Q2 2026 earnings preview&lt;/a&gt; and follows the longer-duration framework in &lt;a class="link" href="https://koreainvestinsights.com/post/fadu-ai-infra-storage-bottleneck-p-q-new-segment-2026-06-02/" &gt;What FADU Must Prove on Price, Volume and New Markets&lt;/a&gt;. The first flow-based comparison with Sandisk is available in &lt;a class="link" href="https://koreainvestinsights.com/post/fadu-sandisk-ai-storage-korea-beta-jeju-semiconductor-2026-05-31/" &gt;Is FADU Korea&amp;rsquo;s Sandisk Beta?&lt;/a&gt;. See also the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-equipment-ip-hub/" &gt;Korea Semiconductor Value Chain Hub&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;FADU&amp;rsquo;s core advantage is not &lt;code&gt;PCIe Gen6 speed&lt;/code&gt; by itself. It is the integration of controller ASIC and firmware that makes NAND behave with low power, consistent latency and high endurance under a customer&amp;rsquo;s workload.&lt;/li&gt;
&lt;li&gt;The current moat is &lt;strong&gt;narrow but expandable&lt;/strong&gt;. Gen5 commercialization is visible in revenue and OCP participation. Gen6 customer qualification and production, 64TB/128TB QLC design wins and customer diversification remain unproven.&lt;/li&gt;
&lt;li&gt;Q2 revenue of KRW 73.16bn beat consensus of KRW 63.0bn by 16.1%. Operating profit of KRW 16.27bn beat KRW 8.6bn by 89.2%. The 22.2% margin is the more important signal.&lt;/li&gt;
&lt;li&gt;The preliminary filing does not disclose net income, product mix, gross margin, cash flow or receivables. Revised 2026 EPS of roughly KRW 1,930 and a 36.5x P/E are analytical estimates, not reported or consensus figures.&lt;/li&gt;
&lt;li&gt;KRW 70,400 is close to the low end of the 2027 base-case range of KRW 72,000-86,000. The stock is not cheap on 2026 earnings alone, but modestly discounted if 2027 growth is achieved.&lt;/li&gt;
&lt;li&gt;Fundamentals improved while flows remain weak. Foreigners sold KRW 70.3bn over ten sessions, and program selling reached KRW 17.9bn during the three sessions after the earnings release.&lt;/li&gt;
&lt;/ul&gt;
&lt;hr&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 FADU has proved that it can commercialize a high-performance Gen5 enterprise SSD controller. It has an option to become a broader platform in Gen6 and QLC, but that option requires customer qualification, production orders, lower customer concentration and independent performance evidence.
 &lt;/div&gt;
&lt;/div&gt;
&lt;hr&gt;
&lt;h2 id="1-q2-the-margin-beat-matters-more-than-the-revenue-beat"&gt;1. Q2: The Margin Beat Matters More Than the Revenue Beat
&lt;/h2&gt;&lt;p&gt;FADU filed preliminary consolidated results on July 20. These numbers have not completed external auditor review and can change in the half-year report.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q2 2026 preliminary&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Consensus&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Difference&lt;/th&gt;
 &lt;th style="text-align: right"&gt;QoQ&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 73.16bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 63.0bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+22.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 16.27bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 8.6bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+89.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+111.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;22.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;13.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+8.6pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.3pp&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First-half revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 132.69bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+209.3% YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First-half operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 23.95bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Turned profitable YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://dart.fss.or.kr/dsaf001/main.do?rcpNo=20260720900116" target="_blank" rel="noopener"
 &gt;DART preliminary earnings filing&lt;/a&gt;; local consensus snapshot based on Naver WiseReport as of July 22, 2026.&lt;/p&gt;
&lt;p&gt;The comparison with the July 4 preview is revealing.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Reference case&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Operating profit&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Operating margin&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Prior midpoint&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 70.5bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 11.8bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Prior bull-case upper bound&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 75.0bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 15.4bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Preliminary result&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 73.16bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 16.27bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;22.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Difference vs midpoint&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+37.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.5pp&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Revenue fell just below the prior KRW 75bn bull threshold, but operating profit exceeded the bull-case upper bound. The surprise is therefore not simply more units. It likely reflects a favorable controller mix, fixed-cost absorption and contract economics. Those drivers must still be confirmed in the half-year filing.&lt;/p&gt;
&lt;p&gt;Management said disclosed new orders exceeded KRW 340bn during the first half, a new server OEM customer was secured, and second-half revenue and profit should exceed the first half. That is useful guidance, but not audited evidence. Controller mix, gross margin, operating cash flow and receivables are the next checks.&lt;/p&gt;
&lt;h2 id="2-what-an-enterprise-ssd-controller-actually-does"&gt;2. What an Enterprise SSD Controller Actually Does
&lt;/h2&gt;&lt;p&gt;NAND is the storage medium. The controller combines traffic management, error correction and power management. It translates host commands into physical NAND operations, distributes workloads across channels, and manages performance, endurance and data integrity.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;CPU and GPU
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ↓ PCIe/NVMe commands
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;FADU controller ASIC + firmware
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ maps logical addresses to physical NAND locations
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ schedules reads and writes across NAND channels
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ manages ECC, bad blocks and endurance
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ controls garbage collection and WAF
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ manages power, thermals and power-loss protection
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; └─ provides QoS, security and telemetry
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ↓
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;NAND packages
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Peak speed is not the only metric. Cloud customers care about performance consistency under congestion, throughput within a power envelope, and data integrity during power failure. Very slow outlier requests can damage service quality even when average latency looks good. This makes p99.99 tail latency an important benchmark.&lt;/p&gt;
&lt;h2 id="3-why-the-controller-matters-more-after-pcie-gen6"&gt;3. Why the Controller Matters More After PCIe Gen6
&lt;/h2&gt;&lt;p&gt;PCIe 6.0 introduced 64GT/s signaling, PAM4, FEC and FLIT mode, doubling Gen5 bandwidth. The practical ceiling for an x4 SSD is around 28GB/s. As the interface gets faster, the bottleneck shifts toward NAND parallelism, power, thermal control and firmware scheduling. &lt;a class="link" href="https://pcisig.com/pci-express-6.0-specification" target="_blank" rel="noopener"
 &gt;PCI-SIG PCIe 6.0&lt;/a&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Technology shift&lt;/th&gt;
 &lt;th&gt;Server requirement&lt;/th&gt;
 &lt;th&gt;Controller challenge&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;PCIe Gen5 to Gen6&lt;/td&gt;
 &lt;td&gt;2x bandwidth, less GPU idle time&lt;/td&gt;
 &lt;td&gt;PAM4/FEC handling and NAND scheduling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI inference and RAG&lt;/td&gt;
 &lt;td&gt;High random reads, low tail latency&lt;/td&gt;
 &lt;td&gt;QoS, queue management and caching&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High-capacity QLC&lt;/td&gt;
 &lt;td&gt;122TB and above, lower cost per TB&lt;/td&gt;
 &lt;td&gt;ECC, write amplification and endurance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FDP and ZNS&lt;/td&gt;
 &lt;td&gt;Host-directed data placement&lt;/td&gt;
 &lt;td&gt;Less garbage collection, consistent performance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;E1.S and E3.S&lt;/td&gt;
 &lt;td&gt;Higher rack density and cooling efficiency&lt;/td&gt;
 &lt;td&gt;Power and thermal throttling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Multi-tenant cloud&lt;/td&gt;
 &lt;td&gt;Per-customer isolation&lt;/td&gt;
 &lt;td&gt;Virtualization, multiple PFs, consistent latency&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stronger security&lt;/td&gt;
 &lt;td&gt;Firmware and data protection&lt;/td&gt;
 &lt;td&gt;Root of Trust, encryption and attestation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;AI enterprise SSD demand is splitting into two segments. High-performance TLC serves checkpoints, RAG, vector databases and AI data pipelines. High-capacity QLC serves data lakes, object storage and read-heavy inference data.&lt;/p&gt;
&lt;p&gt;QLC improves density and cost but has weaker write performance, endurance and error characteristics. This increases the value of ECC, garbage collection, wear leveling and FDP. Micron is already shipping a 245TB QLC SSD after launching 122TB-class products. &lt;a class="link" href="https://investors.micron.com/news-releases/news-release-details/industry-leading-245tb-micron-6600-ion-data-center-ssd-now" target="_blank" rel="noopener"
 &gt;Micron 245TB 6600 ION&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="4-what-fadu-has-already-proved"&gt;4. What FADU Has Already Proved
&lt;/h2&gt;&lt;h3 id="41-architecture-and-firmware"&gt;4.1 Architecture and Firmware
&lt;/h3&gt;&lt;p&gt;FADU describes an architecture built around a high-speed point-to-point network, hardware offload, light cores and simpler firmware instead of a conventional many-core shared interconnect. The objective is to reduce software overhead and memory movement.&lt;/p&gt;
&lt;p&gt;The company&amp;rsquo;s Gen5 FC5161 specifications show sequential reads of 13.7GB/s and random reads of 3.3M IOPS. FADU also presents favorable write throughput and performance-per-watt comparisons, but these are company tests rather than independent, same-condition benchmarks. &lt;a class="link" href="https://kind.krx.co.kr/external/dst/irReference/18509/1Q26%20IR%20Book_KOR_260512.pdf" target="_blank" rel="noopener"
 &gt;FADU Q1 2026 IR&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="42-fdp-and-the-ocp-ecosystem"&gt;4.2 FDP and the OCP Ecosystem
&lt;/h3&gt;&lt;p&gt;The FC5161-based Gen5 SSD is listed in the OCP Marketplace. OCP describes it as the first OCP SSD to support FDP, with 64 Physical Functions, power-loss protection, security, telemetry and thermal management. &lt;a class="link" href="https://www.opencompute.org/products/648/fadu-xdw223-pcie-gen5-nvme-u2-ssd" target="_blank" rel="noopener"
 &gt;OCP FADU XDW223&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;FDP places data more efficiently inside the SSD, reducing internal movement, garbage collection and write amplification. That can improve performance consistency, endurance and power consumption at the same time.&lt;/p&gt;
&lt;h3 id="43-commercial-gen5-revenue"&gt;4.3 Commercial Gen5 Revenue
&lt;/h3&gt;&lt;p&gt;Controller revenue reached approximately KRW 47.9bn in Q1 2026, or 80.4% of total revenue. This is the clearest evidence that FADU has moved from development into commercial supply. The Q2 preliminary filing does not provide product-level revenue, so the controller share must be checked in the half-year report.&lt;/p&gt;
&lt;h3 id="44-gen6-progress"&gt;4.4 Gen6 Progress
&lt;/h3&gt;&lt;p&gt;The FC6161 has completed tape-out. FADU targets 28.5GB/s sequential reads and writes and 6.9M random-read IOPS. The OCP contribution database also lists 2026 security-evaluation entries for the FC6161. &lt;a class="link" href="https://www.opencompute.org/contributions?contributions%5Bpage%5D=17" target="_blank" rel="noopener"
 &gt;OCP Contribution Database&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;This demonstrates design progress and participation in the security-evaluation ecosystem. It does not prove customer qualification, production orders or revenue.&lt;/p&gt;
&lt;h2 id="5-competitors-are-already-moving"&gt;5. Competitors Are Already Moving
&lt;/h2&gt;&lt;p&gt;FADU&amp;rsquo;s performance target is competitive, but Gen6 leadership has not been established.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company and product&lt;/th&gt;
 &lt;th&gt;Public status&lt;/th&gt;
 &lt;th&gt;Implication&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron 9650&lt;/td&gt;
 &lt;td&gt;28GB/s, 5.5M random-read IOPS, high-volume production in 2026&lt;/td&gt;
 &lt;td&gt;Integrated NAND-controller supplier is already shipping Gen6&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung PM1763&lt;/td&gt;
 &lt;td&gt;Mass production announced in July 2026, ninth-generation V-NAND and 4nm controller&lt;/td&gt;
 &lt;td&gt;Gen6 optimized for AI and liquid-cooled infrastructure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Silicon Motion SM8466&lt;/td&gt;
 &lt;td&gt;Gen5, TLC/QLC, FDP/ZNS, 128TB QLC reference design&lt;/td&gt;
 &lt;td&gt;Direct merchant-controller competition&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FADU FC6161&lt;/td&gt;
 &lt;td&gt;Tape-out, targets 28.5GB/s and 6.9M IOPS&lt;/td&gt;
 &lt;td&gt;Strong target, but customer qualification and revenue remain unconfirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://investors.micron.com/news-releases/news-release-details/micron-unveils-portfolio-industry-first-ssds-power-ai-revolution" target="_blank" rel="noopener"
 &gt;Micron AI SSD portfolio&lt;/a&gt;, &lt;a class="link" href="https://news.samsung.com/kr/%EC%82%BC%EC%84%B1%EC%A0%84%EC%9E%90-ai-%EC%9D%B8%ED%94%84%EB%9D%BC%EC%97%90-%EC%B5%9C%EC%A0%81%ED%99%94%EB%90%9C-%EA%B8%B0%EC%97%85%EC%9A%A9-ssd-pm1763-%EC%96%91%EC%82%B0" target="_blank" rel="noopener"
 &gt;Samsung PM1763&lt;/a&gt;, &lt;a class="link" href="https://www.siliconmotion.com/products/enterprise/detail" target="_blank" rel="noopener"
 &gt;Silicon Motion enterprise controller&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;FADU states that its controllers can support both TLC and QLC. The limitation is therefore not the absence of QLC support. The missing evidence is a public 64TB or 128TB QLC customer design win, qualification and production revenue. Silicon Motion already presents a 128TB QLC reference platform.&lt;/p&gt;
&lt;p&gt;Qualification is also a two-sided barrier. Enterprise SSD development and qualification can take more than 18 months, followed by three to five years of supply. That creates switching costs after entry. Meta&amp;rsquo;s open-source SSD qualification framework can reduce testing friction for FADU, but over time it may also reduce process advantages held by incumbents. &lt;a class="link" href="https://www.opencompute.org/blog/introducing-metas-open-source-testing-framework-revolutionizing-ssd-qualifications" target="_blank" rel="noopener"
 &gt;Meta SSD qualification framework&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="6-customer-concentration-remains-a-discount"&gt;6. Customer Concentration Remains a Discount
&lt;/h2&gt;&lt;p&gt;The Q1 business report shows Customer 1 at 51.53% of revenue. A separate financial-note classification shows Customer A at 49.15% and Customer B at 31.87%. The classifications differ, but both show high concentration. &lt;a class="link" href="https://kind.krx.co.kr/external/2026/05/12/000812/20260512001819/11013.htm" target="_blank" rel="noopener"
 &gt;FADU Q1 2026 business report&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Customer qualification can create recurring revenue once production begins, but it also makes quarterly results sensitive to a small number of order schedules. The new server OEM and KRW 340bn of orders are positive. Diversification is not proved until the largest customer falls from the 50% range toward 30% of reported revenue.&lt;/p&gt;
&lt;h2 id="7-moat-score-305-narrow-but-expandable"&gt;7. Moat Score: 3.0/5, Narrow but Expandable
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Area&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Score&lt;/th&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Controller architecture and firmware&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.0/5&lt;/td&gt;
 &lt;td&gt;Proprietary design and commercial Gen5 revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power efficiency and QoS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.5/5&lt;/td&gt;
 &lt;td&gt;Strong company and OCP evidence, limited independent comparison&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Custom development and qualification&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.5/5&lt;/td&gt;
 &lt;td&gt;Long process creates barriers, concentration remains high&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;OCP and FDP ecosystem&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.5/5&lt;/td&gt;
 &lt;td&gt;Public evidence for first FDP-enabled OCP SSD&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer and NAND partner scale&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.5/5&lt;/td&gt;
 &lt;td&gt;Expanding, but still concentrated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Gen6 mass-production proof&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.5/5&lt;/td&gt;
 &lt;td&gt;Tape-out and security evaluation, no public production order&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Commercial QLC adoption&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.0/5&lt;/td&gt;
 &lt;td&gt;Technical support claimed, no 64TB/128TB revenue evidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Overall&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;3.0/5&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Narrow moat&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;There is not enough evidence to call this a wide moat. There is also too much commercial and ecosystem evidence to dismiss FADU as a theme stock without technology. The company is expanding a real but narrow Gen5 position into the next generation.&lt;/p&gt;
&lt;h2 id="8-translating-the-preliminary-beat-into-eps"&gt;8. Translating the Preliminary Beat Into EPS
&lt;/h2&gt;&lt;p&gt;The preliminary filing does not disclose net income. The following is a mechanical bridge, not a revised broker consensus.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Step&lt;/th&gt;
 &lt;th&gt;Calculation&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Result&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Prior 2026 EPS&lt;/td&gt;
 &lt;td&gt;Naver WiseReport-based snapshot&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,817&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating-profit upside&lt;/td&gt;
 &lt;td&gt;KRW 16.27bn actual minus KRW 8.6bn consensus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 7.67bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;After-tax conversion&lt;/td&gt;
 &lt;td&gt;Upside multiplied by 76%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 5.83bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Per-share increase&lt;/td&gt;
 &lt;td&gt;KRW 5.83bn divided by about 50.08m shares&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 116&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mechanical revised EPS&lt;/td&gt;
 &lt;td&gt;KRW 1,817 plus KRW 116&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;About KRW 1,930&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 P/E&lt;/td&gt;
 &lt;td&gt;KRW 70,400 divided by KRW 1,930&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;36.5x&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Adding the Q2 operating-profit upside to the prior 2026 operating-profit estimate of KRW 85.9bn, while leaving second-half forecasts unchanged, gives about KRW 93.6bn. Tax, currency, financial gains and tax credits can change the EPS conversion. The local snapshot also lacks a populated analyst-count field, so the estimate set should be treated as thin.&lt;/p&gt;
&lt;h2 id="9-a-scenario-range-not-a-single-fair-value"&gt;9. A Scenario Range, Not a Single Fair Value
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Earnings assumption&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Multiple&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Price range&lt;/th&gt;
 &lt;th&gt;Required conditions&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bear&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2027 EPS KRW 3,400-3,800&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16-18x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 54,000-68,000&lt;/td&gt;
 &lt;td&gt;Slower growth, Gen6 delay, continued concentration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base&lt;/td&gt;
 &lt;td style="text-align: right"&gt;EPS KRW 4,000-4,300&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18-20x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;KRW 72,000-86,000&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Gen5 growth, backlog conversion, Gen6 progress&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bull&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2027-28 average EPS around KRW 4,600&lt;/td&gt;
 &lt;td style="text-align: right"&gt;24-28x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 110,000-129,000&lt;/td&gt;
 &lt;td&gt;New hyperscaler, Gen6 production, QLC and diversification&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The July 22 close of KRW 70,400 is near the lower end of the base case. The stock remains expensive on mechanically revised 2026 EPS, but modestly discounted if 2027 EPS reaches KRW 4,000 or more. The KRW 130,000 market target is closer to a bull case than to a valuation supported by current earnings alone.&lt;/p&gt;
&lt;h2 id="10-why-the-stock-fell-after-strong-results"&gt;10. Why the Stock Fell After Strong Results
&lt;/h2&gt;&lt;p&gt;FADU traded as high as KRW 79,800 on the release date but closed at KRW 71,600. It rebounded to KRW 77,600 on July 21 and then fell to KRW 70,400 on July 22.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreigners&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Institutions&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Individuals&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Latest 3 sessions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 20.79bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 9.67bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 8.20bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Latest 5 sessions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 37.20bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 18.97bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 14.83bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Latest 10 sessions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 70.28bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 49.28bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 15.24bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Investment trusts purchased about 494,000 shares over ten sessions. However, both foreigners and institutions sold on July 22, while program selling totaled KRW 17.88bn over the three post-earnings sessions.&lt;/p&gt;
&lt;p&gt;Short-sale value represented an average of roughly 4.9% of turnover over the latest five sessions and 3.0% on July 22. New short selling alone does not explain the 9.3% decline. Profit-taking and program flows matter more. Securities-lending balances rose from 1.08m shares on July 16 to 1.53m on July 21.&lt;/p&gt;
&lt;p&gt;Flow, program, short-sale and lending figures come from the Research OS local database and Kiwoom REST provider layer. They are useful directional data, but the short-sale and lending series have not been fully reconciled with official KRX source data. Short-interest balances are unavailable.&lt;/p&gt;
&lt;h2 id="11-execution-framework"&gt;11. Execution Framework
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Dimension&lt;/th&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fundamentals&lt;/td&gt;
 &lt;td&gt;Upgraded&lt;/td&gt;
 &lt;td&gt;Operating profit and margin exceeded the prior bull case&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Valuation&lt;/td&gt;
 &lt;td&gt;Neutral to modestly discounted&lt;/td&gt;
 &lt;td&gt;Expensive on 2026, near low end of 2027 base case&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flows and chart&lt;/td&gt;
 &lt;td&gt;Weak&lt;/td&gt;
 &lt;td&gt;Foreign and program selling, higher lending balance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Approach&lt;/td&gt;
 &lt;td&gt;Confirmation-based&lt;/td&gt;
 &lt;td&gt;Earnings are visible, flow reversal is not&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Three price zones can be used as confirmation levels rather than forecasts.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;KRW 70,000 support&lt;/strong&gt;: close to the July 22 intraday low of KRW 70,200. Stabilization and slower foreign selling would improve the setup.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;KRW 76,800-77,600 recovery&lt;/strong&gt;: regaining the post-earnings supply zone with lower foreign or program selling would provide stronger confirmation.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;KRW 63,900-65,000 risk zone&lt;/strong&gt;: a failure to recover after losing KRW 70,000 could open a retest.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;These levels are a public analytical framework, not personalized investment advice.&lt;/p&gt;
&lt;h2 id="12-four-pieces-of-evidence-that-would-prove-a-wider-moat"&gt;12. Four Pieces of Evidence That Would Prove a Wider Moat
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;FC6161 Gen6 customer qualification and production orders&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A 64TB or 128TB QLC design win&lt;/strong&gt; that turns claimed support into customer revenue.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Largest-customer share falling toward 30%&lt;/strong&gt; as new OEMs and hyperscalers contribute.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Independent evidence&lt;/strong&gt; of p99.99 latency, WAF and performance-per-watt superiority under equal conditions.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Two or more of these would materially improve confidence in the 2027 base case. A Gen6 delay beyond 2027, no QLC design win and persistent customer concentration would compress the multiple even if Gen5 remains profitable.&lt;/p&gt;
&lt;h2 id="13-red-team"&gt;13. Red Team
&lt;/h2&gt;&lt;p&gt;First, the 22.2% margin may not repeat. Product mix, revenue timing and lower one-time spending could have helped the quarter. The half-year report must confirm gross margin and controller mix.&lt;/p&gt;
&lt;p&gt;Second, integrated suppliers have scale advantages. Micron and Samsung can optimize NAND, controller, firmware and finished SSD together. Their position may strengthen as Gen6 and 122TB-245TB QLC products become mainstream.&lt;/p&gt;
&lt;p&gt;Third, standardization can lower qualification barriers. OCP and Meta&amp;rsquo;s open tools helped FADU enter the market, but they can also help future competitors.&lt;/p&gt;
&lt;p&gt;Fourth, orders are not cash. Purchase orders, delivery, acceptance, revenue recognition and collection occur at different times. Operating cash flow and receivables must keep pace with reported profit.&lt;/p&gt;
&lt;h2 id="14-conclusion"&gt;14. Conclusion
&lt;/h2&gt;&lt;p&gt;Q2 strongly demonstrated that FADU can earn money from its Gen5 controller business. The 22.2% operating margin is more important than KRW 73.16bn of revenue because it suggests a favorable controller mix and operating leverage.&lt;/p&gt;
&lt;p&gt;The quarter did not prove Gen6 production, QLC design wins or customer diversification. FADU&amp;rsquo;s current moat is 3.0 out of 5: narrow, commercially real and capable of expanding. The stock is not cheap on 2026 earnings alone. It enters the lower end of the base-case valuation only if 2027 EPS can exceed KRW 4,000.&lt;/p&gt;
&lt;p&gt;The most precise description is this:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;FADU is a high-performance enterprise SSD controller supplier with proven Gen5 commercialization and an option to become a broader Gen6 and QLC platform. The Q2 surprise proved the profitability of the current business, not the success of that future option.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="evidence-map"&gt;Evidence Map
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Classification&lt;/th&gt;
 &lt;th&gt;Items&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Confirmed facts&lt;/td&gt;
 &lt;td&gt;Preliminary Q2 revenue and operating profit, Gen5 controller revenue, OCP FDP listing, FC6161 tape-out/security listing, Q1 customer concentration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Analysis&lt;/td&gt;
 &lt;td&gt;Meaning of the margin beat, mechanical EPS bridge, 2027 scenario range, 3.0/5 moat score&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Unconfirmed&lt;/td&gt;
 &lt;td&gt;Q2 product gross margin, net income and cash flow; Gen6 qualification and orders; 64TB/128TB QLC adoption; independent performance comparison&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>Big Tech AI Earnings: 2027 Demand Is Contracted, but 2028 Free Cash Flow Is the Real Test</title><link>https://koreainvestinsights.com/post/big-tech-ai-earnings-capex-roi-memory-2028-fcf-2026-07-22/</link><pubDate>Wed, 22 Jul 2026 17:30:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/big-tech-ai-earnings-capex-roi-memory-2028-fcf-2026-07-22/</guid><description>&lt;p&gt;Amazon, Microsoft, and Alphabet plan to spend roughly $570-580 billion on capital investment in 2026. Their latest reported cloud growth rates were 28%, 39% in constant currency, and 63%, respectively. The market is no longer debating whether current AI demand is strong. It is asking whether the data centers built now will be well utilized, whether revenue can outrun depreciation and power costs, and whether memory profits can survive the supply response expected from 2028.&lt;/p&gt;
&lt;p&gt;The last four quarters suggest that much of the demand through 2027 is already supported by contracts, reservations, and supply constraints. Beyond 2028, scarcity is no longer enough. Utilization, inference revenue, customer concentration, and cash collection become the decisive variables.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Related context
This is a follow-up to the &lt;a class="link" href="https://koreainvestinsights.com/en/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/" &gt;July Big Tech earnings and memory thesis preview&lt;/a&gt;. The financing side is covered in &lt;a class="link" href="https://koreainvestinsights.com/en/post/hyperscaler-financing-race-ai-capex-memory-bottleneck-2026-07-07/" &gt;Hyperscaler financing and the memory bottleneck&lt;/a&gt;, while the semiconductor supply response is developed in the &lt;a class="link" href="https://koreainvestinsights.com/en/post/weekly-semiconductor-deep-dive-strong-earnings-lower-multiple-2028-pivot-2026-07-19/" &gt;weekly semiconductor deep dive&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="summary"&gt;Summary
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;AI demand is likely to remain stronger than expected through 2027. Backlogs, long-term commitments, supply shortages, and customer expansion point in the same direction.&lt;/li&gt;
&lt;li&gt;The most durable demand is likely to come from enterprise inference, agents, coding, search, and data services rather than one-off frontier training runs.&lt;/li&gt;
&lt;li&gt;The three companies plan approximately $570-580 billion of 2026 capital spending. Spending can grow again in 2027, although the growth rate should slow from an unusually high base.&lt;/li&gt;
&lt;li&gt;Return on investment should be measured through incremental cloud gross profit and first-party revenue or cost savings, less depreciation, power, and financing costs.&lt;/li&gt;
&lt;li&gt;Current evidence supports the highest probability of post-2028 free cash flow recovery at Alphabet, followed by Microsoft and Amazon. Alphabet shows the clearest combination of monetization and unit-cost improvement. Microsoft has the strongest contracted visibility but more concentration and lease complexity. Amazon has the largest rebound option but the greatest current cash-flow pressure.&lt;/li&gt;
&lt;li&gt;Memory purchases should continue across HBM, server DRAM, and enterprise SSDs. Custom accelerators change the product mix but do not remove the need for memory.&lt;/li&gt;
&lt;li&gt;The principal risk after 2028 is price rather than bit demand. New capacity can reduce average selling prices and margins even if memory shipments keep rising.&lt;/li&gt;
&lt;li&gt;Credit conditions are cautionary, not distressed. Bond books remain covered, but lower cover ratios and wider long-end spreads show that capital markets are beginning to impose discipline.&lt;/li&gt;
&lt;/ul&gt;
&lt;hr&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core judgment&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Through 2027, supply constraints and contracted demand support continued spending. From 2028, the decisive variables shift to utilization, cash collection, and returns after depreciation. For memory, strong volume and normalizing prices can coexist.
 &lt;/div&gt;
&lt;/div&gt;
&lt;hr&gt;
&lt;h2 id="1-the-three-earnings-calls-and-the-questions-that-matter"&gt;1. The three earnings calls and the questions that matter
&lt;/h2&gt;&lt;p&gt;Alphabet reports at 5:30 a.m. KST on July 23, Microsoft at 6:30 a.m. KST on July 30, and Amazon at 6:00 a.m. KST on July 31, based on company investor-relations announcements.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Earnings time&lt;/th&gt;
 &lt;th&gt;Primary question&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td&gt;July 23, 05:30 KST&lt;/td&gt;
 &lt;td&gt;How quickly do 63% cloud growth and a $462.3 billion backlog convert into revenue and cash?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td&gt;July 30, 06:30 KST&lt;/td&gt;
 &lt;td&gt;When does Azure supply improve, and does backlog growth reaccelerate outside OpenAI?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td&gt;July 31, 06:00 KST&lt;/td&gt;
 &lt;td&gt;Can AWS protect margins and restore free cash flow while Amazon spends about $200 billion?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The market is trying to answer three questions. How long can AI demand grow above expectations, and from which customers and workloads? How much longer can capital spending rise, and can the investments earn adequate returns? Will that spending translate into recurring purchases of HBM, server DRAM, and enterprise SSDs?&lt;/p&gt;
&lt;p&gt;The last four quarters provide a constructive answer to the first question. The second and third require company-specific analysis because a dollar spent on accelerators has a different return profile from a dollar spent on land, power, networking, or long-lived buildings.&lt;/p&gt;
&lt;h2 id="2-four-quarters-of-operating-evidence"&gt;2. Four quarters of operating evidence
&lt;/h2&gt;&lt;h3 id="21-amazon-aws-accelerated-while-cash-left-first"&gt;2.1 Amazon: AWS accelerated while cash left first
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;AWS revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YoY growth&lt;/th&gt;
 &lt;th style="text-align: right"&gt;AWS operating income&lt;/th&gt;
 &lt;th&gt;Cash-flow signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$30.9B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$10.2B&lt;/td&gt;
 &lt;td&gt;TTM FCF $18.2B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$33.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$11.4B&lt;/td&gt;
 &lt;td&gt;AI infrastructure spending kept rising&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$35.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$12.4B&lt;/td&gt;
 &lt;td&gt;Amazon signaled about $200B of 2026 investment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$37.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$14.2B&lt;/td&gt;
 &lt;td&gt;TTM FCF fell to $1.2B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;AWS growth accelerated for four consecutive quarters. Its Q1 operating margin reached 37.8%, showing that revenue and profit expanded despite supply constraints.&lt;/p&gt;
&lt;p&gt;The cash profile moved in the opposite direction. Amazon generated $148.5 billion of trailing operating cash flow, but net property and equipment spending reached $147.3 billion. AWS-centered long-term commitments stood at $364 billion with a weighted-average remaining life of 5.5 years. Revenue visibility is high, but spending occurs before the associated revenue arrives.&lt;/p&gt;
&lt;p&gt;Amazon&amp;rsquo;s 2025 shareholder letter stated that the company was not spending $200 billion on a hunch and had substantial customer commitments. Graviton, Trainium, and Nitro together exceeded a $20 billion annual revenue run rate and were growing at a triple-digit rate. Custom silicon is both a cost-control tool and a way to keep workloads inside AWS.&lt;/p&gt;
&lt;p&gt;Amazon can produce the largest cash-flow rebound if the new assets fill quickly, but it also has the widest gap to close. AWS needs to maintain growth around the high-20% range and defend margins in the mid-to-high 30s while 2026 investments begin monetizing in 2027 and 2028.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-First-Quarter-Results/default.aspx" target="_blank" rel="noopener"
 &gt;Amazon Q1 2026 results&lt;/a&gt;, &lt;a class="link" href="https://www.aboutamazon.com/news/company-news/amazon-ceo-andy-jassy-2025-letter-to-shareholders" target="_blank" rel="noopener"
 &gt;Amazon 2025 shareholder letter&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1018724/000101872426000014/amzn-20260331.htm" target="_blank" rel="noopener"
 &gt;Amazon Q1 2026 10-Q&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="22-microsoft-the-strongest-contract-visibility-and-the-largest-concentration-question"&gt;2.2 Microsoft: the strongest contract visibility and the largest concentration question
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Azure growth, constant currency&lt;/th&gt;
 &lt;th style="text-align: right"&gt;RPO&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Capital spending&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Free cash flow&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FY25 Q4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$76.4B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about +39%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$368B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$24.2B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$25.6B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY26 Q1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$77.7B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+39%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$392B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$34.9B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$25.7B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY26 Q2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$81.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+38%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$625B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$37.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5.9B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY26 Q3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$82.9B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+39%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$627B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$31.9B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$15.8B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Azure grew 40% as reported and 39% in constant currency in the latest quarter. Microsoft&amp;rsquo;s AI business exceeded a $37 billion annual revenue run rate and grew 123%. Commercial RPO reached $627 billion with an average duration of 2.5 years, and roughly 25% is expected to convert to revenue over the next 12 months.&lt;/p&gt;
&lt;p&gt;Contract quality needs to be separated from contract size. OpenAI accounted for approximately 45% of commercial RPO in FY26 Q2, while RPO excluding OpenAI grew 26% in FY26 Q3. A major strategic customer and a broader enterprise expansion are both present.&lt;/p&gt;
&lt;p&gt;Microsoft also has the most complex capital-spending definition. Q3 cash property and equipment spending was $30.9 billion, while spending including finance leases reached $31.9 billion. Management said roughly two-thirds of capital spending went to shorter-lived assets such as GPUs and CPUs, with one-third allocated to land and buildings. Short-lived assets monetize earlier but also return to the replacement cycle sooner.&lt;/p&gt;
&lt;p&gt;Azure is expected to remain supply constrained through 2026, with modest growth acceleration possible in the second half of the calendar year. Q4 capital spending is expected to exceed $40 billion. Calendar 2026 investment is approximately $190 billion, including about $25 billion of component inflation. Microsoft expects double-digit revenue and operating-income growth in FY27.&lt;/p&gt;
&lt;p&gt;The company has the strongest contracted visibility, but post-2028 free cash flow depends on independent enterprise demand, Copilot and agent usage, lease-adjusted cash generation, and whether gross profit grows faster than depreciation.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://www.microsoft.com/en-us/investor/events/fy-2026/earnings-fy-2026-q3" target="_blank" rel="noopener"
 &gt;Microsoft FY26 Q3 earnings call&lt;/a&gt;, &lt;a class="link" href="https://www.microsoft.com/en-us/investor/earnings/FY-2026-Q3/press-release-webcast" target="_blank" rel="noopener"
 &gt;Microsoft FY26 Q3 results&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="23-alphabet-demand-margin-and-unit-cost-improvement-are-visible-together"&gt;2.3 Alphabet: demand, margin, and unit-cost improvement are visible together
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Google Cloud revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YoY growth&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Operating income&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Margin&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Backlog&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Capital spending&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$13.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+32%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.8B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$106B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$22.4B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$15.2B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+34%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$3.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$155B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$24.0B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$17.7B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+48%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$240B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$27.9B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$20.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+63%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$6.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;32.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$462.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$35.7B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Alphabet delivered the clearest simultaneous improvement. Cloud growth accelerated from 32% to 63%, the operating margin rose from 20.7% to 32.9%, and backlog increased from $106 billion to $462.3 billion.&lt;/p&gt;
&lt;p&gt;Q1 generative AI product revenue grew 800%. New customers doubled, deals between $100 million and $1 billion doubled, and Alphabet signed multiple deals above $1 billion. Existing customers consumed 45% more than their original commitments. Gemini Enterprise paid monthly users rose 40% quarter over quarter.&lt;/p&gt;
&lt;p&gt;Unit economics also improved. Alphabet said the cost per AI response fell by more than 30%. Higher demand and lower unit cost are the best combination for infrastructure returns. The same infrastructure supports Search, YouTube, Workspace, and advertising, so external cloud revenue understates part of the return.&lt;/p&gt;
&lt;p&gt;Alphabet guides to $180-190 billion of 2026 capital spending and expects another significant increase in 2027. Q1 uncommenced asset commitments were approximately $108.6 billion, and uncommenced data-center leases were $75.6 billion, scheduled to begin between 2026 and 2031. Total RPO was $467.6 billion, with just over half expected to be recognized within 24 months.&lt;/p&gt;
&lt;p&gt;The main risk is the speed of spending itself. Q1 operating cash flow was $45.8 billion and capital spending was $35.7 billion, leaving quarterly free cash flow of $10.1 billion. Trailing free cash flow remained a substantial $64.4 billion, but another large increase in 2027 spending could delay the recovery.&lt;/p&gt;
&lt;p&gt;Current evidence nevertheless places Alphabet first in proving AI investment returns. The next call needs to clarify how much backlog growth came from a few large contracts and how quickly new facilities are filling.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://abc.xyz/investor/events/event-details/2026/2026-Q1-Earnings-Call-2026-nW8kCrBAKS/default.aspx" target="_blank" rel="noopener"
 &gt;Alphabet Q1 2026 earnings call&lt;/a&gt;, &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1652044/000165204426000048/goog-20260331.htm" target="_blank" rel="noopener"
 &gt;Alphabet Q1 2026 10-Q&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="3-where-ai-demand-can-persist"&gt;3. Where AI demand can persist
&lt;/h2&gt;&lt;p&gt;AI demand should be divided into five layers with different customers and durations.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Demand layer&lt;/th&gt;
 &lt;th&gt;Customers&lt;/th&gt;
 &lt;th&gt;Expected durability&lt;/th&gt;
 &lt;th&gt;Key evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Frontier training and post-training&lt;/td&gt;
 &lt;td&gt;Leading model labs and first-party models&lt;/td&gt;
 &lt;td&gt;Strong but concentrated&lt;/td&gt;
 &lt;td&gt;Cluster reservations, long-term power contracts, model roadmaps&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Enterprise inference and agents&lt;/td&gt;
 &lt;td&gt;Finance, retail, manufacturing, healthcare, government&lt;/td&gt;
 &lt;td&gt;Most durable candidate&lt;/td&gt;
 &lt;td&gt;Paid users, calls, spend per customer, renewals&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Development and coding&lt;/td&gt;
 &lt;td&gt;Software companies and developers&lt;/td&gt;
 &lt;td&gt;Rapid adoption, strong price competition&lt;/td&gt;
 &lt;td&gt;Seats, agent runtime, generated code, retention&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Data, storage, and general cloud&lt;/td&gt;
 &lt;td&gt;Existing cloud customers&lt;/td&gt;
 &lt;td&gt;Stable and bundled with AI&lt;/td&gt;
 &lt;td&gt;Database, storage, networking, migrations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First-party services&lt;/td&gt;
 &lt;td&gt;Search, advertising, productivity, commerce, recommendations&lt;/td&gt;
 &lt;td&gt;Revenue and cost savings combined&lt;/td&gt;
 &lt;td&gt;Conversion, cost per response, productivity, gross margin&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="31-supply-should-remain-the-constraint-through-2027"&gt;3.1 Supply should remain the constraint through 2027
&lt;/h3&gt;&lt;p&gt;AWS, Azure, and Google Cloud have all described supply constraints. Alphabet customers used 45% more than their initial commitments. Microsoft expects constraints through the end of 2026. Amazon&amp;rsquo;s long-term commitments have an average 5.5 years remaining.&lt;/p&gt;
&lt;p&gt;This supports a capacity-led revenue model through at least 2027. As data centers open and power is connected, revenue should follow. Near-term spending reductions are also difficult because construction, leases, power, servers, and memory are already under contract.&lt;/p&gt;
&lt;h3 id="32-2028-is-more-likely-to-be-a-change-in-demand-quality-than-a-demand-cliff"&gt;3.2 2028 is more likely to be a change in demand quality than a demand cliff
&lt;/h3&gt;&lt;p&gt;More facilities and semiconductor supply arrive from 2028. Once scarcity eases, investors must test whether customers actually use reserved capacity, renew contracts, and increase usage fast enough to offset lower inference prices.&lt;/p&gt;
&lt;p&gt;Demand can remain high while profitability falls. Model efficiency and custom silicon reduce cost per task, while cloud competition can lower revenue per token. If call volume grows faster than pricing declines, utilization and revenue still rise.&lt;/p&gt;
&lt;p&gt;The useful relationship is:&lt;/p&gt;
&lt;p&gt;&lt;code&gt;AI revenue growth = workload growth + customer growth - price decline per workload&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;Enterprise agents, coding, and embedded inference have the best chance of producing high-frequency, recurring use.&lt;/p&gt;
&lt;h3 id="33-independent-customers-are-the-most-important-proof"&gt;3.3 Independent customers are the most important proof
&lt;/h3&gt;&lt;p&gt;Strategic customer contracts support early demand, but reciprocal investment and revenue commitments can overstate final demand. Confidence after 2028 requires more usage from enterprises, governments, developers, and smaller customers without financing links to the infrastructure provider.&lt;/p&gt;
&lt;p&gt;Alphabet&amp;rsquo;s doubled new-customer count, doubled large-deal count, and commitment overages are the clearest current evidence. Microsoft&amp;rsquo;s 26% RPO growth excluding OpenAI is more important than total RPO alone. Amazon needs to disclose more about how much of AWS commitments comes from AI versus general cloud demand.&lt;/p&gt;
&lt;h2 id="4-how-much-longer-can-capital-spending-grow"&gt;4. How much longer can capital spending grow?
&lt;/h2&gt;&lt;h3 id="41-approximately-570-580-billion-in-2026"&gt;4.1 Approximately $570-580 billion in 2026
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026 investment plan&lt;/th&gt;
 &lt;th&gt;Definition caveat&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $200B&lt;/td&gt;
 &lt;td&gt;Primarily AI and AWS, with some spending elsewhere&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $190B&lt;/td&gt;
 &lt;td&gt;Cash spending and finance leases can both be included&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$180-190B&lt;/td&gt;
 &lt;td&gt;Servers, data centers, networks, and custom silicon&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $570-580B&lt;/td&gt;
 &lt;td&gt;Accounting definitions differ across companies&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The absolute amount can grow again, but the 60-100% growth rates seen in 2026 are difficult to repeat. The base is larger, while power, sites, transformers, networking equipment, and skilled labor remain constrained.&lt;/p&gt;
&lt;p&gt;The following 2027 ranges are analytical scenarios, not company guidance.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Combined 2027 spending&lt;/th&gt;
 &lt;th&gt;Conditions&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Downside&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$530-580B&lt;/td&gt;
 &lt;td&gt;Slower bookings, power delays, digestion of 2026 prebuys&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$630-665B&lt;/td&gt;
 &lt;td&gt;Continued constraints, slower growth rate, execution of existing commitments&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Upside&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Above $700B&lt;/td&gt;
 &lt;td&gt;Agent usage surge, additional sovereign demand, custom silicon and GPU expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A slower capital-spending growth rate is not by itself a bearish signal. The warning comes when slower spending, cloud growth, and backlog appear together.&lt;/p&gt;
&lt;h3 id="42-a-simple-return-framework"&gt;4.2 A simple return framework
&lt;/h3&gt;&lt;p&gt;&lt;code&gt;AI investment return = incremental cloud gross profit + first-party revenue and savings - depreciation - power - financing cost&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;Capital spending hits the cash-flow statement first. Depreciation enters the income statement after servers and data centers begin operating. Cash-flow troughs and accounting-margin troughs therefore occur at different times.&lt;/p&gt;
&lt;p&gt;Free cash flow is pressured first in 2026. Revenue and depreciation rise together in 2027-2029. Around 2030, accelerators purchased in 2026 can begin returning to the replacement cycle. Cash flow can recover before reported margins fully absorb depreciation.&lt;/p&gt;
&lt;h3 id="43-current-evidence-by-company"&gt;4.3 Current evidence by company
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Positive evidence&lt;/th&gt;
 &lt;th&gt;Remaining risk&lt;/th&gt;
 &lt;th&gt;Confidence in post-2028 FCF recovery&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td&gt;Cloud +63%, margin 32.9%, cost per response down more than 30%&lt;/td&gt;
 &lt;td&gt;Fast spending growth, leases, possible large-deal concentration&lt;/td&gt;
 &lt;td&gt;Highest&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td&gt;Azure +39%, RPO $627B, AI ARR above $37B&lt;/td&gt;
 &lt;td&gt;OpenAI concentration, finance leases, short-lived assets&lt;/td&gt;
 &lt;td&gt;Medium-high&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td&gt;AWS +28%, margin 37.8%, commitments $364B&lt;/td&gt;
 &lt;td&gt;TTM FCF $1.2B, $200B spending, monetization lag&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="5-the-dashboard-for-2028-and-beyond"&gt;5. The dashboard for 2028 and beyond
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th&gt;Constructive signal&lt;/th&gt;
 &lt;th&gt;Warning signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Independent-customer RPO&lt;/td&gt;
 &lt;td&gt;Growth outside strategically financed customers&lt;/td&gt;
 &lt;td&gt;A few customers drive most of the increase&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Next-12-month recognition&lt;/td&gt;
 &lt;td&gt;A larger share of RPO converts soon&lt;/td&gt;
 &lt;td&gt;Contract duration rises while near-term conversion slows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cash collection&lt;/td&gt;
 &lt;td&gt;Cash receipts outgrow contract assets&lt;/td&gt;
 &lt;td&gt;Receivables and contract assets grow without deposits&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating cash flow&lt;/td&gt;
 &lt;td&gt;Four-quarter OCF growth exceeds capex growth by at least 10 points&lt;/td&gt;
 &lt;td&gt;OCF trails investment growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud economics&lt;/td&gt;
 &lt;td&gt;Gross profit growth exceeds depreciation and power growth&lt;/td&gt;
 &lt;td&gt;Revenue grows while margin falls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Utilization&lt;/td&gt;
 &lt;td&gt;New capacity fills rapidly&lt;/td&gt;
 &lt;td&gt;Low utilization after power connection&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Lease-adjusted FCF&lt;/td&gt;
 &lt;td&gt;Recovery remains after finance leases&lt;/td&gt;
 &lt;td&gt;Reported FCF improves while lease obligations rise&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Custom silicon&lt;/td&gt;
 &lt;td&gt;Lower cost and higher customer use&lt;/td&gt;
 &lt;td&gt;R&amp;amp;D rises without adoption&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory contracts&lt;/td&gt;
 &lt;td&gt;Commitments convert to shipments and cash&lt;/td&gt;
 &lt;td&gt;Deferrals, renegotiation, or cancellations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Supply normalization&lt;/td&gt;
 &lt;td&gt;Shipments and utilization hold despite lower prices&lt;/td&gt;
 &lt;td&gt;ASP and utilization fall together&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Replacement cycle&lt;/td&gt;
 &lt;td&gt;First-generation assets earn returns before replacement&lt;/td&gt;
 &lt;td&gt;Replacement begins before initial payback&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The strongest combination is independent RPO growth, new capacity going online, stable cloud margins, and OCF growth overtaking capital-spending growth. The most dangerous combination is easing supply constraints, slower growth, higher depreciation, rising customer concentration, and increasing debt or leases.&lt;/p&gt;
&lt;h2 id="6-credit-markets-are-imposing-discipline-not-signaling-a-crisis"&gt;6. Credit markets are imposing discipline, not signaling a crisis
&lt;/h2&gt;&lt;p&gt;Apollo reports that hyperscaler bond cover ratios fell from around 5 times in February 2026 to below 2 times in July. Demand still exceeds supply, but investors have become more selective.&lt;/p&gt;
&lt;p&gt;Amazon issued $25 billion of bonds on July 7. Orders peaked near $62 billion and ended around $41 billion after pricing. Its longest maturities offered an additional 18-21 basis points of yield. Funding remained available, but at a higher cost.&lt;/p&gt;
&lt;p&gt;The Bank of England&amp;rsquo;s July 2026 Financial Stability Report said AI-related issuance had not yet impaired broader investment-grade terms or availability. Barclays projected approximately $240 billion of 2026 investment-grade hyperscaler financing, comparable to major bank issuance.&lt;/p&gt;
&lt;p&gt;The signal is yellow, not red. Capital markets remain open but increasingly demand backlog quality, utilization, and cash conversion. This should delay weaker projects and widen capital-cost differences rather than end AI spending outright.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://www.apollo.com/wealth/insights-news/insights/daily-spark/cover-ratios-for-hyperscaler-bonds-declining" target="_blank" rel="noopener"
 &gt;Apollo hyperscaler bond cover ratios&lt;/a&gt;, &lt;a class="link" href="https://www.bankofengland.co.uk/financial-stability-report/2026/july-2026" target="_blank" rel="noopener"
 &gt;Bank of England July 2026 Financial Stability Report&lt;/a&gt;, &lt;a class="link" href="https://fortune.com/2026/07/08/amazons-25-billion-surprise-bond-sale-dangled-extra-yield-to-lure-in-buyers/" target="_blank" rel="noopener"
 &gt;Amazon bond demand&lt;/a&gt;&lt;/p&gt;
&lt;h2 id="7-does-the-spending-become-recurring-memory-demand"&gt;7. Does the spending become recurring memory demand?
&lt;/h2&gt;&lt;h3 id="71-ai-systems-require-a-memory-hierarchy"&gt;7.1 AI systems require a memory hierarchy
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Memory layer&lt;/th&gt;
 &lt;th&gt;Function&lt;/th&gt;
 &lt;th&gt;Demand direction&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM&lt;/td&gt;
 &lt;td&gt;Bandwidth close to GPUs and ASICs&lt;/td&gt;
 &lt;td&gt;Strongest in HBM3E and HBM4&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server DRAM&lt;/td&gt;
 &lt;td&gt;CPU work, agent orchestration, preprocessing, caches, metadata&lt;/td&gt;
 &lt;td&gt;Expands with agents and multi-model systems&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Enterprise SSD and NAND&lt;/td&gt;
 &lt;td&gt;Checkpoints, vector databases, documents, video, long-context storage&lt;/td&gt;
 &lt;td&gt;Expands with inference volume and retention&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Training is HBM intensive. Enterprise inference and agents expand server DRAM and SSD demand alongside HBM. Multiple agents repeat tasks and preserve longer context, increasing memory outside the accelerator.&lt;/p&gt;
&lt;p&gt;TPUs, Trainium, and Maia can displace some external GPUs but still require HBM, server DRAM, and storage. Custom silicon changes specifications, packaging, and bargaining power rather than eliminating memory demand.&lt;/p&gt;
&lt;h3 id="72-procurement-is-indirect-and-arrives-with-a-lag"&gt;7.2 Procurement is indirect and arrives with a lag
&lt;/h3&gt;&lt;p&gt;Hyperscalers do not purchase all HBM directly. Procurement runs through Nvidia, custom ASIC partners, server ODMs, and packaging suppliers. Capital spending therefore reaches memory-vendor revenue with a lag through design, allocation, wafer starts, packaging, server assembly, and installation.&lt;/p&gt;
&lt;p&gt;Micron&amp;rsquo;s strategic customer agreements illustrate the connection. It has 16 agreements, generally covering 2026-2030, representing roughly 20% of DRAM volume and one-third of NAND volume. Fourteen agreements contain approximately $100 billion of cumulative minimum-price revenue, supported by about $22 billion of customer deposits and financing commitments.&lt;/p&gt;
&lt;p&gt;Long-term agreements raise the earnings floor. Take-or-pay and minimum prices reduce cancellation and downside pricing. Ceiling prices can also limit upside when spot prices surge. These contracts reduce the severity of a downturn but do not permanently lock in today&amp;rsquo;s scarcity margins.&lt;/p&gt;
&lt;p&gt;Source: &lt;a class="link" href="https://investors.micron.com/static-files/631b1a32-5537-46ae-8f40-82e42fc79dfe" target="_blank" rel="noopener"
 &gt;Micron FY3Q26 prepared remarks&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="73-the-2028-memory-risk-is-price-not-necessarily-demand"&gt;7.3 The 2028 memory risk is price, not necessarily demand
&lt;/h3&gt;&lt;p&gt;Additional DRAM, NAND, and HBM packaging capacity should arrive from 2027 and 2028. Bit shipments can keep growing while average selling prices and margins normalize.&lt;/p&gt;
&lt;p&gt;&lt;code&gt;Memory profit = bit shipments × price per bit - expansion, depreciation, and input costs&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;In 2026 and 2027, demand is growing faster than supply, supporting both volume and price. From 2028, supply growth can reduce pricing even if bit demand stays high. Micron&amp;rsquo;s current gross margin above 80% reflects scarcity and should not be treated as a steady-state margin.&lt;/p&gt;
&lt;p&gt;The semiconductor conclusion therefore requires more than a direction for Big Tech capital spending. Utilization, memory supply growth, contract pricing, inventory, and producer capital discipline must be read together.&lt;/p&gt;
&lt;h2 id="8-company-specific-earnings-checklist"&gt;8. Company-specific earnings checklist
&lt;/h2&gt;&lt;h3 id="alphabet"&gt;Alphabet
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Cloud growth near 60%, with both new customers and usage increasing.&lt;/li&gt;
&lt;li&gt;Backlog rising from $462.3 billion without a deterioration in near-term recognition.&lt;/li&gt;
&lt;li&gt;Continued explanation for $180-190 billion of 2026 spending and the 2027 step-up.&lt;/li&gt;
&lt;li&gt;Cloud margin staying above 30% and cost per response falling further.&lt;/li&gt;
&lt;li&gt;Evidence that external TPU sales and internal usage both scale.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="microsoft"&gt;Microsoft
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Azure growth near 39-40% after additional supply comes online.&lt;/li&gt;
&lt;li&gt;RPO growth reaccelerating outside OpenAI.&lt;/li&gt;
&lt;li&gt;More than $40 billion of quarterly spending converting into revenue quickly.&lt;/li&gt;
&lt;li&gt;Paid Copilot and agent seats accompanied by actual usage and revenue.&lt;/li&gt;
&lt;li&gt;Operating cash flow growing faster than capital spending and finance leases.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="amazon"&gt;Amazon
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;AWS growth in the high 20s and margins in the mid-to-high 30s.&lt;/li&gt;
&lt;li&gt;Faster conversion of the $364 billion commitment base.&lt;/li&gt;
&lt;li&gt;Trainium adoption and revenue demonstrating a cost advantage over external GPUs.&lt;/li&gt;
&lt;li&gt;A clear 2027-2028 monetization schedule for roughly $200 billion of investment.&lt;/li&gt;
&lt;li&gt;Operating cash flow rising while net investment growth slows.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="9-three-paths-for-2028-2031"&gt;9. Three paths for 2028-2031
&lt;/h2&gt;&lt;p&gt;These probabilities are analytical judgments, not company guidance.&lt;/p&gt;
&lt;h3 id="base-case-55"&gt;Base case, 55%
&lt;/h3&gt;&lt;p&gt;Capital spending rises again in 2027 but at a slower rate. New data centers fill, and enterprise inference and agent revenue expand. Alphabet recovers free cash flow first, followed by Microsoft. Amazon recovers later because of a longer monetization lag.&lt;/p&gt;
&lt;p&gt;Semiconductor revenue remains high, but new supply normalizes memory prices and margins from 2028. HBM and server DRAM volumes remain strong while explosive spot-price increases fade.&lt;/p&gt;
&lt;h3 id="upside-case-20"&gt;Upside case, 20%
&lt;/h3&gt;&lt;p&gt;Agent, coding, search, and enterprise inference usage grows faster than unit-price declines. Sovereign and enterprise customers add demand. Combined 2027 capital spending exceeds $700 billion while operating cash flow also expands.&lt;/p&gt;
&lt;p&gt;HBM4, advanced packaging, server DRAM, and enterprise SSD shortages persist into 2028. Memory contract volumes and prices rise, producing another round of semiconductor earnings upgrades.&lt;/p&gt;
&lt;h3 id="downside-case-25"&gt;Downside case, 25%
&lt;/h3&gt;&lt;p&gt;Model efficiency and price competition reduce revenue per task quickly. Strategic contracts are renegotiated, and new data centers operate below expected utilization. Depreciation and power costs are already committed, pressuring margins and free cash flow.&lt;/p&gt;
&lt;p&gt;Memory supply added in 2027-2029 exceeds demand growth, pushing down ASP and utilization together. Long-term agreements soften the decline but do not protect uncontracted volume and spot pricing.&lt;/p&gt;
&lt;h2 id="10-conclusion"&gt;10. Conclusion
&lt;/h2&gt;&lt;p&gt;AI demand is likely to remain stronger than expected through 2027. Cloud growth, backlogs, supply constraints, and long-term infrastructure commitments support that view. The customer base is broadening from frontier labs to enterprise inference, coding, search, and general cloud workloads.&lt;/p&gt;
&lt;p&gt;Capital spending should remain elevated in 2027, although its growth rate can slow. A slowdown from a very high base is not a cycle break. The real warning would be simultaneous weakness in backlog, utilization, cloud margins, and operating cash flow.&lt;/p&gt;
&lt;p&gt;Current evidence gives Alphabet the best chance of post-2028 free cash flow recovery. It has shown demand growth, margin improvement, and lower unit costs together. Microsoft has stronger contract visibility but more OpenAI concentration and finance-lease complexity. Amazon has powerful AWS economics and significant rebound potential, but it must close the gap between $200 billion of investment and trailing free cash flow of only $1.2 billion.&lt;/p&gt;
&lt;p&gt;Memory purchases should continue, but the decisive post-2028 issue is pricing after supply expands. Custom silicon and inference efficiency change the mix across HBM, server DRAM, and enterprise SSDs rather than eliminating demand. Strategic customer agreements raise the floor but do not make current scarcity margins permanent.&lt;/p&gt;
&lt;p&gt;The most important figures in these earnings calls are not capital spending totals alone. Independent-customer backlog, next-12-month conversion, utilization, cloud margins, and lease-adjusted free cash flow must move together. If they improve, 2028 becomes the start of cash harvesting. If customer concentration rises while utilization and cash collection lag, today&amp;rsquo;s earnings strength can return as tomorrow&amp;rsquo;s depreciation burden.&lt;/p&gt;
&lt;hr&gt;

 &lt;blockquote&gt;
 &lt;p&gt;This article is based on public company investor-relations materials, SEC filings, and industry sources. Backlogs and long-term agreements improve visibility but remain subject to usage, recognition timing, pricing, and contract terms. This is informational analysis, not a recommendation to buy or sell securities.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>NVIDIA Vera CPU Production Expansion and Simmtech: SoCAMM2 is About Module Units, Not Bits</title><link>https://koreainvestinsights.com/post/nvidia-vera-cpu-socamm2-simmtech-module-unit-thesis-2026-07-19/</link><pubDate>Sun, 19 Jul 2026 18:05:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/nvidia-vera-cpu-socamm2-simmtech-module-unit-thesis-2026-07-19/</guid><description>&lt;p&gt;One NVIDIA Vera Rubin NVL72 contains 36 Vera CPUs and 54TB of LPDDR5X memory. That is 1.5TB per CPU. However, in June, TrendForce reported that NVIDIA would halve the capacity of its next-generation Vera SoCAMM modules while increasing total module shipments and Vera CPU production.&lt;/p&gt;
&lt;p&gt;On the surface, reduced memory capacity appears negative for component suppliers. When measuring PCB demand purely in terms of bits, it is. However, what Simmtech manufactures is not DRAM chips but PCBs for SoCAMM modules. Simmtech&amp;rsquo;s volume is more sensitive to the number of CPUs produced and the number of modules than to total memory capacity. Even if LPDDR shortage forces lower per-CPU capacity, if NVIDIA ships more Vera CPUs and modules, PCB demand may not decline.&lt;/p&gt;
&lt;p&gt;Looking at official materials, industry research, analyst estimates, and actual stock prices and supply-demand data shows that the business logic remains intact, but the price bottom has not yet been confirmed.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Connected Context
This article is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/ai-pcb-thesis-system-bom-common-bottleneck-2026-05-05/" &gt;AI substrate and PCB investment thesis: System BOM as a common bottleneck&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-ai-pcb-ecosystem-ten-companies-2026-05-05/" &gt;Korean AI PCB ecosystem: ten companies&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/vera-rubin-vr200-bom-memory-pcb-mlcc-korea-alpha-2026-05-21/" &gt;Vera Rubin VR200 bill of materials verification&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/memory-supply-gap-spillover-substrate-essd-legacy-hana-three-reports-2026-07-14/" &gt;Memory supply gap spilling beyond HBM&lt;/a&gt;. Related articles can be found in the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-ai-pcb-substrate-hub/" &gt;AI substrate and PCB hub&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM hub&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;exclusive analysis hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;NVIDIA&amp;rsquo;s official specifications show Vera Rubin NVL72 uses 36 Vera CPUs and 54TB of LPDDR5X. Standalone Vera CPU racks support up to 4 nodes per 1U tray, expanding to 256 CPUs per rack maximum. Vera&amp;rsquo;s evolution from a supporting CPU in GPU racks to an independent server platform is confirmed.&lt;/li&gt;
&lt;li&gt;TrendForce&amp;rsquo;s reported SoCAMM capacity reduction is a response to LPDDR supply shortage, not slowing demand. NVIDIA intends to lower per-module capacity while increasing total module shipments and Vera CPU production.&lt;/li&gt;
&lt;li&gt;Simmtech lists SoCAMM PCB on its official product page. However, simultaneous supply from three memory vendors and high market share are analyst estimates, not company disclosures. Actual customer-specific revenue and market share remain undisclosed.&lt;/li&gt;
&lt;li&gt;On July 16th, Simmtech closed at 107,400 KRW, down 11.68% in a single day. Over 20 trading days, institutions bought 236.2 billion KRW in net, and real money (pension, mutual funds, insurance) bought 186.4 billion KRW, while foreign investors sold 161.1 billion KRW and programs sold 185.9 billion KRW. Mid-term institutional buying pressure collides with short-term selling pressure.&lt;/li&gt;
&lt;li&gt;The 2026E PER of 27.1x is not cheap. Applying 2027E expected EPS of 6,655 KRW yields 16.1x, but only if SoCAMM revenue and profit margins expand as planned.&lt;/li&gt;
&lt;li&gt;The verdict is &lt;strong&gt;Conditional Buy&lt;/strong&gt;. A safer approach confirms support near 104,000 KRW with easing foreign and program selling, or recovery through 113,800 KRW on increased volume.&lt;/li&gt;
&lt;/ul&gt;
&lt;hr&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;This Article's Verdict&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Directly connecting SoCAMM2 capacity reduction to Simmtech PCB demand decline uses the wrong unit. Simmtech's revenue formula approaches Vera CPU units × modules per CPU × PCBs per module × Simmtech market share more than DRAM bit counts. However, the last two items in this formula have not yet been confirmed by company disclosures.
 &lt;/div&gt;
&lt;/div&gt;
&lt;hr&gt;
&lt;h2 id="1-numbers-to-discard-and-numbers-to-keep"&gt;1. Numbers to Discard and Numbers to Keep
&lt;/h2&gt;&lt;p&gt;The 2030 server CPU market size of 170 billion dollars, 1:1 CPU-to-GPU ratio, and 2027 DRAM shortage appearing in attached analysis are already reflected in major statements from multiple brokerages and corporate presentations. They are useful for understanding industry direction but are not sources of excess returns specific to Simmtech.&lt;/p&gt;
&lt;p&gt;Four numbers are more important when evaluating Simmtech:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Why It Matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Vera CPUs shipped&lt;/td&gt;
 &lt;td&gt;Determines the installed base for the SoCAMM platform&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SoCAMM modules per CPU&lt;/td&gt;
 &lt;td&gt;Direct volume variable for Simmtech module PCBs&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Simmtech&amp;rsquo;s customer-specific supply share&lt;/td&gt;
 &lt;td&gt;Determines Simmtech&amp;rsquo;s portion of industry growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Module PCB unit price and cost&lt;/td&gt;
 &lt;td&gt;Determines the rate at which revenue growth converts to operating profit&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Using assumptions of 5 to 6 million Vera CPUs and 8 modules per CPU yields 40 to 48 million modules annually. At 5.5 million units, that is 44 million modules. However, this calculation is not an official NVIDIA shipment plan. It is a scenario derived from analyst and industry CPU shipment estimates and publicly disclosed system configurations.&lt;/p&gt;
&lt;p&gt;Therefore, the 44 million unit figure should not be used directly as a revenue projection. It is only an example for gauging order magnitude.&lt;/p&gt;
&lt;h2 id="2-official-nvidia-materials-confirm-veras-expansion-path"&gt;2. Official NVIDIA Materials Confirm Vera&amp;rsquo;s Expansion Path
&lt;/h2&gt;&lt;h3 id="21-nvl72-contains-36-vera-cpus"&gt;2.1 NVL72 Contains 36 Vera CPUs
&lt;/h3&gt;&lt;p&gt;NVIDIA&amp;rsquo;s &lt;a class="link" href="https://www.nvidia.com/en-us/data-center/vera-rubin-nvl72/" target="_blank" rel="noopener"
 &gt;official Vera Rubin NVL72 page&lt;/a&gt; presents the following configuration:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Official Specification&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Rubin GPUs&lt;/td&gt;
 &lt;td style="text-align: right"&gt;72&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Vera CPUs&lt;/td&gt;
 &lt;td style="text-align: right"&gt;36&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CPU Memory&lt;/td&gt;
 &lt;td style="text-align: right"&gt;LPDDR5X 54TB&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory per Vera CPU&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.5TB&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Vera CPU Cores&lt;/td&gt;
 &lt;td style="text-align: right"&gt;88&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The structure is not simply one CPU per GPU. In NVL72, one Vera CPU supports two GPUs. Nevertheless, a single 72-GPU rack requires 36 CPUs. This shows that the CPU and LPDDR layers have risen to major components in AI system cost structure.&lt;/p&gt;
&lt;h3 id="22-vera-sells-beyond-gpu-racks"&gt;2.2 Vera Sells Beyond GPU Racks
&lt;/h3&gt;&lt;p&gt;The more significant change is the standalone Vera CPU system. &lt;a class="link" href="https://developer.nvidia.com/blog/?p=114004" target="_blank" rel="noopener"
 &gt;NVIDIA&amp;rsquo;s technical blog&lt;/a&gt; explains that Vera CPU racks can hold up to 4 nodes per 1U tray and scale to 256 Vera CPUs per rack. Single-socket and dual-socket systems support up to 1.5TB LPDDR5X per socket.&lt;/p&gt;
&lt;p&gt;NVIDIA announced that Cisco, Dell, HPE, Lenovo, and Supermicro plan to launch Vera-based systems in the second half of 2026. If the schedule holds, Vera demand sources extend beyond NVL72 assembly volumes.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Vera Demand Sources
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;CPU layer in Vera Rubin NVL72
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ Standalone Vera CPU servers
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ CPU racks for inference and agents
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ Single and dual-socket systems from major OEMs
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;This is why Simmtech&amp;rsquo;s investment logic does not depend on GPU rack production alone. Expansion of standalone CPU systems broadens the demand base for SoCAMM modules and PCBs.&lt;/p&gt;
&lt;h2 id="3-what-does-socamm-capacity-reduction-mean"&gt;3. What Does SoCAMM Capacity Reduction Mean?
&lt;/h2&gt;&lt;p&gt;&lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260610-13090.html" target="_blank" rel="noopener"
 &gt;TrendForce&amp;rsquo;s June 10, 2026 analysis&lt;/a&gt; found that only about 60% of NVIDIA&amp;rsquo;s 2027 LPDRAM requirements could be met with reserved allocations. To address the supply shortage, NVIDIA is reported to have chosen the following adjustments:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Lower the memory capacity of each SoCAMM module for Vera.&lt;/li&gt;
&lt;li&gt;Increase total module shipments.&lt;/li&gt;
&lt;li&gt;Produce more Vera CPUs with the same LPDRAM supply.&lt;/li&gt;
&lt;li&gt;Raise market penetration of standalone Vera CPU racks.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;In this structure, two types of demand can move in different directions.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Demand Unit&lt;/th&gt;
 &lt;th&gt;Impact of Capacity Reduction&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM bits and GB&lt;/td&gt;
 &lt;td&gt;Per-CPU installed capacity may decline&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SoCAMM modules&lt;/td&gt;
 &lt;td&gt;May increase if CPU shipment growth exceeds capacity reduction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Module PCBs&lt;/td&gt;
 &lt;td&gt;Proportional to modules, so may increase&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sockets, connectors, module testing&lt;/td&gt;
 &lt;td&gt;Increase with module count&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Expressing this in Simmtech terms gives the following formula:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Simmtech SoCAMM PCB Volume
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;= Vera CPU Shipment
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;× SoCAMM modules per CPU
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;× PCBs per module
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;× Simmtech supply share
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Conversely, memory vendors&amp;rsquo; SoCAMM revenue approximates CPU count × per-CPU installed GB × price per GB. The same news can have different impacts on DRAM vendors and module PCB vendors.&lt;/p&gt;
&lt;h3 id="mismatch-between-official-specifications-and-trendforce-reporting"&gt;Mismatch Between Official Specifications and TrendForce Reporting
&lt;/h3&gt;&lt;p&gt;NVIDIA&amp;rsquo;s current official page shows a maximum 1.5TB LPDDR5X per Vera CPU. TrendForce reported next-generation module capacity reduction. With both documents existing simultaneously, it is not yet certain which case applies:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;NVIDIA&amp;rsquo;s page continues to display maximum configuration or existing design&lt;/li&gt;
&lt;li&gt;Capacity reduction applies only to specific customers or product configurations&lt;/li&gt;
&lt;li&gt;Final production specifications are not yet finalized&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;This inconsistency is not noise to be hidden but a verification item for investment judgment. The actual per-module capacity, modules per CPU, and customer-specific configuration of production systems must be verified through launched products and company disclosures.&lt;/p&gt;
&lt;h2 id="4-what-does-simmtech-supply-in-socamm2"&gt;4. What Does Simmtech Supply in SoCAMM2?
&lt;/h2&gt;&lt;h3 id="41-what-the-company-officially-confirms"&gt;4.1 What the Company Officially Confirms
&lt;/h3&gt;&lt;p&gt;&lt;a class="link" href="https://www.simmtech.com/product/board06.aspx" target="_blank" rel="noopener"
 &gt;Simmtech&amp;rsquo;s official product page&lt;/a&gt; classifies SoCAMM PCB as a next-generation memory module product for AI. The key specifications presented by the company are as follows:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Simmtech Official Description&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Base Memory&lt;/td&gt;
 &lt;td&gt;LPDDR&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Application&lt;/td&gt;
 &lt;td&gt;Next-generation memory modules for AI&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Layer Count&lt;/td&gt;
 &lt;td&gt;10–12 layers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Key Requirements&lt;/td&gt;
 &lt;td&gt;Low dielectric constant and loss, low thermal expansion coefficient, impedance control&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Surface Treatment&lt;/td&gt;
 &lt;td&gt;Selective ENIG&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Confirmation that SoCAMM PCB appears on Simmtech&amp;rsquo;s product list is established. The company also manufactures general memory module PCBs and memory package substrates. When AI server LPDDR demand grows in both modules and packages, the company has a product foundation to respond.&lt;/p&gt;
&lt;h3 id="42-what-is-only-confirmed-by-analyst-estimates"&gt;4.2 What Is Only Confirmed by Analyst Estimates
&lt;/h3&gt;&lt;p&gt;The claim that Simmtech participates in SoCAMM2 production from Samsung Electronics, SK Hynix, and Micron simultaneously and holds high market share is not a company disclosure. It is an analysis and market estimate in &lt;a class="link" href="https://www.simmtech.com/upload/media/file/639179687810021917.pdf" target="_blank" rel="noopener"
 &gt;the brokerage report posted on Simmtech&amp;rsquo;s homepage&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;Therefore, these expressions must be distinguished:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Expression&lt;/th&gt;
 &lt;th&gt;Evidence Level&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Simmtech possesses SoCAMM PCB products&lt;/td&gt;
 &lt;td&gt;Company official confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Simmtech participates in SoCAMM2 production from three memory vendors&lt;/td&gt;
 &lt;td&gt;Analyst estimate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Simmtech holds #1 supply share&lt;/td&gt;
 &lt;td&gt;Analyst estimate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer-specific revenue and actual market share&lt;/td&gt;
 &lt;td&gt;Undisclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Exact timing of SoCAMM contribution to Simmtech profit&lt;/td&gt;
 &lt;td&gt;Undisclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This distinction is necessary because customer qualification and production volume drive the economics of the substrate business. The ability to manufacture a product and maintaining high market share for repeat revenue are separated by stages of customer qualification, yield, unit price, and utilization rate.&lt;/p&gt;
&lt;h3 id="43-sk-hynixs-mass-production-confirms-platform-reality"&gt;4.3 SK Hynix&amp;rsquo;s Mass Production Confirms Platform Reality
&lt;/h3&gt;&lt;p&gt;According to &lt;a class="link" href="https://news.skhynix.com/mass-production-socamm2-192gb/" target="_blank" rel="noopener"
 &gt;SK Hynix&amp;rsquo;s official announcement&lt;/a&gt;, the company began mass production of 192GB SoCAMM2 using 1cnm LPDDR5X in April 2026. The product targets the NVIDIA Vera Rubin platform.&lt;/p&gt;
&lt;p&gt;This shows that SoCAMM2 is not a long-term research project but has entered actual mass production. However, SK Hynix&amp;rsquo;s module mass production does not automatically prove Simmtech&amp;rsquo;s customer-specific supply volumes. Simmtech demonstrating SoCAMM revenue, high-layer board mix, and utilization rate improvements in second and third quarter results is separately necessary.&lt;/p&gt;
&lt;h2 id="5-results-are-recovering-but-still-in-the-stage-of-proving-expectations"&gt;5. Results Are Recovering but Still in the Stage of Proving Expectations
&lt;/h2&gt;&lt;p&gt;Simmtech recorded 1.41 trillion KRW in full-year 2025 revenue and 11.9 billion KRW in operating profit, with a net loss of 164.2 billion KRW. In the first quarter 2026, consolidated revenue recovered to approximately 422.4 billion KRW and operating profit to approximately 13.7 billion KRW. Second quarter consensus expects faster recovery pace.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26 Actual&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q26 Consensus&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Sequential Change&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~422.4 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;466.9 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~+10.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating Profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~13.7 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;47.3 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~+245%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating Margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~3.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~10.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.9 percentage points&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;First quarter actuals are from &lt;a class="link" href="https://kind.krx.co.kr/external/2026/05/14/000890/20260514001959/11013.htm" target="_blank" rel="noopener"
 &gt;Simmtech&amp;rsquo;s quarterly report&lt;/a&gt;; second quarter estimates are from Naver and WiseReport consensus collected by Research OS local database on July 16th.&lt;/p&gt;
&lt;p&gt;Operating profit growth far exceeds revenue growth. This indicates the need for simultaneous increases in high-margin product mix, factory utilization, and fixed cost absorption. Even if SoCAMM volume increases, profit margin may fall below expectations due to raw material and gold plating costs, initial yield, and customer-specific price negotiations.&lt;/p&gt;
&lt;h3 id="the-profit-path-demanded-by-annual-consensus"&gt;The Profit Path Demanded by Annual Consensus
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Year&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Operating Profit&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Net Income&lt;/th&gt;
 &lt;th style="text-align: right"&gt;EPS&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026E PER (July 16 close)&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.932 trillion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;189.6 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;149.3 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,968 KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;27.1x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.289 trillion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Consensus not aggregated&lt;/td&gt;
 &lt;td style="text-align: right"&gt;251.0 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6,655 KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.1x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.566 trillion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Consensus not aggregated&lt;/td&gt;
 &lt;td style="text-align: right"&gt;322.8 billion KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8,588 KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12.5x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For the current stock price to look cheap, 2027 and 2028 earnings must actually materialize. The current 2026E PER of 27.1x and PBR of 5.64x substantially reflects early recovery. An average price target of 171,250 KRW is 59.5% higher than current price, but when targets are rising rapidly, they are difficult to use as a safety margin.&lt;/p&gt;
&lt;h2 id="6-july-16th-stock-price-overheating-reduced-but-trend-recovery-not-yet-confirmed"&gt;6. July 16th Stock Price: Overheating Reduced but Trend Recovery Not Yet Confirmed
&lt;/h2&gt;&lt;p&gt;Simmtech closed at 107,400 KRW on July 16th, declining 11.68% that day with an intraday low of 103,700 KRW. This is 34.6% lower than the July 1st high of 164,200 KRW.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period and Indicator&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1-day return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.68%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5-day&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.91%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;10-day&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.87%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20-day&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-15.76%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;60-day&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+40.94%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;120-day&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+109.77%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;vs. 5-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.42%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;vs. 20-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.28%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;vs. 60-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.48%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;vs. 120-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27.39%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RSI14&lt;/td&gt;
 &lt;td style="text-align: right"&gt;44.4&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;RSI has exited overbought territory. However, price remains below 5-day, 20-day, and 60-day moving averages. Since it is 27% above the 120-day line, all upside from the long-term move has not disappeared. Both &amp;ldquo;a sharp decline means cheap&amp;rdquo; and &amp;ldquo;the trend has recovered&amp;rdquo; cannot yet be confirmed.&lt;/p&gt;
&lt;p&gt;On a trading basis, July 14th low of 99,600 KRW and July 16th low of 103,700 KRW represent the first support zone. The 113,800 KRW level confirms whether broken short-term support recovers; 121,600 to 124,000 KRW is a resistance zone where July 15th close and the 20-day moving average overlap.&lt;/p&gt;
&lt;h2 id="7-supply-and-demand-institutions-bought-but-foreign-investors-and-programs-still-sell"&gt;7. Supply and Demand: Institutions Bought But Foreign Investors and Programs Still Sell
&lt;/h2&gt;&lt;p&gt;Cumulative net buying amounts through July 16th from Research OS local Kiwoom database are as follows. Units are in billions of KRW.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Institutions&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Individuals&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Real Money&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Programs&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1 day&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.83&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.88&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.57&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.91&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.55&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-25.79&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.02&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.95&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.98&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-28.71&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;10 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-93.39&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+80.32&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.79&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+75.77&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-40.78&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-161.09&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+236.21&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-84.25&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+186.44&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-185.94&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Real money is the sum of pension funds, mutual funds, and insurance. The table reveals two things simultaneously.&lt;/p&gt;
&lt;p&gt;First, on a 20-day basis, institutional and real money buying is substantial. During the post-peak correction, domestically oriented long-term capital absorbed volume.&lt;/p&gt;
&lt;p&gt;Second, foreign investor and program selling has continued at similar scale. On the July 16th plunge, foreign investors and institutions sold together while individuals absorbed 15.57 billion KRW. It is difficult to judge the bottom confirmed by institutional buying alone.&lt;/p&gt;
&lt;p&gt;Short-selling transaction share also rose to 21.2% on July 16th. The 20-day average is 9.3%. Short-rebate inventory is approximately 5.48 million shares or 14.6% of issued shares, but not all short-rebate inventory is short-sale inventory. This figure should only be consulted as a reference for available selling volume.&lt;/p&gt;
&lt;h2 id="8-simmtech-revisited-through-p--q--c"&gt;8. Simmtech Revisited Through P × Q × C
&lt;/h2&gt;&lt;h3 id="p-unit-price-of-high-layer-socamm-pcb"&gt;P: Unit Price of High-Layer SoCAMM PCB
&lt;/h3&gt;&lt;p&gt;SoCAMM PCB requires 10–12 layers, low dielectric loss, low thermal expansion, and precise impedance control. Higher manufacturing difficulty compared to general memory module PCB can help improve product mix. However, customer-specific unit prices and price increase structures are undisclosed.&lt;/p&gt;
&lt;h3 id="q-vera-cpu-and-socamm-module-count"&gt;Q: Vera CPU and SoCAMM Module Count
&lt;/h3&gt;&lt;p&gt;Q is the largest upside variable. Beyond NVL72 mass production, standalone Vera CPU racks and major OEM servers are scheduled for the second half of 2026. If NVIDIA follows TrendForce&amp;rsquo;s report in lowering per-module capacity while increasing module shipments, Simmtech&amp;rsquo;s Q can grow faster than DRAM bit count.&lt;/p&gt;
&lt;h3 id="c-raw-materials-gold-plating-yield-utilization"&gt;C: Raw Materials, Gold Plating, Yield, Utilization
&lt;/h3&gt;&lt;p&gt;The item determining whether revenue growth converts to operating profit. PCBs are sensitive to gold and copper prices, BT materials, plating costs, and initial yield. To achieve 10.1% operating margin consensus for Q2, increased high-margin products and cost pass-through must be confirmed together.&lt;/p&gt;
&lt;h2 id="9-three-scenarios"&gt;9. Three Scenarios
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;Conditions&lt;/th&gt;
 &lt;th&gt;Earnings and Price Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bullish&lt;/td&gt;
 &lt;td&gt;Vera Rubin and standalone CPU servers ship as scheduled; total SoCAMM module count increases; Simmtech maintains three-memory-vendor market share&lt;/td&gt;
 &lt;td&gt;2027E EPS of 6,655 KRW or higher becomes achievable and justifies leading PER of ~16x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base&lt;/td&gt;
 &lt;td&gt;Vera shipments grow but LPDDR shortage and customer-specific ramps misalign; Simmtech share and margin remain at consensus levels&lt;/td&gt;
 &lt;td&gt;Business growth remains valid but stock can fluctuate widely based on earnings and supply-demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bearish&lt;/td&gt;
 &lt;td&gt;Vera or SoCAMM mass production delays; reduced per-module capacity does not lead to increased module count; Simmtech share or yield underperforms&lt;/td&gt;
 &lt;td&gt;27x 2026E PER becomes a burden and probability of revisiting July lows increases&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;In these scenarios, the most important variable is not the 170 billion dollar server CPU market. It is Simmtech&amp;rsquo;s actual shipment volume, customer-specific revenue, high-layer board mix, and operating margin. Even if large industry numbers prove correct, if the portion accruing to the company is small, the stock will struggle.&lt;/p&gt;
&lt;h2 id="10-investment-verdict-and-verification-sequence"&gt;10. Investment Verdict and Verification Sequence
&lt;/h2&gt;&lt;h3 id="verdict-conditional-buy"&gt;Verdict: Conditional Buy
&lt;/h3&gt;&lt;p&gt;Business logic has become clearer. NVIDIA official materials confirmed Vera&amp;rsquo;s standalone server expansion, and TrendForce&amp;rsquo;s report shows capacity reduction as a supply shortage response, not demand decline. Simmtech officially possesses SoCAMM PCB products.&lt;/p&gt;
&lt;p&gt;However, the stock remains under pressure from foreign investor and program selling, and it is not cheap on 2026 earnings basis. Therefore, it is better to consider two entry paths.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Path&lt;/th&gt;
 &lt;th&gt;Verification Condition&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Support Confirmation&lt;/td&gt;
 &lt;td&gt;Hold the low between 103,700–106,000 KRW and confirm easing of foreign investor and program selling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Trend Confirmation&lt;/td&gt;
 &lt;td&gt;Recover through 113,800 KRW on increased volume with either institutions or foreign investors turning net buyers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Recovery through 121,600–124,000 KRW would substantially reverse near-term trend damage. Conversely, breaking below 99,600 KRW on a closing basis without recovery would require reassessing price hypothesis.&lt;/p&gt;
&lt;h3 id="catalysts"&gt;Catalysts
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Achievement of 466.9 billion KRW revenue and 47.3 billion KRW operating profit in 2Q26 earnings&lt;/li&gt;
&lt;li&gt;Company commentary on SoCAMM revenue, high-layer board mix, and utilization rate&lt;/li&gt;
&lt;li&gt;Actual market launch of Vera-based OEM servers in second half 2026&lt;/li&gt;
&lt;li&gt;NVIDIA&amp;rsquo;s final production specifications and expansion of standalone Vera CPU rack shipments&lt;/li&gt;
&lt;li&gt;Mass production and customer qualification updates from SK Hynix, Samsung Electronics, and Micron for SoCAMM2&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="invalidation-conditions"&gt;Invalidation Conditions
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Repeat orders fail to follow initial SoCAMM volume&lt;/li&gt;
&lt;li&gt;Reduced per-module capacity does not translate to increased total module shipments&lt;/li&gt;
&lt;li&gt;Simmtech&amp;rsquo;s customer-specific market share confirms below analyst estimates&lt;/li&gt;
&lt;li&gt;Q2 operating profit materially misses consensus with recurring raw material and yield burdens&lt;/li&gt;
&lt;li&gt;Vera Rubin or standalone Vera CPU system launch delays significantly&lt;/li&gt;
&lt;li&gt;99,600 KRW breaks below support with concurrent expansion of foreign investor and program selling&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="11-final-conclusion"&gt;11. Final Conclusion
&lt;/h2&gt;&lt;p&gt;Vera CPU supply expansion reinforces Simmtech&amp;rsquo;s SoCAMM2 investment thesis. Particularly, the path of lowering module capacity due to LPDDR shortage while increasing total module and CPU shipments is difficult to characterize as unfavorable for substrate vendors. DRAM vendors sell bits and GB; Simmtech sells PCBs that go into each module. These are different units.&lt;/p&gt;
&lt;p&gt;However, many items remain unconfirmed. Vera shipment volumes, final modules per CPU, Simmtech&amp;rsquo;s actual customer-specific market share, SoCAMM revenue, and high-layer board profit margin cannot be finalized by public materials alone. A calculation like 44 million modules is a useful scenario but not official guidance.&lt;/p&gt;
&lt;p&gt;Current stock price has corrected nearly 35% from peak, but there is no evidence that foreign investor and program selling has ended. Therefore, this idea should be viewed as &amp;ldquo;a conditional buy candidate requiring confirmation of both earnings and supply-demand&amp;rdquo; rather than &amp;ldquo;a growth stock that has fallen sharply.&amp;rdquo; Q2 operating profit and SoCAMM revenue commentary is the first verification, and second-half 2026 Vera standalone server shipments represent the second verification.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="evidence-and-limitations"&gt;Evidence and Limitations
&lt;/h2&gt;&lt;h3 id="confirmed-facts"&gt;Confirmed Facts
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;NVIDIA official specifications present 36 Vera CPUs and 54TB LPDDR5X in NVL72.&lt;/li&gt;
&lt;li&gt;NVIDIA disclosed standalone CPU systems accommodating up to 256 Vera CPUs per rack and OEM launch plans for second half 2026.&lt;/li&gt;
&lt;li&gt;TrendForce reported adjustments to lower per-module capacity while increasing total module shipments and Vera CPU production due to LPDRAM shortage.&lt;/li&gt;
&lt;li&gt;SK Hynix announced mass production of 192GB SoCAMM2 for Vera Rubin.&lt;/li&gt;
&lt;li&gt;Simmtech lists SoCAMM PCB on its official product list.&lt;/li&gt;
&lt;li&gt;July 16th Simmtech close, supply-demand, and consensus have been reverified from local database.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="estimates"&gt;Estimates
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Annual calculation of 40 to 48 million modules using assumptions of 5 to 6 million Vera CPUs and 8 modules per CPU&lt;/li&gt;
&lt;li&gt;Claims that Simmtech supplies simultaneously to three memory vendors and maintains high market share&lt;/li&gt;
&lt;li&gt;Magnitude of capacity reduction impact on Simmtech revenue and profit following module count increase&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="items-not-confirmed-by-public-materials"&gt;Items Not Confirmed by Public Materials
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;NVIDIA&amp;rsquo;s official 2027 Vera CPU shipment guidance&lt;/li&gt;
&lt;li&gt;Final production system modules per CPU and capacity per module&lt;/li&gt;
&lt;li&gt;Simmtech&amp;rsquo;s customer-specific SoCAMM revenue, market share, unit price, and profit margin&lt;/li&gt;
&lt;li&gt;Q2 earnings announcement date and company SoCAMM revenue guidance&lt;/li&gt;
&lt;li&gt;Precise scope of application for TrendForce&amp;rsquo;s reported capacity reduction and NVIDIA&amp;rsquo;s official 1.5TB specification&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Informational analysis based on public materials and local market data; not investment advice reflecting individual investor timelines, prices, or risk tolerance.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>Weekly Semiconductor Deep Dive: Strong Earnings, Compressed Multiples, and the 2028 Inflection</title><link>https://koreainvestinsights.com/post/weekly-semiconductor-deep-dive-strong-earnings-lower-multiple-2028-pivot-2026-07-19/</link><pubDate>Sun, 19 Jul 2026 15:53:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/weekly-semiconductor-deep-dive-strong-earnings-lower-multiple-2028-pivot-2026-07-19/</guid><description>&lt;p&gt;On July 13, KOSPI fell 9.05% and SK Hynix dropped 15.37%. Yet in the same week, TSMC reported Q2 revenue of $40.2 billion and a gross margin of 67.7%, while ASML raised its annual revenue outlook to €43–45 billion. Commodity DRAM and NAND contract prices also rose sharply. It was a week in which physical orders and stock prices moved in opposite directions.&lt;/p&gt;
&lt;p&gt;This divergence is less a signal that the semiconductor boom is over, and more a sign that the market&amp;rsquo;s question has changed. Investors are no longer looking only at how much will be earned in 2026 and 2027. They are first calculating how much supply will expand in 2028, whether customers can sustain high memory prices and data center capex, and how much inference efficiency gains will erode the scarcity premium of HBM.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Connected Context
This post consolidates &lt;a class="link" href="https://koreainvestinsights.com/post/semiconductor-bull-bear-four-clocks-capital-intensity-cycle-2026-07-17/" &gt;Five Clocks of the Semiconductor Bull and Bear Case&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/memory-fair-value-fcfe-terminal-samsung-hynix-micron-2026-07-17/" &gt;Memory Fair Value via FCFE and Normalized Earnings&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-sk-hynix-2027-2028-integrated-scenarios-risk-adjusted-valuation-2026-07-13/" &gt;Samsung Electronics and SK Hynix: 2027–2028 Scenarios&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/hbm-2030-supply-demand-267eb-demand-model-crosscheck-2026-07-13/" &gt;HBM 2030 Supply-Demand Model Cross-Check (26.7 EB)&lt;/a&gt; into a weekly synthesis updated with new data from the third week of July. Related posts are available at the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Official data confirmed from July 13–19 does not support the claim that AI semiconductor demand has peaked. TSMC and ASML results and guidance, DRAM and NAND prices, and server memory demand all reinforce the earnings-strength thesis for 2026–2027.&lt;/li&gt;
&lt;li&gt;The same data also amplifies 2028 supply risk. ASML&amp;rsquo;s capacity expansion plans and TSMC&amp;rsquo;s heavy capex are evidence of current orders — and simultaneously a reservation of future supply and depreciation.&lt;/li&gt;
&lt;li&gt;HBM long-term contracts can raise the floor on volume and earnings, but may slow the pace at which spot price spikes translate into blended ASPs. Volume stability and a cap on price upside coexist.&lt;/li&gt;
&lt;li&gt;China plays three roles simultaneously: an end-demand destination, a competitor in commodity memory, and a market that could be politically decoupled. Its larger impact is on 2028 commodity DRAM and NAND price ceilings and global supply discipline, not on near-term HBM substitution.&lt;/li&gt;
&lt;li&gt;Efficiency gains from open models such as Kimi K3 do not simply eliminate semiconductor demand. They can shift the value of some HBM toward server DRAM, eSSD, networking, and power. The key question is not whether cost-per-task falls, but how much total workload expands and what share is deployed on self-built infrastructure.&lt;/li&gt;
&lt;li&gt;Current base-case scenario probabilities: Extended Scarcity 25%, Strong Earnings / Compressed Multiple 40%, Efficiency and Supply Normalization 25%, Financial / Demand Dislocation 10%. The most probable path is one where strong earnings in 2026–2027 and duration discounting for 2028 coexist.&lt;/li&gt;
&lt;/ul&gt;
&lt;hr&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Key Statement of the Week&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The physical evidence underpinning the memory supercycle passed its first full-scale stress test. Stock prices, however, have begun asking not about the magnitude of scarcity but about how long excess earnings will last. 2026–2027 is the period for earnings delivery; 2028 is the period that simultaneously tests supply, efficiency, and customer profitability.
 &lt;/div&gt;
&lt;/div&gt;
&lt;hr&gt;
&lt;h2 id="1-scope-and-evidence-rules"&gt;1. Scope and Evidence Rules
&lt;/h2&gt;&lt;p&gt;The analysis covers July 13–19, 2026. Company IR materials, government and congressional filings, industry data, broker estimates, and market supply-demand figures drawn from two weekly analysis documents were re-categorized. Where multiple posts cited the same underlying source, the confirmation count was not incremented.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Grade&lt;/th&gt;
 &lt;th&gt;Source&lt;/th&gt;
 &lt;th&gt;How Used in This Report&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;A&lt;/td&gt;
 &lt;td&gt;Company IR, government / congressional / regulatory filings, SEC disclosures&lt;/td&gt;
 &lt;td&gt;Baseline for facts and official plans&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;B&lt;/td&gt;
 &lt;td&gt;Industry research such as TrendForce, primary broker reports&lt;/td&gt;
 &lt;td&gt;Supporting evidence for pricing, share, and company-level estimates&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C&lt;/td&gt;
 &lt;td&gt;Cross-referenced press reports, channel check summaries&lt;/td&gt;
 &lt;td&gt;Used for directional confirmation only; not used as standalone investment signals&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;D&lt;/td&gt;
 &lt;td&gt;Social media, circulating summary images&lt;/td&gt;
 &lt;td&gt;Used only to identify questions requiring further verification&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Orders from TSMC, ASML, and memory manufacturers can reflect the same hyperscaler capex plan multiple times. Official plans signal demand intent but do not guarantee actual utilization or cash flow. The sharp stock price declines were not attributed solely to technical supply-demand dynamics. Flow imbalances can amplify drawdowns, but they cannot eliminate a fundamental re-rating of earnings duration.&lt;/p&gt;
&lt;h2 id="2-a-map-of-the-weeks-events"&gt;2. A Map of the Week&amp;rsquo;s Events
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Date&lt;/th&gt;
 &lt;th&gt;Confirmed Developments&lt;/th&gt;
 &lt;th&gt;Questions the Market Began Asking&lt;/th&gt;
 &lt;th&gt;Current Assessment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;July 13&lt;/td&gt;
 &lt;td&gt;HBM 2030 demand model at 26.7 EB re-verified; SK Hynix Q2 earnings estimate cut; KOSPI sharp decline&lt;/td&gt;
 &lt;td&gt;Long-term shortage, or excessive extrapolation?&lt;/td&gt;
 &lt;td&gt;2027 shortage has strong support; 2030 shortage magnitude carries low conviction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 14&lt;/td&gt;
 &lt;td&gt;SK Hynix estimates adjusted to reflect HBM long-term contracts; IBM and Ericsson flagging memory cost burden; reports of Apple testing CXMT chips&lt;/td&gt;
 &lt;td&gt;Are long-term contracts an earnings floor or a price ceiling?&lt;/td&gt;
 &lt;td&gt;Volume visibility is high, but ASP upside constraints and customer cost burden emerge alongside&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 15&lt;/td&gt;
 &lt;td&gt;ASML Q2 results and 2027–2028 capacity expansion plans&lt;/td&gt;
 &lt;td&gt;Strong orders, or the start of a supply response?&lt;/td&gt;
 &lt;td&gt;Both are correct — simultaneously reinforces current demand and amplifies 2028 supply risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 16&lt;/td&gt;
 &lt;td&gt;TSMC Q2 results and Q3 guidance; U.S. House letter requesting restrictions on Chinese memory procurement&lt;/td&gt;
 &lt;td&gt;AI demand is strong — why is memory stock price weak?&lt;/td&gt;
 &lt;td&gt;Debate centers on earnings attribution, supply, and cost of capital — not demand collapse&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 17&lt;/td&gt;
 &lt;td&gt;Kimi K3 launch; Chinese open-model efficiency gains and API pricing repricing&lt;/td&gt;
 &lt;td&gt;Does efficiency reduce semiconductor demand?&lt;/td&gt;
 &lt;td&gt;Cost-per-task falls, but aggregate usage and self-built deployments may expand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 18&lt;/td&gt;
 &lt;td&gt;China AI API cost analysis; review of memory trio&amp;rsquo;s China exposure&lt;/td&gt;
 &lt;td&gt;Is China a customer or a competitor?&lt;/td&gt;
 &lt;td&gt;Both roles simultaneously; impacts 2028 normalized value more than 2026 EPS&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 19&lt;/td&gt;
 &lt;td&gt;Channel checks circulating claiming strong server DRAM and HBM4 orders&lt;/td&gt;
 &lt;td&gt;Did the sell-off price in a Q3 earnings collapse?&lt;/td&gt;
 &lt;td&gt;Probability appears lower, but primary source unconfirmed — maintained at Grade C&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The most significant shift in this table is not the direction of individual events, but the time horizon. Until early July, the debate centered on whether H2 2026 earnings would beat expectations. It has since shifted to whether 2028 earnings represent a normalized earnings level, and how long elevated prices and margins can be sustained.&lt;/p&gt;
&lt;h2 id="3-evidence-strengthened-this-week"&gt;3. Evidence Strengthened This Week
&lt;/h2&gt;&lt;h3 id="31-tsmc-and-asml-showed-that-ai-orders-have-translated-into-real-activity"&gt;3.1 TSMC and ASML Showed That AI Orders Have Translated into Real Activity
&lt;/h3&gt;&lt;p&gt;TSMC&amp;rsquo;s Q2 results demonstrate that AI demand has moved beyond the planning stage.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;TSMC Q2 2026 and Q3 2026 Outlook&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$40.2 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;67.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 operating margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;60.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 revenue guidance&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$44.6–$45.8 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 gross margin guidance&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65–67%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The high gross margins and next-quarter growth guidance indicate that the quality of orders for leading-edge process nodes and advanced packaging remains intact. &lt;a class="link" href="https://investor.tsmc.com/english/quarterly-results/2026/q2" target="_blank" rel="noopener"
 &gt;TSMC Q2 2026 Official Results&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;ASML confirmed the same direction. Q2 revenue of €9.326 billion and net income of €2.918 billion were reported, and the 2026 revenue outlook was set at €43–45 billion. The capacity roadmap reveals the timeline at which current orders translate into future supply.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Equipment&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026 Capacity&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2027 Plan&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2028 Under Review&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Low-NA EUV&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~65 units&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~30% expansion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Additional ~30% expansion under review&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DUV immersion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~130 units&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~30% expansion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Additional ~30% expansion under review&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;ASML stated that AI investment is driving demand for leading-edge logic and memory, and that customer capacity commitments are converting into long-term demand visibility. These figures weaken the strong bear case that AI orders consist entirely of duplicate bookings. &lt;a class="link" href="https://www.asml.com/en/news/press-releases/2026/q2-2026-financial-results" target="_blank" rel="noopener"
 &gt;ASML Q2 2026 Official Results&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="32-memory-prices-remain-elevated"&gt;3.2 Memory Prices Remain Elevated
&lt;/h3&gt;&lt;p&gt;TrendForce projected Q2 2026 commodity DRAM contract prices to rise 58–63% quarter-over-quarter, and NAND prices to rise 70–75%. The Q3 server DRAM price growth forecast was revised to 13–18%, but the upward trajectory is maintained. &lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260331-12995.html" target="_blank" rel="noopener"
 &gt;Q2 DRAM/NAND Outlook&lt;/a&gt;, &lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260709-13140.html" target="_blank" rel="noopener"
 &gt;Q3 Server DRAM Outlook&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;It is important to distinguish between a deceleration in the rate of price increases and an actual price decline. Given how rapidly prices rose in Q2, the quarter-over-quarter growth rate in Q3 may slow. Nonetheless, if absolute prices remain elevated, suppliers&amp;rsquo; operating income can remain strong. The risk the market is pricing is not a decline in earnings, but a deceleration in earnings growth.&lt;/p&gt;
&lt;h3 id="33-hbm-ramp-cannibalizes-commodity-dram-supply"&gt;3.3 HBM Ramp Cannibalizes Commodity DRAM Supply
&lt;/h3&gt;&lt;p&gt;HBM requires more wafer area, stacking, packaging, and testing capacity than commodity DRAM to produce the same number of bits. As the three memory manufacturers increase their HBM mix, HBM shipments expand while the effective bit supply of commodity DRAM can shrink.&lt;/p&gt;
&lt;p&gt;As a result, the bottleneck broadens in the following sequence:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;GPU / ASIC
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; HBM
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; Server DRAM
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; eSSD and NAND
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; Packaging / Substrate / Test
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; Networking / Power / Cooling
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;AI agents and long-context inference do not rely solely on HBM for model weights. They also consume server DRAM and eSSD for KV cache and data serving, and consume more networking to connect multiple accelerators and sparse mixture-of-expert models. This is why aggregate memory tier volumes can continue to expand even if HBM unit pricing comes under pressure.&lt;/p&gt;
&lt;h3 id="34-2027-shortage-is-a-strong-conclusion-25-shortage-by-2030-is-a-scenario"&gt;3.4 2027 Shortage Is a Strong Conclusion; 2.5× Shortage by 2030 Is a Scenario
&lt;/h3&gt;&lt;p&gt;Comparing 2030 HBM demand of 26.7 EB against supply of 10.6 EB yields demand at 2.52× supply. The arithmetic is reproducible. The issue lies not in the result but in whether the underlying assumptions hold simultaneously. The base model assumes 24× token growth, 5× model size expansion, 4× context length extension, and a 70% HBM retention rate for KV cache — all simultaneously.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Assumption Change&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2030 HBM Demand&lt;/th&gt;
 &lt;th style="text-align: right"&gt;vs. 10.6 EB Supply&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Base model&lt;/td&gt;
 &lt;td style="text-align: right"&gt;26.7 EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.52×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Model size 2×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18.5 EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.75×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KV efficiency 6×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;22.3 EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.10×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM retention rate 50%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;22.9 EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.16×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Tokens 12×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.2 EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.53×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Composite bear assumptions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.5 EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.62×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Reducing any single variable still leaves a shortage, but if multiple bearish assumptions hold simultaneously, oversupply becomes possible. The investable statement is therefore: &amp;ldquo;The market is likely to remain tight through 2027.&amp;rdquo; The statement &amp;ldquo;Supply will be definitively 2.5× short by 2030&amp;rdquo; cannot be directly plugged into a price target.&lt;/p&gt;
&lt;h2 id="4-risks-strengthened-this-week"&gt;4. Risks Strengthened This Week
&lt;/h2&gt;&lt;h3 id="41-earnings-growth-rate-has-become-more-important-than-strong-absolute-earnings"&gt;4.1 Earnings Growth Rate Has Become More Important Than Strong Absolute Earnings
&lt;/h3&gt;&lt;p&gt;The SK Hynix Q2 operating income estimate cut originated from adjustments to ASP assumptions rather than from a collapse in shipment volumes.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Report&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q2 2026 Operating Income Estimate&lt;/th&gt;
 &lt;th&gt;Key Adjustment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Mirae Asset Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;From ₩70.7 trillion to ₩62.3 trillion&lt;/td&gt;
 &lt;td&gt;DRAM ASP from +40.6% to +32.9%; NAND from +55% to +50%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Investment &amp;amp; Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;₩60.4 trillion&lt;/td&gt;
 &lt;td&gt;DRAM ASP +28.9%; NAND +50.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;An operating income in the ₩60 trillion range is very large in absolute terms. However, stock prices responded more to the magnitude of the revision from prior estimates than to the absolute level of earnings. The 15.37% decline in SK Hynix on July 13 is better explained by the combination of shifting expectations and concentrated technical supply-demand dynamics than by any conclusion that the upcycle has ended.&lt;/p&gt;
&lt;p&gt;This is the central logic of the weekly analysis. Strong demand alone does not drive stock prices higher. Demand must beat market expectations, and evidence that elevated earnings persist into 2028 is required.&lt;/p&gt;
&lt;h3 id="42-long-term-contracts-create-both-an-earnings-floor-and-a-price-ceiling"&gt;4.2 Long-Term Contracts Create Both an Earnings Floor and a Price Ceiling
&lt;/h3&gt;&lt;p&gt;HBM long-term supply agreements provide volume visibility, customer lock-in, predictability of payback periods, and downside protection when spot prices fall. On the other side, the following concerns apply:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Even if spot prices spike, a company&amp;rsquo;s blended ASP may rise only gradually.&lt;/li&gt;
&lt;li&gt;Pricing formulas, renegotiation mechanisms, minimum purchase commitments, and cancellation terms are not disclosed.&lt;/li&gt;
&lt;li&gt;Customer compliance with minimum purchase obligations and credit risk remain with the supplier.&lt;/li&gt;
&lt;li&gt;Up-front payments and long-term volume commitments may not flow through to accounting earnings and operating cash flow at the same pace.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Interpreting long-term contracts as an unconditional reinforcement of pricing power is only half the picture. Contracts raise the earnings floor and reduce volatility, but they may not capture the full upside of a spot price spike. The SK Hynix Q2 estimate revision was the first instance of this gap appearing in reported numbers. The detailed figures are discussed in &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-2q-earnings-cut-mirae-kis-lta-target-price-2026-07-14/" &gt;SK Hynix Q2 Earnings Cut and Maintained Price Targets&lt;/a&gt;.&lt;/p&gt;
&lt;h3 id="43-equipment-orders-are-both-current-demand-and-future-supply"&gt;4.3 Equipment Orders Are Both Current Demand and Future Supply
&lt;/h3&gt;&lt;p&gt;Reading ASML&amp;rsquo;s capacity expansion solely as near-term bullish evidence misses the time dimension.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Low-NA EUV 2027 capacity ≈ 65 units × 1.30 = 84.5 units
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;DUV immersion 2027 capacity ≈ 130 units × 1.30 = 169 units
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Equipment manufacturers recognize orders as revenue in 2026–2027 first. Actual bit production at memory manufacturers can increase from 2028 onward, after equipment installation, yield stabilization, and packaging capacity additions. The same capex that initially boosts equipment and materials earnings later pressures memory prices and margins.&lt;/p&gt;
&lt;p&gt;The metrics to watch for 2028 are therefore not the announced investment totals. They are equipment installation timing, new wafer starts, yield ramp, packaging throughput, the increase in depreciation, and actual bit output.&lt;/p&gt;
&lt;h3 id="44-high-memory-prices-can-erode-customer-demand"&gt;4.4 High Memory Prices Can Erode Customer Demand
&lt;/h3&gt;&lt;p&gt;When memory prices rise, customers can respond in four ways:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Pull forward inventory purchases before prices rise further.&lt;/li&gt;
&lt;li&gt;Raise server and end-product prices.&lt;/li&gt;
&lt;li&gt;Reduce memory content or downgrade specifications.&lt;/li&gt;
&lt;li&gt;Delay purchases and projects.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The first behavior makes near-term orders appear stronger but simply pulls forward future demand. The other three represent pathways through which suppliers&amp;rsquo; pricing power ultimately destroys customer demand. The IBM and Ericsson examples showed that memory supply constraints have begun to affect end-customer budget allocation and margins. We are currently in a phase where pre-buying and specification adjustments are visible before large-scale order cancellations.&lt;/p&gt;
&lt;h3 id="45-ai-capex-bottleneck-is-shifting-to-cost-of-capital"&gt;4.5 AI Capex Bottleneck Is Shifting to Cost of Capital
&lt;/h3&gt;&lt;p&gt;Even if AI demand is real, if the investing entity&amp;rsquo;s cash flow cannot keep pace with depreciation and financing costs, the pace of investment will slow. Going forward, the following four items matter more than the absolute capex figure from hyperscalers:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Whether AI and cloud gross margin growth outpaces the increase in depreciation.&lt;/li&gt;
&lt;li&gt;Whether data center power delivery and GPU utilization ramp according to plan.&lt;/li&gt;
&lt;li&gt;Whether returns on new projects, including borrowing costs, exceed the cost of capital.&lt;/li&gt;
&lt;li&gt;Whether the credit quality and up-front payments from customers that have signed HBM long-term contracts remain stable.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The strength of orders at TSMC and ASML means they are not evidence of current demand collapse. However, in U.S. Big Tech earnings releases at the end of July, investors should examine AI gross margins, free cash flow, and the slope of 2027 investment guidance alongside — not instead of — total capex figures.&lt;/p&gt;
&lt;h2 id="5-was-the-share-price-plunge-a-supply-demand-shock-or-a-fundamental-repricing"&gt;5. Was the Share Price Plunge a Supply-Demand Shock or a Fundamental Repricing?
&lt;/h2&gt;&lt;p&gt;This week&amp;rsquo;s sharp decline in Korean semiconductor stocks bears the hallmarks of leverage unwinding, program selling, and a market-cap structure heavily concentrated in a handful of large-cap names — all of which amplified the drawdown. Simultaneous selling by foreign and institutional investors caused single-day price moves far in excess of any change in underlying business value.&lt;/p&gt;
&lt;p&gt;Yet concluding that &amp;ldquo;it was merely a supply-demand shock, so everything rebounds&amp;rdquo; carries its own danger. If relative strength fails to recover even after strong earnings and pricing data, it may signal that the market is actively lowering its estimates for 2028 EPS and normalized multiples.&lt;/p&gt;
&lt;p&gt;The diagnostic framework is straightforward.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Signals pointing toward a supply-demand shock&lt;/th&gt;
 &lt;th&gt;Signals pointing toward a fundamental repricing&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign and program selling decelerates quickly&lt;/td&gt;
 &lt;td&gt;Selling continues despite positive news&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Relative strength recovers on volume&lt;/td&gt;
 &lt;td&gt;Semiconductors continue to underperform the broader market&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;12-month forward EPS revisions remain positive&lt;/td&gt;
 &lt;td&gt;Samsung Electronics, SK Hynix, and Micron EPS all revised down simultaneously&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Contract prices rise and inventories stabilize&lt;/td&gt;
 &lt;td&gt;Inventories build while contract prices decline simultaneously&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Big-Tech CAPEX and AI profitability both improve&lt;/td&gt;
 &lt;td&gt;CAPEX cuts or deteriorating AI margins&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This distinction is precisely why price and earnings estimates must be tracked together. Watching only supply-demand dynamics risks missing long-term risks; watching only results risks missing the additional downside created by crowded positioning.&lt;/p&gt;
&lt;h2 id="6-china-is-three-variables-at-once"&gt;6. China Is Three Variables at Once
&lt;/h2&gt;&lt;h3 id="61-china-is-a-major-demand-market"&gt;6.1 China Is a Major Demand Market
&lt;/h3&gt;&lt;p&gt;China exposure figures disclosed in company filings use different definitions, making direct comparisons difficult.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;China Exposure Metric&lt;/th&gt;
 &lt;th&gt;Caveat&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30.1%&lt;/td&gt;
 &lt;td&gt;Based on standalone consolidated revenue — not memory-only share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1Q26 24.3%, full-year 2025 19.7%&lt;/td&gt;
 &lt;td&gt;Based on sales entity location&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td style="text-align: right"&gt;FY25 China + Hong Kong 10.1%&lt;/td&gt;
 &lt;td&gt;Based on customer headquarters location&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung and SK Hynix are directly exposed if Chinese demand contracts. Micron&amp;rsquo;s direct exposure is comparatively low, but it cannot escape the impact if global pricing declines. Revenue-share figures alone are insufficient to assess China risk.&lt;/p&gt;
&lt;h3 id="62-china-is-a-commodity-memory-competitor"&gt;6.2 China Is a Commodity Memory Competitor
&lt;/h3&gt;&lt;p&gt;CXMT has announced mass production of LPDDR5X at 8,533 Mbps and 9,600 Mbps, with its 10,667 Mbps product reportedly in customer qualification. Its estimated share of global DRAM revenue for Q1 2026 stands at approximately 8%. However, an Apple supply agreement, broad adoption in flagship global products, and HBM yield have not been confirmed. &lt;a class="link" href="https://www.cxmt.com/en/news/info_19.html" target="_blank" rel="noopener"
 &gt;CXMT LPDDR5X Official Announcement&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The near-term threat is concentrated in two channels rather than HBM displacement:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Localization of commodity DRAM and NAND in Chinese smartphones and PCs.&lt;/li&gt;
&lt;li&gt;Using CXMT as a fourth-supplier card in price negotiations with the established memory trio.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;If volume displaced from China migrates to other markets, it could pull down global contract prices for Samsung, SK Hynix, and Micron. This price contagion may matter more as a second-order effect than a direct decline in China revenue. A detailed exposure breakdown is available in &lt;a class="link" href="https://koreainvestinsights.com/post/china-memory-localization-exposure-samsung-hynix-micron-2026-07-18/" &gt;China Memory Localization and the Big Three&lt;/a&gt;.&lt;/p&gt;
&lt;h3 id="63-china-can-be-segmented-by-policy"&gt;6.3 China Can Be Segmented by Policy
&lt;/h3&gt;&lt;p&gt;On July 16, members of the House Select Committee on China sent a letter to the Department of Commerce calling for tightened controls on YMTC, accelerated review of CXMT for Entity List inclusion, restrictions on Chinese DRAM and HBM procurement in AI systems, data centers, federal IT, and critical infrastructure, and coordination with South Korea, Japan, and the EU. This is a policy demand — not an implemented purchasing ban. &lt;a class="link" href="https://chinaselectcommittee.house.gov/media/letters/moolenaar-whitesides-to-secretary-lutnick-hold-firm-on-chinese-memory-chips-ban" target="_blank" rel="noopener"
 &gt;House Select Committee on China Official Letter&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;If enacted, the global memory market could bifurcate along the following lines:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Market&lt;/th&gt;
 &lt;th&gt;Possible Structure&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;United States and allies&lt;/td&gt;
 &lt;td&gt;Verified supply chains, higher prices, market-share defense by the memory trio&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Chinese domestic market&lt;/td&gt;
 &lt;td&gt;Localization, lower prices, medium-term oversupply risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Micron stands to benefit most directly from U.S. policy. Samsung and SK Hynix carry both the advantage of being allied-nation suppliers and the exposure of their China businesses. This policy factor is less a catalyst that directly lifts 2026 HBM earnings and more a variable that reduces the China-oversupply discount rate priced into 2028 valuations.&lt;/p&gt;
&lt;h2 id="7-kimi-k3-and-efficiency-demand-shifting-not-demand-destruction"&gt;7. Kimi K3 and Efficiency: Demand Shifting, Not Demand Destruction
&lt;/h2&gt;&lt;p&gt;According to Kimi&amp;rsquo;s official announcement, K3 has 2.8 trillion total parameters, activates 16 of 896 experts per token, and supports a 1 million-token context window. The sparse expert architecture reduces compute per request, but still requires large memory and high-bandwidth networking to store all weights and route expert selection. &lt;a class="link" href="https://www.kimi.com/blog/kimi-k3" target="_blank" rel="noopener"
 &gt;Kimi K3 Official Announcement&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The semiconductor impact of efficiency gains is better understood through the following equation:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Total semiconductor demand = silicon cost per task × total number of tasks × fraction self-hosted
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;ul&gt;
&lt;li&gt;Sparse architectures, caching, and quantization reduce GPU time and HBM footprint per task.&lt;/li&gt;
&lt;li&gt;Falling API prices increase AI usage volumes and the number of agents.&lt;/li&gt;
&lt;li&gt;Open-weight models may increase self-hosted server deployments by enterprises and governments.&lt;/li&gt;
&lt;li&gt;Memory tiering shifts some HBM workload onto server DRAM, eSSD, and remote caches.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Current firm orders and CAPEX data suggest that volume growth is outpacing efficiency gains. Over a longer horizon, however, the volume elasticity of server DRAM, eSSD, networking, and power may prove more favorable than the pricing power of premium GPUs and HBM. Efficiency is less a variable that &amp;ldquo;reduces total semiconductor consumption&amp;rdquo; and more one that reshapes &amp;ldquo;which semiconductors capture the revenue.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;Two caveats apply here. The KV-cache savings rates and theoretical throughput from Kimi&amp;rsquo;s linear research models cannot be directly applied to K3 in production. Additionally, self-reported performance benchmarks should not be weighted equally with independent verification. Weights, licensing terms, independent throughput results, and actual tokens and total cost per task require confirmation.&lt;/p&gt;
&lt;h2 id="8-five-clocks--when-they-are-watched-together-the-contradictions-resolve"&gt;8. Five Clocks — When They Are Watched Together, the Contradictions Resolve
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Clock&lt;/th&gt;
 &lt;th&gt;This Week&amp;rsquo;s Observation&lt;/th&gt;
 &lt;th&gt;Investment Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Pricing&lt;/td&gt;
 &lt;td&gt;Sharp 2Q DRAM and NAND price increases; 3Q pace of gains decelerating&lt;/td&gt;
 &lt;td&gt;2026 earnings look strong, but the rate of increase may slow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investment&lt;/td&gt;
 &lt;td&gt;TSMC expanding investment; ASML orders and capacity growing&lt;/td&gt;
 &lt;td&gt;Current demand confirmed while 2027–2028 supply response is booked&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Construction&lt;/td&gt;
 &lt;td&gt;Power, land, and packaging bottlenecks persist&lt;/td&gt;
 &lt;td&gt;Announced investment lags before translating into actual production ramp&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Monetization&lt;/td&gt;
 &lt;td&gt;AI usage rising; customer costs and financing charges also rising&lt;/td&gt;
 &lt;td&gt;AI gross margin and free cash flow require scrutiny alongside revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Share price&lt;/td&gt;
 &lt;td&gt;Sluggish reaction to positive news; concentrated supply-demand shock in Korea&lt;/td&gt;
 &lt;td&gt;Market is discounting duration and positioning rather than current earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The pricing and investment clocks are running fast, while the construction and monetization clocks are lagging. The share-price clock is the forward-looking reflection of whether that gap narrows or widens ahead. That is why strong results and raised CAPEX guidance can trigger selling on the same day.&lt;/p&gt;
&lt;p&gt;The &amp;ldquo;55% industry strength / valuation weakness coexistence&amp;rdquo; from the prior piece and &amp;ldquo;P2 at 40%&amp;rdquo; below are not two different readings of the same table. The 55% was a broad category in which multiples remain compressed even as the industry stays strong. The more granular distribution here splits that category further into the early phases of P2 and P3. Because new data from TSMC and ASML confirmed current demand while simultaneously expanding the implied supply response, a portion of the long-term scarcity probability has been shifted toward supply normalization.&lt;/p&gt;
&lt;h2 id="9-scenario-update"&gt;9. Scenario Update
&lt;/h2&gt;&lt;p&gt;The probabilities below are judgment values reflecting information confirmed as of July 19 — not historical base rates.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability&lt;/th&gt;
 &lt;th&gt;Conditions for Validity&lt;/th&gt;
 &lt;th&gt;Expected Market Behavior&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P1. Sustained Scarcity&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td&gt;AI usage, CAPEX, and utilization rates continue to overwhelm supply and efficiency gains; HBM4 pricing and orders hold&lt;/td&gt;
 &lt;td&gt;High-beta memory names such as SK Hynix and Micron, alongside packaging and equipment, outperform&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P2. Strong Earnings, Compressed Multiples&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td&gt;Earnings remain strong in 2026–2027 but concerns over 2028 supply and customer profitability persist&lt;/td&gt;
 &lt;td&gt;Range-bound trading and elevated volatility despite earnings growth; Samsung Electronics carries relative advantage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P3. Efficiency and Supply Normalization&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td&gt;2027–2028 capacity expansion, Chinese commodity supply, customer price resistance, and declining AI costs converge&lt;/td&gt;
 &lt;td&gt;HBM scarcity premium narrows; equipment and materials orders peak before memory itself&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P4. Financial or Demand Dislocation&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Big-Tech CAPEX cuts, a credit event, rising inventories, and simultaneous contract price and EPS downgrades&lt;/td&gt;
 &lt;td&gt;Sector-wide semiconductor decline; cash and non-semiconductor diversification become more important&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The previous detailed distribution was 35% / 40% / 15% / 10%. This revision moves 10 percentage points from P1 to P3. TSMC and ASML order data did not raise the probability of P4. Conversely, equipment capacity expansion and the pricing burden on customers have increased the probability of supply and efficiency normalization relative to prolonged scarcity.&lt;/p&gt;
&lt;h2 id="10-roles-by-stock-and-value-chain"&gt;10. Roles by Stock and Value Chain
&lt;/h2&gt;&lt;h3 id="samsung-electronics-risk-adjusted-core"&gt;Samsung Electronics: Risk-Adjusted Core
&lt;/h3&gt;&lt;p&gt;Samsung absorbs commodity DRAM and NAND price appreciation across a broad market share base while retaining the option value of catching up in HBM and recovering its foundry business. It also holds relative defensive characteristics should open models and low-cost inference proliferate, shifting some HBM workload toward commodity memory and storage.&lt;/p&gt;
&lt;p&gt;Key risks include exposure to China and commodity memory, uncertainty around HBM4 customer qualification, and the high capital burden of foundry operations. Samsung offers more defensive characteristics than SK Hynix in P2 and P3, but diversification does not equal profitability if HBM catch-up has not been validated in earnings.&lt;/p&gt;
&lt;h3 id="sk-hynix-pure-beta-on-sustained-scarcity"&gt;SK Hynix: Pure Beta on Sustained Scarcity
&lt;/h3&gt;&lt;p&gt;SK Hynix&amp;rsquo;s leading HBM market share, co-development relationships with customers, and long-term supply agreements give it the strongest earnings leverage under P1. Conversely, it also carries the highest sensitivity to earnings and multiple compression if supply expansion and efficiency gains accelerate under P3.&lt;/p&gt;
&lt;p&gt;Additional judgment requires tracking HBM4 and HBM4E customer qualifications, yield, average selling price terms embedded in long-term contracts, and foreign-investor and program-trading flows alongside relative strength. Business quality and price-embedded margin of safety must be assessed separately.&lt;/p&gt;
&lt;h3 id="micron-intersection-of-us-policy-and-pure-play-memory"&gt;Micron: Intersection of U.S. Policy and Pure-Play Memory
&lt;/h3&gt;&lt;p&gt;Micron is the most direct beneficiary of U.S.-based manufacturing, policy support, and restrictions on Chinese memory procurement. It offers a clean read-through to the HBM, DRAM, and NAND upcycle. However, its limited business diversification means it is also directly exposed to any normalization in global average selling prices. Even with a low China revenue share, Micron cannot easily avoid the impact if surplus Chinese-origin volumes depress global pricing.&lt;/p&gt;
&lt;h3 id="sandisk-tactical-beta-on-nand-and-essd"&gt;SanDisk: Tactical Beta on NAND and eSSD
&lt;/h3&gt;&lt;p&gt;SanDisk is directly exposed to NAND supply constraints and AI server eSSD demand. It stands to benefit most if low-cost inference and memory tiering shift incremental storage volumes away from HBM. At the same time, NAND exhibits greater price volatility than DRAM and carries more direct YMTC exposure, giving it a more tactical character relative to core positions.&lt;/p&gt;
&lt;h3 id="tsmc-and-asml-demand-validators-and-leading-indicators-of-future-supply"&gt;TSMC and ASML: Demand Validators and Leading Indicators of Future Supply
&lt;/h3&gt;&lt;p&gt;TSMC demonstrates the quality of AI demand and the depth of the leading-edge foundry moat. ASML is the earliest observable data point for customers&amp;rsquo; long-term capacity expansion intentions. Both companies&amp;rsquo; results serve as supporting evidence for the memory thesis, but they are not simple coincident indicators. High CAPEX and growing equipment supply signal not only future memory demand but also future supply expansion.&lt;/p&gt;
&lt;h3 id="korean-equipment-and-materials-actual-orders-and-recurring-revenue-over-published-plans"&gt;Korean Equipment and Materials: Actual Orders and Recurring Revenue Over Published Plans
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Category&lt;/th&gt;
 &lt;th&gt;Candidate Names&lt;/th&gt;
 &lt;th&gt;Data to Confirm&lt;/th&gt;
 &lt;th&gt;Current Role&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Recurring consumables&lt;/td&gt;
 &lt;td&gt;Hana Materials, Wooldex&lt;/td&gt;
 &lt;td&gt;Customer utilization rates, OEM channels, revenue and operating margin correlation&lt;/td&gt;
 &lt;td&gt;Priority research group&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Process equipment&lt;/td&gt;
 &lt;td&gt;KC Tech, Nextin&lt;/td&gt;
 &lt;td&gt;Actual orders, customer qualification, backlog&lt;/td&gt;
 &lt;td&gt;Conditional confirmation group&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM back-end&lt;/td&gt;
 &lt;td&gt;Techwing&lt;/td&gt;
 &lt;td&gt;Additional Cube Prober orders and production revenue&lt;/td&gt;
 &lt;td&gt;Event confirmation group&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Infrastructure construction&lt;/td&gt;
 &lt;td&gt;Hanyang E&amp;amp;E, others&lt;/td&gt;
 &lt;td&gt;Major contract wins, order-to-revenue ratio, margins&lt;/td&gt;
 &lt;td&gt;Order-event group rather than core&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The confirmation sequence for investments is: capacity announcement → actual order placement → backlog build → revenue recognition → recurring consumables. Equipment names may peak in earnings earlier than memory production itself. Rather than buying broadly on fab plans alone, the focus should narrow to companies where actual orders and margins are confirmed.&lt;/p&gt;
&lt;h2 id="11-counter-arguments-where-this-thesis-could-be-wrong"&gt;11. Counter-Arguments: Where This Thesis Could Be Wrong
&lt;/h2&gt;&lt;h3 id="demand-evidence-may-have-been-double-counted"&gt;Demand Evidence May Have Been Double-Counted
&lt;/h3&gt;&lt;p&gt;Orders from TSMC, ASML, and the memory trio may all originate from the same Big-Tech capital expenditure plans. Adding orders across multiple tiers of the supply chain as separate demand signals overstates actual end demand. Power connection timelines, GPU utilization, server shipments, and cloud AI gross margins are the terminal verification data points.&lt;/p&gt;
&lt;h3 id="the-2030-demand-model-may-underestimate-efficiency-gains"&gt;The 2030 Demand Model May Underestimate Efficiency Gains
&lt;/h3&gt;&lt;p&gt;If any one of — token count, model size, context length, or HBM retention rate — is reduced, a shortage persists. But if all four decline simultaneously, a 2030 oversupply scenario becomes plausible. The directional case for long-term shortage and the numerical magnitude of that shortage must be assessed separately.&lt;/p&gt;
&lt;h3 id="capex-may-grow-faster-than-returns"&gt;CAPEX May Grow Faster Than Returns
&lt;/h3&gt;&lt;p&gt;Suppliers must spend more on new fabs and packaging even when order books are full. Customers must absorb data center depreciation and interest expense. If revenue growth fails to translate into free cash flow growth, high earnings may attract depressed multiples.&lt;/p&gt;
&lt;h3 id="china-policy-may-not-be-implemented"&gt;China Policy May Not Be Implemented
&lt;/h3&gt;&lt;p&gt;The House Committee letter is a policy signal — not an executive order or coordinated allied action. If CXMT enters only a limited set of consumer products within China and enforcement remains loose, pricing leverage for the incumbents may erode before any meaningful market-share shift occurs.&lt;/p&gt;
&lt;h3 id="open-model-company-disclosures-may-not-be-independently-reproducible"&gt;Open-Model Company Disclosures May Not Be Independently Reproducible
&lt;/h3&gt;&lt;p&gt;Kimi K3&amp;rsquo;s official specifications are confirmed, but throughput, real-world token consumption, and total serving cost require independent verification. Conversely, if efficiency gains arrive faster than expected and usage elasticity is lower than assumed, the HBM scarcity premium could compress more severely.&lt;/p&gt;
&lt;h3 id="the-entire-price-decline-may-have-been-attributed-to-forced-liquidation"&gt;The Entire Price Decline May Have Been Attributed to Forced Liquidation
&lt;/h3&gt;&lt;p&gt;Leverage unwinding and program selling explain the magnitude of the drop. But if prices fail to recover even in the face of positive news, the market may be actively lowering its 2028 earnings and normalized-multiple assumptions. A supply-demand explanation cannot substitute for validating earnings durability.&lt;/p&gt;
&lt;h2 id="12-observable-conditions-that-would-change-the-thesis"&gt;12. Observable Conditions That Would Change the Thesis
&lt;/h2&gt;&lt;p&gt;Rather than any single article or a one-day bounce, the key is whether two independent data axes move together.&lt;/p&gt;
&lt;h3 id="conditions-that-raise-the-probability-of-a-bullish-outcome"&gt;Conditions That Raise the Probability of a Bullish Outcome
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;Big-Tech CAPEX growth and improvement in AI and cloud gross margins are confirmed simultaneously.&lt;/li&gt;
&lt;li&gt;Data center power connections, GPU utilization rates, and server shipments all rise together.&lt;/li&gt;
&lt;li&gt;HBM4, server DRAM, and eSSD prices, along with 12-month forward EPS, continue to be revised upward.&lt;/li&gt;
&lt;li&gt;Samsung Electronics, SK Hynix, and Micron rally on positive news and recover relative strength versus the semiconductor index.&lt;/li&gt;
&lt;li&gt;Foreign-investor and program selling decelerates, and a trend recovery accompanied by volume emerges.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="conditions-that-raise-the-probability-of-p3-normalization"&gt;Conditions That Raise the Probability of P3 Normalization
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;2027–2028 equipment delivery and new bit capacity come online faster than planned.&lt;/li&gt;
&lt;li&gt;HBM yields and packaging productivity improve sharply while supply from all three incumbents expands simultaneously.&lt;/li&gt;
&lt;li&gt;Following memory price increases, customers reduce their content per device and purchasing deferrals become widespread.&lt;/li&gt;
&lt;li&gt;CXMT DDR5 and LPDDR5X customer qualifications and sales outside China increase.&lt;/li&gt;
&lt;li&gt;Server DRAM and HBM contract price growth rates and 2028 EPS are simultaneously revised lower.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="p4-warning-conditions"&gt;P4 Warning Conditions
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;U.S. Big-Tech companies cut CAPEX or cancel or delay data center projects.&lt;/li&gt;
&lt;li&gt;Credit spreads and refinancing risk for AI infrastructure customers spike sharply.&lt;/li&gt;
&lt;li&gt;DRAM and NAND inventory builds and contract price declines occur simultaneously.&lt;/li&gt;
&lt;li&gt;12-month forward EPS for Samsung Electronics, SK Hynix, and Micron are all revised downward together.&lt;/li&gt;
&lt;li&gt;Semiconductor relative strength and market breadth (advance-decline) both deteriorate simultaneously.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="13-data-to-track-from-late-july-onward"&gt;13. Data to Track from Late July Onward
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;th&gt;Data to Track&lt;/th&gt;
 &lt;th&gt;Judgment That Changes&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Late July&lt;/td&gt;
 &lt;td&gt;U.S. Big-Tech earnings and conference calls&lt;/td&gt;
 &lt;td&gt;CAPEX, AI gross margin, free cash flow, memory price resistance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Late July&lt;/td&gt;
 &lt;td&gt;Additional Kimi K3 technical disclosures and independent evaluations&lt;/td&gt;
 &lt;td&gt;Which is larger — efficiency gains or usage growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Late July&lt;/td&gt;
 &lt;td&gt;Samsung Electronics and SK Hynix detailed IR disclosures&lt;/td&gt;
 &lt;td&gt;HBM4, long-term contracts, DRAM and NAND ASP, shipments, CAPEX&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;August onward&lt;/td&gt;
 &lt;td&gt;3Q contract pricing and distribution inventory levels&lt;/td&gt;
 &lt;td&gt;Whether the pace of price gains has peaked and pre-buying is normalizing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Quarterly&lt;/td&gt;
 &lt;td&gt;ASML bookings, TSMC CAPEX, memory equipment orders&lt;/td&gt;
 &lt;td&gt;Speed of 2027–2028 supply response&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ongoing&lt;/td&gt;
 &lt;td&gt;CXMT and YMTC regulatory actions and customer qualifications&lt;/td&gt;
 &lt;td&gt;Chinese supply&amp;rsquo;s price contagion to global markets and market bifurcation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="14-final-judgment"&gt;14. Final Judgment
&lt;/h2&gt;&lt;p&gt;This week served as the first comprehensive stress test of the memory supercycle narrative. Share prices moved sharply, but TSMC and ASML results, DRAM and NAND pricing, and server memory demand all tilted the weight of evidence toward 2026–2027 physical shortages remaining intact.&lt;/p&gt;
&lt;p&gt;At the same time, the same data made 2028 risks more clearly defined. Equipment capacity and CAPEX are expanding, long-term HBM supply agreements slow the pass-through of price increases, customers must absorb elevated memory prices alongside data center financing costs, and Chinese commodity memory and open-model efficiency gains will exert pressure on 2028 pricing and multiples.&lt;/p&gt;
&lt;p&gt;The choice between &amp;ldquo;the boom is over&amp;rdquo; and &amp;ldquo;this is a permanent supercycle&amp;rdquo; is therefore a false one. The current base-case path is as follows:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;2026–2027: Physical shortage and high earnings
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; 2027–2028: Equipment, capacity, and yield response
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; Pace of price increases decelerates
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; Absolute earnings remain above historical levels, but multiples compress
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;-&amp;gt; Return dispersion across names driven by product mix, contract structure, and China exposure
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;From this point, the central task is not to re-prove that AI demand exists. It is to measure simultaneously the speed of the supply response, the monetization trajectory for customers, contract quality, 2028 earnings estimates, and how share prices react to positive news. Earnings today are speaking to 2026 and 2027. The market is already asking about 2028.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="key-sources"&gt;Key Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://investor.tsmc.com/english/quarterly-results/2026/q2" target="_blank" rel="noopener"
 &gt;TSMC Q2 2026 Official Earnings&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.asml.com/en/news/press-releases/2026/q2-2026-financial-results" target="_blank" rel="noopener"
 &gt;ASML Q2 2026 Official Earnings and Capacity Plans&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260331-12995.html" target="_blank" rel="noopener"
 &gt;TrendForce Q2 2026 DRAM and NAND Price Outlook&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260709-13140.html" target="_blank" rel="noopener"
 &gt;TrendForce Q3 2026 Server DRAM Outlook&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://chinaselectcommittee.house.gov/media/letters/moolenaar-whitesides-to-secretary-lutnick-hold-firm-on-chinese-memory-chips-ban" target="_blank" rel="noopener"
 &gt;U.S. House Select Committee on China — Letter Calling for Restrictions on Chinese Memory Procurement&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.bis.gov/ear/title-15/subtitle-b/chapter-vii/subchapter-c/part-744/ss-74416-entity-list" target="_blank" rel="noopener"
 &gt;U.S. Department of Commerce BIS Entity List Regulations&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.kimi.com/blog/kimi-k3" target="_blank" rel="noopener"
 &gt;Kimi K3 Official Announcement&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.cxmt.com/en/news/info_19.html" target="_blank" rel="noopener"
 &gt;CXMT LPDDR5X Official Announcement&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="limitations-of-public-data"&gt;Limitations of Public Data
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;The long-term contract pricing formulas for HBM from Samsung Electronics, SK Hynix, and Micron — including floor and ceiling prices, minimum purchase volumes, cancellation clauses, and customer-specific credit terms — have not been disclosed publicly.&lt;/li&gt;
&lt;li&gt;There are no reliable official figures for CXMT&amp;rsquo;s HBM yield or its progress toward Apple qualification.&lt;/li&gt;
&lt;li&gt;Full unit cost by model and actual per-accelerator serving costs for Chinese AI API providers are not publicly reported.&lt;/li&gt;
&lt;li&gt;The 26.7 EB HBM demand figure for 2030 is not a consensus forecast but a scenario constructed by combining several bullish assumptions.&lt;/li&gt;
&lt;li&gt;The July 19 channel checks on server DRAM and HBM4 could not be independently sourced from primary documents and are used solely for directional reference.&lt;/li&gt;
&lt;li&gt;The information cutoff date is July 19, 2026. Scenarios should be revisited following earnings releases from U.S. large-cap tech companies and memory makers expected in late July.&lt;/li&gt;
&lt;/ol&gt;

 &lt;blockquote&gt;
 &lt;p&gt;This article is an informational analysis based on publicly available sources and does not constitute a recommendation to buy or sell any specific security. Prices, supply-demand dynamics, earnings estimates, and policy conditions are subject to change; readers should verify the latest disclosures and assess findings against their own risk tolerance.&lt;/p&gt;

 &lt;/blockquote&gt;</description></item><item><title>U.S. House Letter Seeking Chinese Memory Procurement Restrictions: CXMT, YMTC, and the 2028 Cycle</title><link>https://koreainvestinsights.com/post/us-house-chinese-memory-procurement-ban-cxmt-ymtc-cycle-impact-2026-07-17/</link><pubDate>Fri, 17 Jul 2026 20:45:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/us-house-chinese-memory-procurement-ban-cxmt-ymtc-cycle-impact-2026-07-17/</guid><description>&lt;p&gt;On July 16, 2026, House Select Committee on China Chairman John Moolenaar and Ranking Member George Whitesides sent a letter to the Secretary of Commerce. The two lawmakers called for tighter export controls on CXMT and YMTC, a prohibition on U.S. companies procuring their memory for AI systems and data centers, and allied participation — from South Korea, Japan, and the EU — in the same supply-chain rules.&lt;/p&gt;
&lt;p&gt;The letter is real. Its language is strong. It is not, however, a regulation, an executive order, or a statute. &lt;strong&gt;This is a bipartisan policy demand addressed to the executive branch, not an enacted Chinese memory embargo.&lt;/strong&gt; Conflating the two risks materially overstating the near-term significance for equity prices.&lt;/p&gt;
&lt;p&gt;That does not mean the letter is inconsequential. If these demands translate into actual policy, the United States and its allies will find it harder to fill memory shortfalls with Chinese supply. This is not an event that immediately increases HBM shipments, but it does reduce the concern that CXMT and YMTC capacity additions in 2027–2028 will collapse global pricing. The China oversupply discount that investors have been applying to normalized earnings and target multiples for Samsung, SK Hynix, and Micron could narrow.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Context: This article follows &lt;a class="link" href="https://koreainvestinsights.com/post/apple-cxmt-dram-china-test-memory-pricing-korea-2026-07-14/" &gt;Will Apple Really Use CXMT?&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/cxmt-ipo-memory-price-risk-hbm-client-dram-2026-06-21/" &gt;CXMT IPO and Memory Price Risk&lt;/a&gt;. Those earlier pieces addressed Chinese memory supply expansion and Apple&amp;rsquo;s use of alternative suppliers as a negotiating lever. This piece examines how the memory cycle changes if the United States moves to close that supply channel through policy. Related material is available at the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;The July 16 House Select Committee letter has been officially confirmed. It is, however, a policy demand from two members of Congress addressed to the Commerce Department — not an enacted regulation.&lt;/li&gt;
&lt;li&gt;Three demands are on the table: strengthen controls on YMTC and accelerate CXMT&amp;rsquo;s Entity List review; prohibit U.S. persons and entities from procuring Chinese DRAM, HBM, and other memory for AI systems, data centers, federal IT, and critical infrastructure; coordinate with South Korea, Japan, and the EU on equivalent standards.&lt;/li&gt;
&lt;li&gt;The Entity List restricts the export, re-export, and in-country transfer of items subject to U.S. Export Administration Regulations when a listed entity is a party to the transaction. It does not automatically prohibit U.S. companies from purchasing finished Chinese memory products domestically. A purchase ban requires a separate executive order, agency directive, or legislation.&lt;/li&gt;
&lt;li&gt;If enacted, the policy is positive for market share and pricing power at Samsung, SK Hynix, and Micron. SanDisk and Kioxia also benefit as YMTC&amp;rsquo;s path into global NAND and enterprise SSD markets becomes more constrained.&lt;/li&gt;
&lt;li&gt;Product-level directness varies. Mobile and PC DRAM and NAND could feel the impact immediately; advanced HBM faces limited near-term supply alternatives regardless. This is more relevant to defending post-2028 competitive risk and multiples for SK Hynix than to 2026–2027 HBM EPS.&lt;/li&gt;
&lt;li&gt;The latest pricing outlook already signals a deceleration. TrendForce projected Q2 commodity DRAM contract prices up 58–63% quarter-over-quarter and NAND up 70–75%, but trimmed Q3 server DRAM growth to 13–18%. This letter is a policy signal that increases the plausibility of a sustained supply shortage into 2027 rather than a catalyst for another sharp near-term price spike.&lt;/li&gt;
&lt;li&gt;Our current stance is &lt;strong&gt;Watchlist — incorporate upon policy confirmation&lt;/strong&gt;. Chasing memory equities on the letter alone is premature; the correct sequence is to watch for a CXMT Entity List notice, a U.S. corporate procurement restriction, allied coordination, Apple&amp;rsquo;s sourcing decisions, and Q3 contract pricing.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;img alt="Transmission channels from the U.S. congressional Chinese memory policy letter" loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/images/posts/us-house-chinese-memory-policy-impact-2026-07-17.png"&gt;
&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;One-line view&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 This letter is not a HBM demand shock. It is a policy signal aimed at reducing the risk that Chinese-origin oversupply in 2027–2028 destroys global memory profitability.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-three-demands-confirmed-in-the-letter"&gt;1. Three Demands Confirmed in the Letter
&lt;/h2&gt;&lt;p&gt;The signatories are Republican John Moolenaar, Chairman of the House Select Committee on China, and Democrat George Whitesides. The letter is addressed to Commerce Secretary Howard Lutnick. The bipartisan signal is clear given that members of opposing parties co-signed. It is not, however, a full congressional vote or a Commerce Department decision.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Demand&lt;/th&gt;
 &lt;th&gt;Actual Content&lt;/th&gt;
 &lt;th&gt;Current Status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Tighten supply controls&lt;/td&gt;
 &lt;td&gt;Strengthen Entity List restrictions on YMTC; expedite formal Entity List review for CXMT&lt;/td&gt;
 &lt;td&gt;Policy request, not finalized&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Create demand-side restrictions&lt;/td&gt;
 &lt;td&gt;Prohibit U.S. persons and entities from procuring DRAM, HBM, and other memory from YMTC, CXMT, and similar firms for AI systems, data centers, federal IT, and critical infrastructure&lt;/td&gt;
 &lt;td&gt;Requires separate executive order or agency directive&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Allied coordination&lt;/td&gt;
 &lt;td&gt;Work with South Korea, Japan, and the EU to prevent CXMT and YMTC from entering allied supply chains&lt;/td&gt;
 &lt;td&gt;Diplomatic and policy coordination request&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The official letter also cited concerns that Apple and other U.S. technology companies are exploring purchases of Chinese memory. The demand is therefore not an abstract security slogan — it is an attempt to foreclose demand channels before CXMT and YMTC can achieve customer qualification and meaningful entry into global supply chains. &lt;a class="link" href="https://chinaselectcommittee.house.gov/media/letters/moolenaar-whitesides-to-secretary-lutnick-hold-firm-on-chinese-memory-chips-ban" target="_blank" rel="noopener"
 &gt;Official House Select Committee Letter&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="separating-fact-from-overstatement"&gt;Separating Fact from Overstatement
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Statement&lt;/th&gt;
 &lt;th&gt;Verdict&lt;/th&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. lawmakers demanded restrictions on Chinese memory procurement&lt;/td&gt;
 &lt;td&gt;Fact&lt;/td&gt;
 &lt;td&gt;Official letter dated July 16 confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;A full Chinese memory embargo has been finalized&lt;/td&gt;
 &lt;td&gt;Overstatement&lt;/td&gt;
 &lt;td&gt;No executive branch decision or implementing regulations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT&amp;rsquo;s Entity List designation has been confirmed&lt;/td&gt;
 &lt;td&gt;Unconfirmed&lt;/td&gt;
 &lt;td&gt;Only an accelerated review was requested&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple&amp;rsquo;s devices sold in China are automatically prohibited&lt;/td&gt;
 &lt;td&gt;Overstatement&lt;/td&gt;
 &lt;td&gt;The letter&amp;rsquo;s explicit scope covers AI, data centers, federal IT, and critical infrastructure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;South Korea and Japan have banned Chinese memory&lt;/td&gt;
 &lt;td&gt;False&lt;/td&gt;
 &lt;td&gt;Coordination was requested; no national measures have been announced&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The policy direction is forceful, but multiple implementation thresholds remain. Equity analysis should therefore start from &amp;ldquo;which discount disappears if this is implemented&amp;rdquo; rather than &amp;ldquo;the ban is already in effect.&amp;rdquo;&lt;/p&gt;
&lt;h2 id="2-the-entity-list-and-a-purchase-prohibition-are-distinct-instruments"&gt;2. The Entity List and a Purchase Prohibition Are Distinct Instruments
&lt;/h2&gt;&lt;p&gt;There is a reason the letter requests supply-side and demand-side controls separately: the two instruments operate differently under U.S. law.&lt;/p&gt;
&lt;p&gt;The Bureau of Industry and Security&amp;rsquo;s Entity List imposes license requirements on the export, re-export, or in-country transfer of items subject to the Export Administration Regulations when a listed entity is a party to the transaction. The covered items include U.S.-origin semiconductor equipment, components, design tools, software, and process support. &lt;a class="link" href="https://www.bis.gov/ear/title-15/subtitle-b/chapter-vii/subchapter-c/part-744/ss-74416-entity-list" target="_blank" rel="noopener"
 &gt;BIS Entity List Regulations&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;This mechanism pressures CXMT and YMTC on capacity and technology-node transitions. If equipment parts cannot be replaced on schedule and process support is unavailable, yields, wafer starts, and the pace of leading-edge conversion can slow.&lt;/p&gt;
&lt;p&gt;It does not, however, automatically prohibit U.S. companies from purchasing finished DRAM or NAND chips that were manufactured in China. That is precisely why the lawmakers included a separate procurement prohibition as their second demand. The structure is as follows:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Entity List tightening
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Restricts the path by which U.S.-origin equipment and technology reach Chinese memory makers
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Pressures capacity, yield, and technology-node transition
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Separate procurement ban
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Restricts the path by which U.S. companies purchase finished Chinese memory products
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Pressures customer qualification, revenue, and global market share
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The maximum effect requires both instruments operating simultaneously — making production harder at the top of the supply chain while cutting off customers at the bottom. If only the Entity List is tightened without a purchase restriction, Chinese firms can route around it using non-U.S. equipment and domestic customers. If only a procurement ban is imposed without strong equipment controls, Chinese domestic capacity can continue to expand.&lt;/p&gt;
&lt;h3 id="how-far-does-this-reach-for-apple"&gt;How Far Does This Reach for Apple?
&lt;/h3&gt;&lt;p&gt;The letter&amp;rsquo;s explicitly enumerated scope covers AI systems, data centers, federal IT, and critical infrastructure. Extending the prohibition to cover standard consumer iPhones and iPads sold within China would require either a broader definition in the implementing rules or a separate action.&lt;/p&gt;
&lt;p&gt;It is therefore premature to conclude that &lt;a class="link" href="https://koreainvestinsights.com/post/apple-cxmt-dram-china-test-memory-pricing-korea-2026-07-14/" &gt;Apple&amp;rsquo;s CXMT testing&lt;/a&gt; has been made immediately impossible. What has changed materially, however, is the regulatory stability Apple needs to use CXMT as a negotiating lever. If a supplier could be added to the Entity List at any time, and if certain products could become subject to procurement restrictions without advance notice, proceeding with volume qualification and long-term contracts becomes much more difficult.&lt;/p&gt;
&lt;h2 id="3-the-transmission-path-to-the-memory-cycle"&gt;3. The Transmission Path to the Memory Cycle
&lt;/h2&gt;&lt;p&gt;The core transmission channel is straightforward.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Procurement restrictions on Chinese memory
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Narrowed customer qualification and sales channels for CXMT and YMTC globally
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Market share defense at Samsung, SK Hynix, Micron, SanDisk, and Kioxia
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Delayed inflection point for DRAM and NAND price declines
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Higher expected normalized earnings and cash flows in 2027–2028
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Compression of the China oversupply discount
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The memory shortage itself is not new information. TrendForce projected Q2 2026 commodity DRAM contract prices up 58–63% quarter-over-quarter and NAND contract prices up 70–75%, driven by capacity prioritization toward HBM, server DRAM, and enterprise SSD at the expense of standard DRAM and NAND. &lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260331-12995.html" target="_blank" rel="noopener"
 &gt;TrendForce Q2 Outlook&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The latest forecast also shows a deceleration in the rate of price increase. On July 9, TrendForce projected Q3 server DRAM contract prices rising 13–18%. It noted that long-term supply agreements are capping some price upside, but also that server DRAM tightness could extend into 2027 — with RDIMM bit supply growth of 15–20% potentially failing to keep pace with server CPU shipment growth. Prices could continue to rise quarter-over-quarter from H2 2026 through H2 2027, but at a more gradual pace. &lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260709-13140.html" target="_blank" rel="noopener"
 &gt;TrendForce Q3 Outlook&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Two conclusions follow from this data.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;The near-term price spike is already in the market. The letter alone is insufficient to restart Q2-style rates of increase.&lt;/li&gt;
&lt;li&gt;The more important question is how quickly Chinese capacity additions end the 2027 supply shortage. If the United States and its allies exclude Chinese memory, the duration of the shortage extends.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="4-product-level-impact-is-not-uniform"&gt;4. Product-Level Impact Is Not Uniform
&lt;/h2&gt;&lt;p&gt;Treating CXMT and YMTC as a single &amp;ldquo;Chinese memory&amp;rdquo; block obscures the actual impact. CXMT&amp;rsquo;s core product is DRAM; YMTC&amp;rsquo;s is NAND. Customer qualification timelines and supply bottlenecks also differ by product.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Product&lt;/th&gt;
 &lt;th&gt;Policy Directness&lt;/th&gt;
 &lt;th&gt;Impact Channel&lt;/th&gt;
 &lt;th&gt;Primary Beneficiaries&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Mobile and PC DRAM&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;td&gt;Limits CXMT&amp;rsquo;s qualification with Apple and PC customers and global volume expansion&lt;/td&gt;
 &lt;td&gt;Samsung, Micron, some SK Hynix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server DRAM&lt;/td&gt;
 &lt;td&gt;Medium-High&lt;/td&gt;
 &lt;td&gt;Data center procurement restrictions block Chinese server memory entry&lt;/td&gt;
 &lt;td&gt;Samsung, SK Hynix, Micron&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM&lt;/td&gt;
 &lt;td&gt;Low near-term, Medium-High long-term&lt;/td&gt;
 &lt;td&gt;Limited near-term supply alternatives; reduces China&amp;rsquo;s post-2028 catch-up risk&lt;/td&gt;
 &lt;td&gt;SK Hynix, Samsung, Micron&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Consumer NAND&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;td&gt;Limits YMTC&amp;rsquo;s entry into global smartphone and PC storage&lt;/td&gt;
 &lt;td&gt;SanDisk, Kioxia, Samsung&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Enterprise SSD&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;td&gt;Likely falls directly within the U.S. data center procurement scope&lt;/td&gt;
 &lt;td&gt;SanDisk, Kioxia, Samsung, SK Hynix/Solidigm&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="hbm-multiple-rather-than-near-term-eps"&gt;HBM: Multiple Rather Than Near-Term EPS
&lt;/h3&gt;&lt;p&gt;CXMT is not currently a viable near-term substitute for leading-edge HBM from Samsung, SK Hynix, or Micron. Customer qualification, stacking yields, advanced packaging, interposers, and long-term supply agreements constitute high barriers to entry. Even if U.S. data centers were prohibited from purchasing CXMT HBM today, SK Hynix&amp;rsquo;s 2026 shipments would not increase immediately as a result.&lt;/p&gt;
&lt;p&gt;The calculus changes when projecting past 2028. The market discounts the risk that Chinese firms deploy cash earned from commodity DRAM and state subsidies to pursue HBM development and close the gap with leading players. If equipment controls and customer exclusion operate simultaneously, the pace of that catch-up and the scale at which it can be commercialized are reduced. This effect appears first in normalized P/E and terminal value rather than current-year EPS.&lt;/p&gt;
&lt;h3 id="more-direct-impact-on-mobile-dram-and-nand"&gt;More Direct Impact on Mobile DRAM and NAND
&lt;/h3&gt;&lt;p&gt;The Chinese alternatives Apple tested are closer to mobile and consumer memory. If CXMT establishes itself as a qualified supplier to Apple or global PC makers, the pricing leverage of the incumbent three erodes. If YMTC enters U.S. data center SSD and consumer device markets, the ceiling on NAND pricing falls as well.&lt;/p&gt;
&lt;p&gt;The letter is designed to block precisely these qualification pathways. Accordingly, by product portfolio, Samsung and Micron are more direct beneficiaries on DRAM, while SanDisk and Kioxia benefit more directly on NAND.&lt;/p&gt;
&lt;h2 id="5-impact-by-listed-company"&gt;5. Impact by Listed Company
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company / Group&lt;/th&gt;
 &lt;th&gt;Impact&lt;/th&gt;
 &lt;th&gt;Investment Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron &lt;code&gt;MU&lt;/code&gt;&lt;/td&gt;
 &lt;td&gt;Strongly positive&lt;/td&gt;
 &lt;td&gt;The only large-scale U.S. memory maker. Full coverage across DRAM, HBM, and NAND makes Micron the most direct beneficiary of security-driven procurement and commercial customer diversification. Whether elevated policy expectations are already in the valuation requires a separate check.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics &lt;code&gt;005930&lt;/code&gt;&lt;/td&gt;
 &lt;td&gt;Positive&lt;/td&gt;
 &lt;td&gt;Slower Chinese entry into mobile DRAM, commodity DRAM, and NAND is directly beneficial. Combined with an HBM recovery, earnings defensibility across the entire product portfolio improves.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix &lt;code&gt;000660&lt;/code&gt;&lt;/td&gt;
 &lt;td&gt;Positive&lt;/td&gt;
 &lt;td&gt;Long-term competitive risk in server DRAM and HBM diminishes. Near-term HBM volume upside is limited; the more significant effect is multiple defense for post-2028 earnings durability. Enterprise SSD exposure through Solidigm is also relevant.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SanDisk &lt;code&gt;SNDK&lt;/code&gt; / Kioxia&lt;/td&gt;
 &lt;td&gt;Positive&lt;/td&gt;
 &lt;td&gt;YMTC&amp;rsquo;s entry into U.S. and allied consumer NAND and enterprise SSD supply chains becomes more constrained. Favorable for NAND pricing and customer share retention.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple, PC OEMs, server OEMs&lt;/td&gt;
 &lt;td&gt;Negative&lt;/td&gt;
 &lt;td&gt;Supplier optionality and pricing leverage are reduced. Higher component costs must be absorbed through product pricing, specifications, or margins.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Applied Materials, Lam Research, KLA&lt;/td&gt;
 &lt;td&gt;Conditionally negative&lt;/td&gt;
 &lt;td&gt;Tighter equipment controls on CXMT and YMTC create headwinds for China revenue. Incremental capacity buildout in the U.S., South Korea, and Japan could offset some demand.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT / YMTC&lt;/td&gt;
 &lt;td&gt;Negative abroad, mixed domestically&lt;/td&gt;
 &lt;td&gt;International qualification and customer acquisition become harder. Domestic Chinese substitution, state subsidies, and localized equipment and materials transitions could accelerate.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="differentiating-samsung-and-sk-hynix"&gt;Differentiating Samsung and SK Hynix
&lt;/h3&gt;&lt;p&gt;This policy is marginally more direct for Samsung. Samsung&amp;rsquo;s exposure surface to Chinese competitors spans mobile DRAM, commodity DRAM, and NAND simultaneously. Blocking Chinese suppliers from global market entry defends both pricing and share across those product lines.&lt;/p&gt;
&lt;p&gt;SK Hynix also benefits. Its current earnings, however, are concentrated in HBM and server DRAM, where CXMT is not yet a credible near-term threat. Near-term consensus estimates do not change materially. Instead, the reduction in post-2028 HBM catch-up risk and commodity DRAM oversupply concern strengthens the case that elevated profitability can persist longer — an argument that applies more to the duration of normalized earnings than to the next quarter&amp;rsquo;s EPS.&lt;/p&gt;
&lt;p&gt;The letter alone is not sufficient grounds to reverse the relative priority between Samsung and SK Hynix. By product portfolio, it is a more direct earnings-defense catalyst for Samsung and more of a structural moat extension for SK Hynix&amp;rsquo;s HBM competitive position.&lt;/p&gt;
&lt;h2 id="6-the-global-memory-market-could-bifurcate"&gt;6. The Global Memory Market Could Bifurcate
&lt;/h2&gt;&lt;p&gt;Chinese production capacity does not disappear. What changes is the geography of permissible sales.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Market&lt;/th&gt;
 &lt;th&gt;Expected Structure&lt;/th&gt;
 &lt;th&gt;Pricing and Earnings Impact&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;United States and allies&lt;/td&gt;
 &lt;td&gt;Centered on qualified non-Chinese suppliers; stricter security procurement and origin rules&lt;/td&gt;
 &lt;td&gt;Shortage conditions persist; higher ASPs and margins become sustainable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China domestic&lt;/td&gt;
 &lt;td&gt;CXMT and YMTC localization accelerates; subsidies and transition to local equipment and materials&lt;/td&gt;
 &lt;td&gt;Market share expansion possible; medium-term oversupply and price competition risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Emerging markets&lt;/td&gt;
 &lt;td&gt;Potential acceptance of Chinese memory as cost and regulation trade-offs are weighed&lt;/td&gt;
 &lt;td&gt;Could become a bypass sales channel for Chinese producers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This bifurcation is not uniformly positive for Samsung, SK Hynix, and Micron. In the U.S. and allied markets, pricing and margins may improve. In Chinese customer markets, however, the transition to domestic suppliers could accelerate. The net effect is a &lt;strong&gt;trade-off between higher global margins and a smaller China addressable market&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;If Chinese volumes concentrate domestically and in emerging markets, regional price divergence for the same DRAM and NAND products could widen. U.S. and allied buyers pay a premium for qualified suppliers, while China&amp;rsquo;s domestic capacity additions drive price competition internally. The premise that global memory prices move as one unified market weakens.&lt;/p&gt;
&lt;h2 id="7-share-prices-may-reflect-this-through-discount-rates-before-eps"&gt;7. Share Prices May Reflect This Through Discount Rates Before EPS
&lt;/h2&gt;&lt;p&gt;Memory equity valuations are not determined by the next quarter&amp;rsquo;s earnings alone. Investors simultaneously assess how long cyclical peak earnings persist, how quickly new supply arrives, and where normalized margins ultimately settle.&lt;/p&gt;
&lt;p&gt;Chinese capacity additions compress valuations through three channels.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;They create concern that commodity DRAM and NAND prices could fall sharply in 2027–2028.&lt;/li&gt;
&lt;li&gt;They raise the probability that state-backed late entrants can challenge incumbents in HBM.&lt;/li&gt;
&lt;li&gt;They give customers leverage to use a fourth supplier as a pricing card in negotiations.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;If procurement restrictions and equipment controls are actually implemented, all three risks diminish. The effect may therefore appear in the duration of normalized earnings and target multiples before it shows up in next-quarter EPS revisions.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Policy not enacted
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Chinese capacity additions flow into global supply
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ 2027–2028 price decline risk persists
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Normalized earnings and P/E discounted
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Policy enacted
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Barrier between Chinese capacity and global customers
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Speed of oversupply transmission slows
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Earnings duration revised upward
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ China oversupply discount compresses
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Chinese production capacity does not vanish entirely. If China&amp;rsquo;s domestic price declines spill over into emerging markets, or if non-U.S. equipment ecosystems mature faster than expected, discount rates could widen again.&lt;/p&gt;
&lt;h2 id="8-three-policy-scenarios"&gt;8. Three Policy Scenarios
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;Development&lt;/th&gt;
 &lt;th&gt;Impact on Memory Equities&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1. Letter only&lt;/td&gt;
 &lt;td&gt;CXMT review delayed; no separate purchase ban; no allied coordination&lt;/td&gt;
 &lt;td&gt;Near-term news effect fades. Existing China oversupply risk intact.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2. U.S.-only partial implementation&lt;/td&gt;
 &lt;td&gt;CXMT added to Entity List; procurement restrictions applied to AI, data centers, and federal IT&lt;/td&gt;
 &lt;td&gt;Direct benefit for Micron. Korean companies also defend U.S. customer share and pricing.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3. Allied joint implementation&lt;/td&gt;
 &lt;td&gt;U.S. actions joined by South Korea, Japan, and EU; customer qualification and equipment/materials channels blocked simultaneously&lt;/td&gt;
 &lt;td&gt;Deep global supply chain bifurcation. Most positive for long-term multiples at the incumbent three and leading NAND players.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The third scenario is the most consequential variable. If only the United States restricts access while China successfully enters South Korean, Japanese, and European supply chains, the global pricing defense is limited. Conversely, if allied nations apply equivalent origin and security standards, CXMT and YMTC capacity additions — however large — struggle to translate into global customer revenue.&lt;/p&gt;
&lt;h2 id="9-catalysts-to-monitor"&gt;9. Catalysts to Monitor
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item to Confirm&lt;/th&gt;
 &lt;th&gt;Significance&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;BIS CXMT Entity List review or Federal Register notice&lt;/td&gt;
 &lt;td&gt;Confirms whether supply-side controls are actually initiated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Commerce Department or White House U.S. corporate purchase restriction&lt;/td&gt;
 &lt;td&gt;Confirms whether demand channels for finished Chinese memory products are closed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple&amp;rsquo;s CXMT or YMTC qualification or procurement halt&lt;/td&gt;
 &lt;td&gt;Confirms whether policy pressure reaches consumer electronics at scale&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;South Korea, Japan, EU coordination announcements&lt;/td&gt;
 &lt;td&gt;Confirms whether Chinese capacity is actually blocked from flowing into global pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 DRAM and NAND contract prices&lt;/td&gt;
 &lt;td&gt;Confirms whether fundamentals — not just policy expectations — support earnings estimates&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT and YMTC domestic capacity additions and local equipment transition&lt;/td&gt;
 &lt;td&gt;Indicates how long equipment controls suppress supply&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Our current stance is &lt;strong&gt;Watchlist — incorporate upon policy confirmation&lt;/strong&gt;. The letter alone does not justify chasing memory equities. Until Commerce Department actions materialize, this is more appropriately used as a basis for reweighting risk scenarios than for revising EPS models upward.&lt;/p&gt;
&lt;h2 id="10-bear-case-and-invalidation-conditions"&gt;10. Bear Case and Invalidation Conditions
&lt;/h2&gt;&lt;h3 id="macro-failure-conditions"&gt;Macro Failure Conditions
&lt;/h3&gt;&lt;p&gt;If memory prices rise too quickly, smartphone, PC, and server customers may reduce shipments or trade down on specifications. A policy that protects suppliers could produce demand destruction, with the impact of declining unit volumes exceeding the benefit of higher ASPs. An AI investment slowdown combined with higher interest rates would not spare server memory.&lt;/p&gt;
&lt;h3 id="company-level-failure-conditions"&gt;Company-Level Failure Conditions
&lt;/h3&gt;&lt;p&gt;If Chinese memory concentrates in China and emerging markets rather than the United States, it could accelerate the erosion of Samsung, SK Hynix, and Micron&amp;rsquo;s China customer base. If domestic Chinese demand is large enough and the local equipment and materials ecosystem matures, the supply-suppression effect of the Entity List could prove shorter-lived than expected.&lt;/p&gt;
&lt;h3 id="policy-invalidation-conditions"&gt;Policy Invalidation Conditions
&lt;/h3&gt;&lt;p&gt;The thesis weakens materially if the following combination emerges:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;CXMT&amp;rsquo;s Entity List designation is delayed indefinitely.&lt;/li&gt;
&lt;li&gt;The United States grants broad exceptions for Apple&amp;rsquo;s China-market consumer products.&lt;/li&gt;
&lt;li&gt;The procurement ban remains limited to federal agencies and does not apply to private AI and data center operators.&lt;/li&gt;
&lt;li&gt;South Korea, Japan, and the EU do not coordinate.&lt;/li&gt;
&lt;li&gt;CXMT and YMTC transition rapidly to non-U.S. equipment, sustaining volume and yields.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Conversely, if CXMT&amp;rsquo;s designation, private-sector data center procurement restrictions, and allied coordination materialize together, this letter graduates from political signaling to a structural variable in the memory cycle.&lt;/p&gt;
&lt;h2 id="11-final-assessment"&gt;11. Final Assessment
&lt;/h2&gt;&lt;p&gt;The July 16 letter is an official document with specific demands. It is not, however, an enacted Chinese memory embargo. Near-term equity catalysts therefore require policy confirmation.&lt;/p&gt;
&lt;p&gt;The investment case is clear. If the United States and its allies choose not to fill memory shortfalls with Chinese supply, Samsung, SK Hynix, and Micron can defend market share and pricing leverage for longer. SanDisk and Kioxia also benefit directly from YMTC&amp;rsquo;s restricted path into NAND and enterprise SSD.&lt;/p&gt;
&lt;p&gt;By product, the impact is more direct for Samsung&amp;rsquo;s mobile DRAM, commodity DRAM, and NAND; for SK Hynix the more significant effect is an extended duration for its HBM competitive advantage. Micron, as the only large-scale U.S. memory maker, is the clearest and most straightforward policy beneficiary.&lt;/p&gt;
&lt;p&gt;In one sentence: &lt;strong&gt;This letter is not news that raises 2026 HBM shipments — it is a policy signal that reduces the risk of Chinese-origin oversupply eroding global memory profitability in 2027–2028.&lt;/strong&gt; Until concrete regulations emerge, it is more accurately read as a basis for recalibrating discount rates than as a catalyst to chase.&lt;/p&gt;
&lt;h2 id="evidence-classification"&gt;Evidence Classification
&lt;/h2&gt;&lt;h3 id="confirmed-facts"&gt;Confirmed Facts
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;On July 16, 2026, Representatives John Moolenaar and George Whitesides sent an official letter to the Secretary of Commerce.&lt;/li&gt;
&lt;li&gt;The letter calls for strengthened controls on YMTC, accelerated Entity List review for CXMT, specific U.S. procurement restrictions, and coordination with South Korea, Japan, and the EU.&lt;/li&gt;
&lt;li&gt;The Entity List restricts the export, re-export, and in-country transfer of items subject to U.S. Export Administration Regulations.&lt;/li&gt;
&lt;li&gt;TrendForce projected Q2 commodity DRAM contract prices up 58–63% and NAND up 70–75% quarter-over-quarter.&lt;/li&gt;
&lt;li&gt;TrendForce projected Q3 server DRAM up 13–18% and flagged potential supply tightness into 2027.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="analysis-and-inference"&gt;Analysis and Inference
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;If enacted, the greatest effect is a reduction in the 2027–2028 China oversupply risk and a compression of the normalized earnings discount rate.&lt;/li&gt;
&lt;li&gt;Samsung faces more direct exposure on mobile DRAM, commodity DRAM, and NAND; for SK Hynix the more important consideration is reduced long-term competitive risk in HBM.&lt;/li&gt;
&lt;li&gt;A bifurcated market structure — high prices in the U.S. and allied markets alongside domestic price competition within China — may emerge.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="unconfirmed-items"&gt;Unconfirmed Items
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Whether and when CXMT will be added to the Entity List&lt;/li&gt;
&lt;li&gt;The legal form, scope, and exceptions of any separate purchase prohibition&lt;/li&gt;
&lt;li&gt;Apple&amp;rsquo;s actual order volume from CXMT and YMTC and allocation by supplier&lt;/li&gt;
&lt;li&gt;Whether South Korea, Japan, and the EU will coordinate&lt;/li&gt;
&lt;li&gt;The specific EPS uplift for Samsung, SK Hynix, and Micron if restrictions are implemented&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="key-sources"&gt;Key Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://chinaselectcommittee.house.gov/media/letters/moolenaar-whitesides-to-secretary-lutnick-hold-firm-on-chinese-memory-chips-ban" target="_blank" rel="noopener"
 &gt;House Select Committee on China: Official Moolenaar–Whitesides Letter&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.bis.gov/ear/title-15/subtitle-b/chapter-vii/subchapter-c/part-744/ss-74416-entity-list" target="_blank" rel="noopener"
 &gt;U.S. Commerce Department BIS: Entity List Regulations&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260331-12995.html" target="_blank" rel="noopener"
 &gt;TrendForce: Q2 2026 DRAM and NAND Price Outlook&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260709-13140.html" target="_blank" rel="noopener"
 &gt;TrendForce: Q3 2026 Server DRAM Outlook&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;This article is provided for informational purposes based on publicly available sources and does not constitute a recommendation to buy or sell any specific security.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>Kimi K3 Resets the AI Price Curve: From Kimi Linear to HBM and Big Tech Strategy</title><link>https://koreainvestinsights.com/post/kimi-k3-linear-api-pricing-semiconductor-big-tech-impact-2026-07-17/</link><pubDate>Fri, 17 Jul 2026 12:31:36 +0900</pubDate><guid>https://koreainvestinsights.com/post/kimi-k3-linear-api-pricing-semiconductor-big-tech-impact-2026-07-17/</guid><description>&lt;p&gt;Kimi K3 launched on July 16, 2026 at $3 per million uncached input tokens and $15 per million output tokens. That is not the familiar ultra-low-price positioning of a Chinese challenger. It matches Claude Sonnet 5&amp;rsquo;s standard price and sits 40% below GPT-5.6 Sol on input and 50% below it on output. Moonshot AI is positioning K3 as a primary enterprise model, not a budget fallback.&lt;/p&gt;
&lt;p&gt;The architecture is designed to support that claim. K3 has 2.8 trillion total parameters but effectively activates 16 of 896 experts per token. Kimi Delta Attention controls the cost of long context, Attention Residuals selectively retrieve earlier representations across depth, and quantization-aware training uses MXFP4 weights with MXFP8 activations. Moonshot recommends a supernode with at least 64 accelerators.&lt;/p&gt;
&lt;p&gt;That combination creates a two-sided semiconductor outcome. Memory and compute per token can fall while the number of institutions able to deploy a frontier-class model, and the amount of work they run, can rise. K3 is both an efficiency technology and an infrastructure workload.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Related reading: &lt;a class="link" href="https://koreainvestinsights.com/post/ai-token-value-memory-value-added-2026-07-09/" &gt;AI token value and memory value capture&lt;/a&gt; / &lt;a class="link" href="https://koreainvestinsights.com/post/ai-token-futures-cost-per-token-korea-semiconductor-thesis-2026-05-30/" &gt;AI token futures and cost per token&lt;/a&gt; / &lt;a class="link" href="https://koreainvestinsights.com/post/us-china-agentic-inference-stack-sram-opportunity-2026-07-09/" &gt;US-China divergence in agentic inference infrastructure&lt;/a&gt; / &lt;a class="link" href="https://koreainvestinsights.com/post/hbm-2030-supply-demand-267eb-demand-model-crosscheck-2026-07-13/" &gt;Cross-checking the 2030 HBM shortage model&lt;/a&gt;&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="executive-summary"&gt;Executive Summary
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;K3 combines 2.8T parameters, native vision and a 1M-token context window. Its products and API are live, but as of July 17 the full weights, technical report and license have not been released. The open-weight thesis must be verified after the July 27 deadline.&lt;/li&gt;
&lt;li&gt;Moonshot&amp;rsquo;s official GDPval-AA v2 score is 1,668, not 1,687. AA-Briefcase is 1,548. BrowseComp 91.2 uses context compaction; the no-compaction 1M-context result is 90.4. The results are strong, but mixed harnesses and company-run evaluations require independent replication.&lt;/li&gt;
&lt;li&gt;Kimi Linear&amp;rsquo;s claims of up to 75% lower KV cache and up to 6.3x theoretical decoding throughput at 1M context come from a 48B-total, 3B-active research model. They should not be presented as measured K3 API performance.&lt;/li&gt;
&lt;li&gt;K3 is not a low-cost model. It matches Sonnet 5&amp;rsquo;s standard price, is 50% more expensive than Sonnet&amp;rsquo;s temporary launch promotion, and is cheaper than GPT-5.6 Sol. Because only max reasoning is currently available and independent tests show high output-token use, cost per completed task may be less attractive than the rate card implies.&lt;/li&gt;
&lt;li&gt;The semiconductor impact pits lower compute and KV cache per request against more self-hosted deployments and higher total workload. Server DRAM and enterprise SSDs are the cleanest second-order beneficiaries because long-lived agent context spills below HBM into disaggregated cache tiers.&lt;/li&gt;
&lt;li&gt;The model and cloud layers experience opposite economics. OpenAI, Anthropic and Gemini API pricing face pressure, while Azure, AWS and Google Cloud can monetize K3 and other models through compute, storage and networking. Meta must defend US open-weight leadership.&lt;/li&gt;
&lt;li&gt;The July 27 proof points are the license, full weights, external evaluation, throughput on NVIDIA and AMD, support on Chinese accelerators, actual output tokens per task, vLLM compatibility and cloud catalog adoption.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;&lt;img alt="Kimi K3 pricing strategy and AI infrastructure impact map" loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/images/posts/kimi-k3-pricing-infrastructure-impact-2026-07-17.png"&gt;
&lt;/p&gt;
&lt;h2 id="1-correcting-the-numbers-before-drawing-conclusions"&gt;1. Correcting the Numbers Before Drawing Conclusions
&lt;/h2&gt;&lt;p&gt;Several numbers changed as the launch circulated through social media. The official values matter because some differences alter the investment interpretation.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Official value&lt;/th&gt;
 &lt;th&gt;Investor interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Total parameters&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.8T&lt;/td&gt;
 &lt;td&gt;Total model size, not per-token active compute&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Expert routing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16 of 896&lt;/td&gt;
 &lt;td&gt;Extremely sparse MoE; routing and communication become first-order constraints&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Context&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1M tokens&lt;/td&gt;
 &lt;td&gt;Useful for repositories and research, but task cost depends on output length and cache hits&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GDPval-AA v2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,668&lt;/td&gt;
 &lt;td&gt;The official table does not show 1,687&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AA-Briefcase&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,548&lt;/td&gt;
 &lt;td&gt;Above GPT-5.6 Sol at 1,495, below Claude Fable 5 at 1,583&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;BrowseComp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;91.2&lt;/td&gt;
 &lt;td&gt;Uses compaction starting at 300K tokens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;BrowseComp without compaction&lt;/td&gt;
 &lt;td style="text-align: right"&gt;90.4&lt;/td&gt;
 &lt;td&gt;The cleaner result for the native 1M-context claim&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Product availability&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Web, Work, Code and API live&lt;/td&gt;
 &lt;td&gt;Commercial testing can begin now&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Weights&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Promised by July 27&lt;/td&gt;
 &lt;td&gt;License and complete artifacts remain unverified as of July 17&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reasoning effort&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Max only&lt;/td&gt;
 &lt;td&gt;Low-cost modes are not yet available&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Moonshot explicitly states that K3 still trails Claude Fable 5 and GPT-5.6 Sol in overall user experience. At the same time, it reports frontier-level performance across coding, knowledge work and multimodal benchmarks. The credible interpretation is not that China has conclusively taken the lead. It is that a Chinese model has entered the performance band immediately below the strongest proprietary systems while promising a 1M context and downloadable weights.&lt;/p&gt;
&lt;p&gt;The benchmark table is not a single controlled tournament. K3 runs at maximum reasoning effort. Depending on the task, models use Kimi Code, Claude Code or Codex, and some competitor scores are the best results across harnesses or are imported from external leaderboards. Independent reproduction remains necessary.&lt;/p&gt;
&lt;p&gt;Still, Terminal-Bench 2.1 at 88.3, FrontierSWE at 81.2, SWE Marathon at 42.0, AutomationBench at 30.8, GPQA-Diamond at 93.5 and MMMU-Pro at 81.6 indicate that K3 is designed for long-horizon tool use and coding, not merely chat. The more important signal is the package: near-frontier capability, 1M context and promised open weights.&lt;/p&gt;
&lt;h2 id="2-how-a-28t-model-becomes-deployable"&gt;2. How a 2.8T Model Becomes Deployable
&lt;/h2&gt;&lt;h3 id="21-total-parameters-are-not-active-parameters"&gt;2.1 Total parameters are not active parameters
&lt;/h3&gt;&lt;p&gt;Computing all 2.8T parameters for every token would be prohibitively expensive. Stable LatentMoE effectively activates 16 of 896 experts, or about 1.8% of the expert pool. The technical report is needed to establish exact active parameter counts, but total parameters clearly do not equal per-token compute.&lt;/p&gt;
&lt;p&gt;Sparse MoE shifts rather than eliminates bottlenecks.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;What improves&lt;/th&gt;
 &lt;th&gt;What becomes harder&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Active compute per token&lt;/td&gt;
 &lt;td&gt;Router quality and expert balance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Compute needed for a quality target&lt;/td&gt;
 &lt;td&gt;Communication across accelerators&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Some inference costs&lt;/td&gt;
 &lt;td&gt;Avoiding idle accelerators and hot experts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Scaling model capacity&lt;/td&gt;
 &lt;td&gt;Keeping the full weight set accessible&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is why Moonshot recommends a supernode with 64 or more accelerators. Even when only a small subset of experts is active, the selected expert is not known in advance and the full model must remain accessible. At four bits, 2.8T weights imply a theoretical minimum of about 1.4 TB before scales, metadata, buffers, cache and replication. Sparse activation reduces arithmetic but does not remove model-residency memory or fabric requirements.&lt;/p&gt;
&lt;h3 id="22-kimi-linear-addresses-long-context-cost"&gt;2.2 Kimi Linear addresses long-context cost
&lt;/h3&gt;&lt;p&gt;The Kimi Linear paper predates K3 and evaluates a 48B-total, 3B-active research model, not K3 itself. It combines Kimi Delta Attention with full Multi-head Latent Attention in a 3:1 ratio.&lt;/p&gt;
&lt;p&gt;Full attention is strong at exact copying and fine-grained retrieval, but KV cache grows with context. Linear attention compresses history into a fixed-size state, reducing sequence-length dependence, but can lose exact detail. Kimi Linear uses three KDA layers followed by one full-attention layer to balance efficiency and expressiveness.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Component&lt;/th&gt;
 &lt;th&gt;Role&lt;/th&gt;
 &lt;th&gt;Trade-off&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KDA&lt;/td&gt;
 &lt;td&gt;Compress long history into fixed-size state&lt;/td&gt;
 &lt;td&gt;Can weaken exact copying and fine retrieval&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Full MLA&lt;/td&gt;
 &lt;td&gt;Restores precise token-to-token recall&lt;/td&gt;
 &lt;td&gt;KV cache still grows with context&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3:1 hybrid&lt;/td&gt;
 &lt;td&gt;Balances efficiency and quality&lt;/td&gt;
 &lt;td&gt;Requires more complex kernels and serving software&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The paper reports up to 75% lower KV cache and up to 6.3x theoretical decoding throughput at 1M context. A measured comparison in the complexity analysis reports 2.3x acceleration. These results establish direction, not guaranteed K3 API performance.&lt;/p&gt;
&lt;p&gt;For investors, lower KV cache means more concurrent requests per accelerator. It also makes longer repositories, more documents and persistent agents economical. Memory per token can fall while total tokens rise.&lt;/p&gt;
&lt;h3 id="23-attention-residuals-improve-depth-efficiency"&gt;2.3 Attention Residuals improve depth efficiency
&lt;/h3&gt;&lt;p&gt;Standard residual connections keep adding earlier layer outputs. In very deep networks, useful representations can be diluted. Attention Residuals allow a layer to select the earlier representations it needs. Block AttnRes groups layers and retains block-level representations, reducing memory overhead.&lt;/p&gt;
&lt;p&gt;Moonshot&amp;rsquo;s research says roughly eight blocks recover most of the gains, and Block AttnRes can match a baseline trained with about 1.25x compute. This improves training capital efficiency, but it does not automatically reduce data-center demand. Better efficiency can be reinvested into larger models and longer tasks.&lt;/p&gt;
&lt;h3 id="24-quantization-is-a-hardware-portability-strategy"&gt;2.4 Quantization is a hardware-portability strategy
&lt;/h3&gt;&lt;p&gt;K3 applies quantization-aware training from supervised fine-tuning onward, using MXFP4 weights and MXFP8 activations. This should control accuracy loss better than aggressive post-training quantization and make deployment across multiple hardware platforms easier.&lt;/p&gt;
&lt;p&gt;Moonshot&amp;rsquo;s kernel arena included NVIDIA H200 and a general-purpose GPU from an alternative vendor. The official post does not name a Chinese chip or claim H200-equivalent economics. What is verified is the strategic intent to optimize outside NVIDIA as well as on NVIDIA.&lt;/p&gt;
&lt;p&gt;That matters for both China and global buyers. China needs frontier-level models that can survive constrained access to top NVIDIA systems. Other buyers want bargaining power across NVIDIA, AMD and custom accelerators.&lt;/p&gt;
&lt;h2 id="3-what-the-price-card-reveals"&gt;3. What the Price Card Reveals
&lt;/h2&gt;&lt;h3 id="31-k3-is-priced-as-a-primary-model"&gt;3.1 K3 is priced as a primary model
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Model&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Cached input per 1M&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Standard input&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Output&lt;/th&gt;
 &lt;th&gt;Current positioning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Kimi K3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$0.30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$15&lt;/td&gt;
 &lt;td&gt;Frontier primary-model pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Claude Sonnet 5 launch promo&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Varies&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$10&lt;/td&gt;
 &lt;td&gt;Temporary through August 31&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Claude Sonnet 5 standard&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Varies&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$15&lt;/td&gt;
 &lt;td&gt;Same headline price as K3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPT-5.6 Sol&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$0.50&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$30&lt;/td&gt;
 &lt;td&gt;67% higher input and 100% higher output than K3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;K3 is more expensive than Sonnet 5&amp;rsquo;s current promotion. It is inaccurate to describe it as universally half-price. The clearer strategy is to match Sonnet&amp;rsquo;s standard tier while undercutting the most expensive frontier API.&lt;/p&gt;
&lt;p&gt;The price is both a confidence signal and a monetization decision. Moonshot is no longer accepting a deep discount simply to acquire usage. It is claiming a position in the enterprise default-model tier.&lt;/p&gt;
&lt;h3 id="32-cost-per-completed-task-matters-more-than-cost-per-token"&gt;3.2 Cost per completed task matters more than cost per token
&lt;/h3&gt;&lt;p&gt;Rate cards assume equal token use. Agents differ in planning length, tool calls, retries and verbosity. K3 currently exposes only max reasoning. Artificial Analysis reports that K3 used about 130 million output tokens in its Intelligence Index evaluation, more than twice the peer median of roughly 63 million. A cheaper output token can still lead to a costly completed task.&lt;/p&gt;
&lt;p&gt;&lt;code&gt;Task cost = input tokens × input rate + output tokens × output rate + tool and retry cost&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;Enterprise buyers will monitor task success, output tokens, time to first token, throughput, tool reliability, cache hit rate and session stability. Moonshot says Mooncake delivers more than 90% cache hits on coding workloads, making cached input one-tenth the uncached price. If that rate holds on enterprise traffic, K3&amp;rsquo;s effective economics improve materially.&lt;/p&gt;
&lt;h3 id="33-mooncake-moves-memory-demand-down-the-hierarchy"&gt;3.3 Mooncake moves memory demand down the hierarchy
&lt;/h3&gt;&lt;p&gt;Mooncake separates prefill and decode clusters and disaggregates KV cache across CPU, DRAM and SSD rather than keeping everything in GPU HBM. Its paper reports up to 525% higher throughput in simulation and 75% more requests on production workloads.&lt;/p&gt;
&lt;p&gt;This explains why AI memory demand extends beyond HBM.&lt;/p&gt;
&lt;p&gt;&lt;code&gt;Accelerator HBM → server DRAM → enterprise SSD → remote cache tier&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;KDA can reduce KV cache per request, yet persistent 1M-context agents increase total cache volume and retention time. Hot data remains in HBM and DRAM; colder context moves to SSD. Efficiency can soften an HBM-only thesis while strengthening the broader memory hierarchy.&lt;/p&gt;
&lt;h2 id="4-chinese-open-weights-move-up-the-enterprise-stack"&gt;4. Chinese Open Weights Move Up the Enterprise Stack
&lt;/h2&gt;&lt;p&gt;OpenRouter reports that Chinese models surpassed US models in token share on its platform in early June 2026, based on more than 450 trillion tokens from January through June 14. DeepSeek&amp;rsquo;s share rose from roughly 9% to 18% and it became the leading provider by mid-May.&lt;/p&gt;
&lt;p&gt;The adoption path has three stages:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Low-cost open models take classification, translation and testing workloads.&lt;/li&gt;
&lt;li&gt;Better coding and agent performance take repetitive enterprise workflows.&lt;/li&gt;
&lt;li&gt;Near-frontier quality and million-token context compete for primary routing.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;K3&amp;rsquo;s price is designed for the third stage. It is not following the most aggressive low-price strategy. It is charging a frontier-tier API price while promising weights that clouds, governments and enterprises can deploy themselves.&lt;/p&gt;
&lt;p&gt;Proprietary APIs and open weights accumulate different strategic assets.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Proprietary frontier API&lt;/th&gt;
 &lt;th&gt;Near-frontier open weights&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Controls the best UX and capability&lt;/td&gt;
 &lt;td&gt;Multiplies deployment routes and hardware options&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Centralizes usage data&lt;/td&gt;
 &lt;td&gt;Lets enterprises retain data and operations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Changes pricing and policy centrally&lt;/td&gt;
 &lt;td&gt;Released files are difficult to withdraw&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Model provider captures gross margin&lt;/td&gt;
 &lt;td&gt;Cloud, chip and application vendors share value&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The strongest proprietary model can remain number one while losing paid token share. Enterprises can route repetitive work to K3-class models and reserve the most difficult legal, design and research tasks for the top closed model. Paid token mix and blended price matter more than leaderboard rank.&lt;/p&gt;
&lt;h2 id="5-semiconductor-demand-efficiency-and-diffusion-arrive-together"&gt;5. Semiconductor Demand: Efficiency and Diffusion Arrive Together
&lt;/h2&gt;&lt;h3 id="51-nvidia-near-term-efficiency-risk-medium-term-deployment-elasticity"&gt;5.1 NVIDIA: near-term efficiency risk, medium-term deployment elasticity
&lt;/h3&gt;&lt;p&gt;Sparse MoE, KDA, quantization and autonomous kernel optimization reduce GPU time per task. Hardware portability also weakens the assumption that every frontier workload must stay on CUDA. Those are negative valuation signals for NVIDIA.&lt;/p&gt;
&lt;p&gt;The positive side is equally material. Moonshot recommends at least 64 accelerators for self-hosted K3. Released weights could drive clouds, governments, laboratories and large enterprises to build their own clusters. Demand concentrated behind one API becomes hardware demand across many data centers.&lt;/p&gt;
&lt;p&gt;&lt;code&gt;Total accelerator demand = lower GPU time per task × higher total workload × more self-hosting institutions&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;The first term is negative; the last two are positive. K3 alone is not enough to cut NVIDIA earnings estimates, but neither is it enough to assume that every efficiency gain creates more GPU demand. Workload elasticity is the deciding variable.&lt;/p&gt;
&lt;h3 id="52-amd-optionality-from-hardware-choice"&gt;5.2 AMD: optionality from hardware choice
&lt;/h3&gt;&lt;p&gt;AMD benefits if MXFP4, MXFP8 and vLLM portability allow enterprises to separate model choice from accelerator choice. A near-frontier open model gives buyers a realistic workload on which to test NVIDIA alternatives.&lt;/p&gt;
&lt;p&gt;The proof must come after the weights. ROCm kernels, expert-parallel communication, 1M-context throughput and 64-accelerator stability need to be measured. AMD&amp;rsquo;s upside grows only if K3 demonstrates superior cost per successful task on MI systems.&lt;/p&gt;
&lt;h3 id="53-hbm-lower-per-request-cache-larger-model-residency"&gt;5.3 HBM: lower per-request cache, larger model residency
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Demand layer&lt;/th&gt;
 &lt;th&gt;K3 efficiency impact&lt;/th&gt;
 &lt;th&gt;Direction&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Weight residency&lt;/td&gt;
 &lt;td&gt;A 2.8T model must remain quickly accessible&lt;/td&gt;
 &lt;td&gt;More accelerators and high-bandwidth memory&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Per-request KV cache&lt;/td&gt;
 &lt;td&gt;KDA lowers cache size and growth&lt;/td&gt;
 &lt;td&gt;Less HBM capacity per request&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Concurrent users and agents&lt;/td&gt;
 &lt;td&gt;Lower cost and open deployment can expand workloads&lt;/td&gt;
 &lt;td&gt;More total HBM and DRAM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The headlines can be negative for SK hynix, Micron and Samsung because 75% lower KV cache and 6.3x throughput sound like lower memory intensity. Those numbers belong to a Kimi Linear research model, not measured K3 production. HBM also stores weights, activations, communication buffers and batches, not only KV cache.&lt;/p&gt;
&lt;p&gt;The medium-term balance is neutral to positive if self-hosted clusters and agent workloads grow faster than efficiency. It turns negative if workload elasticity is weak and compression improves faster than usage.&lt;/p&gt;
&lt;h3 id="54-server-dram-and-enterprise-ssds-are-the-clearest-second-order-beneficiaries"&gt;5.4 Server DRAM and enterprise SSDs are the clearest second-order beneficiaries
&lt;/h3&gt;&lt;p&gt;Long context and cache reuse cannot remain entirely in HBM. Once serving separates prefill and decode and spills cache across CPU, DRAM and SSD, server DRAM and enterprise SSD become operating assets for AI inference.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Samsung can sell HBM, server DRAM, high-capacity SSDs, foundry and packaging.&lt;/li&gt;
&lt;li&gt;SK hynix combines HBM and server DRAM with Solidigm enterprise SSDs.&lt;/li&gt;
&lt;li&gt;Micron supplies HBM, data-center DRAM and SSDs.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;K3 does not show that AI needs less memory. It shows that AI memory is becoming hierarchical. The hottest data stays in HBM, retained context sits in server DRAM, and colder cache moves to enterprise SSD. Vendors with a full memory stack have greater resilience than a single-product HBM thesis.&lt;/p&gt;
&lt;h3 id="55-networking-and-custom-silicon-are-the-hidden-moe-bottlenecks"&gt;5.5 Networking and custom silicon are the hidden MoE bottlenecks
&lt;/h3&gt;&lt;p&gt;Spreading 896 experts across accelerators creates variable communication depending on token routing. This is why Moonshot emphasizes balanced expert-parallel training, static shapes and removal of host synchronization from the critical path. Efficient operation across 64 or more accelerators requires high-bandwidth scale-up and scale-out fabric.&lt;/p&gt;
&lt;p&gt;That is structurally positive for Broadcom, Marvell, NVIDIA networking, optical interconnect and switch silicon. It also encourages inference ASICs optimized for open models. K3&amp;rsquo;s 48-hour small-chip design exercise is not a commercial product, but it illustrates faster model-hardware co-design.&lt;/p&gt;
&lt;h2 id="6-us-big-tech-strategy-and-stock-transmission"&gt;6. US Big Tech Strategy and Stock Transmission
&lt;/h2&gt;&lt;h3 id="61-microsoft-pressure-on-openai-economics-more-azure-usage"&gt;6.1 Microsoft: pressure on OpenAI economics, more Azure usage
&lt;/h3&gt;&lt;p&gt;Microsoft owns exposure to OpenAI IP and economics as well as Azure infrastructure. The April 2026 partnership update keeps Microsoft as a primary cloud partner and extends a non-exclusive IP license through 2032.&lt;/p&gt;
&lt;p&gt;K3 pressures the model layer because repetitive work can move away from OpenAI APIs. It can support the cloud layer if Azure hosts K3 as managed or self-hosted infrastructure.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Microsoft layer&lt;/th&gt;
 &lt;th&gt;K3 impact&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;OpenAI revenue share&lt;/td&gt;
 &lt;td&gt;Negative through price and mix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Copilot&lt;/td&gt;
 &lt;td&gt;Positive if routing costs fall&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Azure AI&lt;/td&gt;
 &lt;td&gt;Positive if multi-model usage rises&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Custom silicon and data centers&lt;/td&gt;
 &lt;td&gt;Positive if open-weight optimization expands&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The near-term stock impact is neutral. The medium-term question is whether Azure converts model price deflation into usage and Copilot margin.&lt;/p&gt;
&lt;h3 id="62-amazon-anthropic-and-aws-have-different-economics"&gt;6.2 Amazon: Anthropic and AWS have different economics
&lt;/h3&gt;&lt;p&gt;Amazon is a major Anthropic investor and the provider of AWS, Bedrock, Trainium and Inferentia. K3 can reduce Claude pricing power and the value of Amazon&amp;rsquo;s Anthropic stake. If enterprises run K3 on AWS, however, EC2, Bedrock, storage, networking and Trainium usage can rise.&lt;/p&gt;
&lt;p&gt;Amazon&amp;rsquo;s optimal strategy is to keep the workload on AWS regardless of which model wins. If K3 arrives with a permissive license and efficient Trainium support, AWS can monetize a competitor&amp;rsquo;s diffusion. The stock impact is neutral to modestly positive, although inference price competition could lower cloud margins even as revenue grows.&lt;/p&gt;
&lt;h3 id="63-alphabet-gemini-pricing-pressure-tpu-and-vertex-defense"&gt;6.3 Alphabet: Gemini pricing pressure, TPU and Vertex defense
&lt;/h3&gt;&lt;p&gt;Google controls a model, a chip, a deployment platform and final demand through Search, Ads and Workspace. Vertex Model Garden supports first-party, third-party and open models.&lt;/p&gt;
&lt;p&gt;K3 pressures Gemini API pricing but can create TPU and Google Cloud demand. Lower model cost also reduces the expense of AI Overviews, ad generation and Workspace agents. The stock impact is neutral to positive because Google is not solely a model vendor. The risk is that Gemini loses both performance and price leadership, raising cloud customer-acquisition cost.&lt;/p&gt;
&lt;h3 id="64-meta-from-open-weight-beneficiary-to-defender"&gt;6.4 Meta: from open-weight beneficiary to defender
&lt;/h3&gt;&lt;p&gt;Meta benefited from open-weight diffusion by commoditizing competitors&amp;rsquo; APIs and lowering its own recommendation, advertising and content costs. If Chinese labs release near-frontier capability and longer context first, Meta risks losing its position as the benchmark US open-weight ecosystem.&lt;/p&gt;
&lt;p&gt;K3 forces two responses: the next Llama must compete on long context, agent tools and deployment cost, and Meta must provide a trusted US alternative for enterprises and governments unwilling to deploy Chinese weights. The near-term earnings impact is limited because advertising drives cash flow. The strategic risk is lower returns on AI infrastructure and developer ecosystem investment if Llama falls behind.&lt;/p&gt;
&lt;h3 id="65-existing-market-positioning-matters-more-than-one-launch"&gt;6.5 Existing market positioning matters more than one launch
&lt;/h3&gt;&lt;p&gt;From June 22 through July 16, before most of the K3 evidence could affect trading, the US AI complex had already diverged sharply.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Return&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Drawdown from period high&lt;/th&gt;
 &lt;th&gt;Positioning signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.1%&lt;/td&gt;
 &lt;td&gt;Strong ad cash flow and AI utilization expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.2%&lt;/td&gt;
 &lt;td&gt;Cloud and software resilience&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.2%&lt;/td&gt;
 &lt;td&gt;AWS and consumer recovery expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.5%&lt;/td&gt;
 &lt;td&gt;Mixed search and cloud positioning&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.1%&lt;/td&gt;
 &lt;td&gt;Relatively resilient&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Broadcom&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.7%&lt;/td&gt;
 &lt;td&gt;Custom AI optimism meets valuation pressure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AMD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-14.3%&lt;/td&gt;
 &lt;td&gt;Alternative accelerator optionality with high volatility&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-29.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-32.0%&lt;/td&gt;
 &lt;td&gt;Correction after elevated memory expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-29.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-32.7%&lt;/td&gt;
 &lt;td&gt;Tension between AI infrastructure growth and financing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Marvell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-38.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-40.1%&lt;/td&gt;
 &lt;td&gt;Severe derating in networking and custom silicon&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;These are not K3-caused returns. They show the positions into which K3 arrived. A relatively resilient NVIDIA may be more sensitive to efficiency headlines, while already-corrected Micron and Marvell could react more strongly if the released weights create verifiable infrastructure demand.&lt;/p&gt;
&lt;h2 id="7-company-impact-map"&gt;7. Company Impact Map
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company or layer&lt;/th&gt;
 &lt;th&gt;Near-term signal&lt;/th&gt;
 &lt;th&gt;Medium-term path&lt;/th&gt;
 &lt;th&gt;What to do now&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA&lt;/td&gt;
 &lt;td&gt;Valuation pressure from efficiency and portability&lt;/td&gt;
 &lt;td&gt;Self-hosted 64+ accelerator clusters offset efficiency&lt;/td&gt;
 &lt;td&gt;Watch workload elasticity, do not change estimates on the launch alone&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AMD&lt;/td&gt;
 &lt;td&gt;Alternative deployment optionality&lt;/td&gt;
 &lt;td&gt;Share upside if ROCm economics are proven&lt;/td&gt;
 &lt;td&gt;Wait for post-July 27 throughput&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Broadcom and Marvell&lt;/td&gt;
 &lt;td&gt;Networking complex already derating&lt;/td&gt;
 &lt;td&gt;MoE expert parallelism raises fabric demand&lt;/td&gt;
 &lt;td&gt;Verify orders and actual K3 cluster topology&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td&gt;Sensitive to KV-cache efficiency headlines&lt;/td&gt;
 &lt;td&gt;HBM, server DRAM and Solidigm SSD hierarchy exposure&lt;/td&gt;
 &lt;td&gt;Value the full memory stack, not HBM alone&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;HBM efficiency risk plus catch-up position&lt;/td&gt;
 &lt;td&gt;HBM, server DRAM, SSD and foundry optionality&lt;/td&gt;
 &lt;td&gt;Broadest portfolio, execution still required&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td&gt;High expectations already corrected&lt;/td&gt;
 &lt;td&gt;Integrated US AI memory and storage exposure&lt;/td&gt;
 &lt;td&gt;Watch deployment volume versus pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td&gt;OpenAI price and mix pressure&lt;/td&gt;
 &lt;td&gt;Azure and Copilot cost leverage&lt;/td&gt;
 &lt;td&gt;Cloud usage matters more than model margin&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td&gt;Pressure on Anthropic asset value&lt;/td&gt;
 &lt;td&gt;AWS, Bedrock and Trainium benefit from model choice&lt;/td&gt;
 &lt;td&gt;The cleanest internal offset&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td&gt;Gemini price pressure&lt;/td&gt;
 &lt;td&gt;TPU, Vertex and Search cost benefits&lt;/td&gt;
 &lt;td&gt;Application-layer defense remains strong&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td&gt;Pressure on US open-weight leadership&lt;/td&gt;
 &lt;td&gt;Llama acceleration or broader open ecosystem&lt;/td&gt;
 &lt;td&gt;Strategic impact exceeds near-term earnings impact&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;There is not enough evidence for a new buy or sell call from this launch alone. Weights, license and external hardware throughput remain unavailable. The order of proof matters more than the direction of the narrative.&lt;/p&gt;
&lt;h2 id="8-three-scenarios"&gt;8. Three Scenarios
&lt;/h2&gt;&lt;h3 id="base-case-strong-model-limited-re-rating"&gt;Base case: strong model, limited re-rating
&lt;/h3&gt;&lt;p&gt;Subjective probability: 55%. Full weights and a reasonably permissive license arrive by July 27. External results reproduce 85% to 95% of official performance. K3 runs on NVIDIA and AMD, but 64+ accelerators, max reasoning and high output-token use keep deployment expensive.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Proprietary APIs face more price pressure on repetitive work.&lt;/li&gt;
&lt;li&gt;Clouds gain K3 hosting and self-deployment demand.&lt;/li&gt;
&lt;li&gt;Per-token efficiency offsets higher workload.&lt;/li&gt;
&lt;li&gt;HBM is neutral to modestly positive.&lt;/li&gt;
&lt;li&gt;Server DRAM, enterprise SSD and networking are positive.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="bull-case-open-weights-enter-primary-enterprise-routing"&gt;Bull case: open weights enter primary enterprise routing
&lt;/h3&gt;&lt;p&gt;Subjective probability: 25%. The license is close to MIT, vLLM and SGLang support is stable, and NVIDIA, AMD and Chinese accelerators show strong throughput. External evaluation confirms performance immediately below the top proprietary models. Lower reasoning modes reduce task cost. Major clouds and enterprise gateways add K3 to default routing.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Proprietary model blended price and token share fall.&lt;/li&gt;
&lt;li&gt;Hyperscaler multi-model usage rises.&lt;/li&gt;
&lt;li&gt;Self-serving clusters increase accelerator demand.&lt;/li&gt;
&lt;li&gt;HBM, networking, server DRAM and enterprise SSD benefit from deployment volume.&lt;/li&gt;
&lt;li&gt;Meta and US open-model programs accelerate investment.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="bear-case-weights-reveal-a-cost-and-quality-gap"&gt;Bear case: weights reveal a cost and quality gap
&lt;/h3&gt;&lt;p&gt;Subjective probability: 20%. Weights are delayed or restricted, external scores fall below the official table, preserved-thinking-history requirements and excessive proactiveness create enterprise failures, and task cost exceeds Sonnet 5 because of max reasoning and verbosity. The 64-accelerator requirement limits self-hosting.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Moonshot may have to retreat from frontier pricing.&lt;/li&gt;
&lt;li&gt;Proprietary APIs retain a UX and reliability premium.&lt;/li&gt;
&lt;li&gt;Incremental semiconductor demand remains limited.&lt;/li&gt;
&lt;li&gt;Existing earnings and capex regain control of stock prices.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="9-the-july-27-verification-checklist"&gt;9. The July 27 Verification Checklist
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Proof point&lt;/th&gt;
 &lt;th&gt;Strong signal&lt;/th&gt;
 &lt;th&gt;Weak signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Weights and license&lt;/td&gt;
 &lt;td&gt;Full release on time with commercial modification rights&lt;/td&gt;
 &lt;td&gt;Delay, restrictions or missing components&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;External evaluation&lt;/td&gt;
 &lt;td&gt;Official scores broadly reproduced under one harness&lt;/td&gt;
 &lt;td&gt;Large drop versus company table&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cost per task&lt;/td&gt;
 &lt;td&gt;Lower reasoning modes and fewer output tokens&lt;/td&gt;
 &lt;td&gt;Max-only reasoning and persistent verbosity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA throughput&lt;/td&gt;
 &lt;td&gt;Stable expert parallelism on 64-accelerator nodes&lt;/td&gt;
 &lt;td&gt;Fabric bottlenecks and low utilization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AMD throughput&lt;/td&gt;
 &lt;td&gt;Cost advantage under ROCm and vLLM&lt;/td&gt;
 &lt;td&gt;Immature kernels or accuracy loss&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Chinese accelerators&lt;/td&gt;
 &lt;td&gt;Named hardware and measured throughput&lt;/td&gt;
 &lt;td&gt;Only “alternative GPU” language&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Caching&lt;/td&gt;
 &lt;td&gt;Near-90% hits on enterprise traffic&lt;/td&gt;
 &lt;td&gt;Sharp decline outside coding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud adoption&lt;/td&gt;
 &lt;td&gt;AWS, Azure, Google Cloud or Oracle listings&lt;/td&gt;
 &lt;td&gt;Limited to a few Chinese platforms&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Competitive pricing&lt;/td&gt;
 &lt;td&gt;Standard price cuts or wider cache discounts&lt;/td&gt;
 &lt;td&gt;Temporary promotions only&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory orders&lt;/td&gt;
 &lt;td&gt;Upward revisions to HBM, server DRAM and SSD volumes&lt;/td&gt;
 &lt;td&gt;Efficiency rises while volumes stagnate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="10-the-strongest-counterargument"&gt;10. The Strongest Counterargument
&lt;/h2&gt;&lt;p&gt;The strongest bear case is straightforward: semiconductor demand falls if efficiency improves faster than usage. Model compression, linear attention, quantization, cache reuse and inference ASICs could process the same number of useful tasks with far fewer GPUs and less HBM. Open-weight competition can reduce API prices without producing enough incremental paid work to justify AI capex.&lt;/p&gt;
&lt;p&gt;Evidence for that outcome would include slower inference revenue despite falling prices, higher accelerator utilization but fewer new clusters, HBM bit growth below model-efficiency gains, weak conversion of cloud AI backlog into revenue, and enterprises using open models only to cut costs rather than create new workflows.&lt;/p&gt;
&lt;p&gt;The bull case requires price declines to unlock new work: repository-scale coding, research automation, video editing, chip design and workflows that were previously uneconomic. Workload elasticity relative to model efficiency is the common proof point for both accelerator and HBM investing.&lt;/p&gt;
&lt;h2 id="11-conclusion"&gt;11. Conclusion
&lt;/h2&gt;&lt;p&gt;Kimi K3 does not prove that China has surpassed the strongest US proprietary models. Moonshot acknowledges the remaining UX gap. As of July 17, the full weights and technical report are not available, and a mixed-harness benchmark table cannot declare a winner.&lt;/p&gt;
&lt;p&gt;It does change market structure. A 2.8T, 1M-context, near-frontier model is live at Sonnet&amp;rsquo;s standard price, with weights promised within ten days. Chinese open weights are moving from cheap second-tier substitutes toward primary enterprise workloads.&lt;/p&gt;
&lt;p&gt;For semiconductors, the event is more demand reallocation than demand destruction. KDA and quantization reduce HBM and GPU time per request. Sparse MoE and 64-accelerator supernodes increase model-residency memory and fabric. Mooncake pushes cache into server DRAM and enterprise SSDs. The HBM-only story becomes less simple, while the full memory and data-center hierarchy remains exposed to wider deployment.&lt;/p&gt;
&lt;p&gt;US Big Tech faces the same split. Model APIs absorb price pressure; clouds and applications absorb lower model cost. Microsoft, Amazon and Alphabet can host K3 even if their preferred models lose share. Meta must defend its strategic role as the trusted US open-weight standard.&lt;/p&gt;
&lt;p&gt;On July 27, the important evidence is not the existence of a weight file. It is a permissive license, reproducible evaluation, multi-hardware throughput and competitive cost per completed task. If those conditions hold, K3 becomes an event that changes both the AI price curve and the semiconductor demand path. If they do not, it remains an impressive product launch rather than an industry reset.&lt;/p&gt;
&lt;h2 id="sources-and-limitations"&gt;Sources and Limitations
&lt;/h2&gt;&lt;p&gt;Primary materials: &lt;a class="link" href="https://www.kimi.com/blog/kimi-k3" target="_blank" rel="noopener"
 &gt;Kimi K3 launch&lt;/a&gt;, &lt;a class="link" href="https://platform.kimi.ai/docs/pricing/chat-k3" target="_blank" rel="noopener"
 &gt;Kimi K3 API documentation&lt;/a&gt;, &lt;a class="link" href="https://arxiv.org/abs/2510.26692" target="_blank" rel="noopener"
 &gt;Kimi Linear paper&lt;/a&gt;, &lt;a class="link" href="https://github.com/MoonshotAI/Kimi-Linear" target="_blank" rel="noopener"
 &gt;Kimi Linear GitHub&lt;/a&gt;, &lt;a class="link" href="https://arxiv.org/abs/2603.15031" target="_blank" rel="noopener"
 &gt;Attention Residuals paper&lt;/a&gt;, &lt;a class="link" href="https://arxiv.org/abs/2407.00079" target="_blank" rel="noopener"
 &gt;Mooncake paper&lt;/a&gt;, &lt;a class="link" href="https://www.anthropic.com/news/claude-sonnet-5" target="_blank" rel="noopener"
 &gt;Claude Sonnet 5 pricing&lt;/a&gt;, &lt;a class="link" href="https://developers.openai.com/api/docs/models/gpt-5.6-sol" target="_blank" rel="noopener"
 &gt;GPT-5.6 Sol pricing&lt;/a&gt;, &lt;a class="link" href="https://openrouter.ai/blog/insights/deepseek-v4-adoption/" target="_blank" rel="noopener"
 &gt;OpenRouter model adoption analysis&lt;/a&gt;, &lt;a class="link" href="https://blogs.microsoft.com/blog/2026/04/27/the-next-phase-of-the-microsoft-openai-partnership/" target="_blank" rel="noopener"
 &gt;Microsoft-OpenAI partnership&lt;/a&gt;, &lt;a class="link" href="https://cloud.google.com/vertex-ai/generative-ai/docs/model-garden/explore-models" target="_blank" rel="noopener"
 &gt;Google Vertex Model Garden&lt;/a&gt;, and &lt;a class="link" href="https://www.aboutamazon.com/news/company-news/amazon-aws-anthropic-ai" target="_blank" rel="noopener"
 &gt;Amazon-Anthropic partnership&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;The semiconductor and stock transmission analysis is an inference from public architecture, serving and market data. Exact active parameters, weight size, license terms, AMD and Chinese-accelerator throughput, and enterprise cost per completed task remain blocked as of July 17. This article is for research and information purposes only and is not investment advice.&lt;/p&gt;</description></item><item><title>Will Apple Really Use CXMT? China-Only Memory Testing and the Pricing Risk for Korean DRAM Makers</title><link>https://koreainvestinsights.com/post/apple-cxmt-dram-china-test-memory-pricing-korea-2026-07-14/</link><pubDate>Tue, 14 Jul 2026 22:10:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/apple-cxmt-dram-china-test-memory-pricing-korea-2026-07-14/</guid><description>&lt;p&gt;Apple is reportedly testing DRAM made by ChangXin Memory Technologies, or CXMT, for devices sold in China. No supply contract has been announced. Those two facts must be read together: CXMT has entered Apple&amp;rsquo;s qualification process, but it has not become the main memory supplier for global iPhones or Macs.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Context: This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/cxmt-ipo-memory-price-risk-hbm-client-dram-2026-06-21/" &gt;CXMT&amp;rsquo;s IPO and memory price risk&lt;/a&gt;. That report argued that CXMT capacity is more likely to cap client DRAM and LPDDR pricing before it disrupts HBM. This article examines how a buyer as large as Apple could use that emerging supply. Related collections are the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Apple could eventually qualify CXMT as a commercial supplier. The most plausible first step is limited volume in standard devices sold in China. A rapid move into global flagship products remains unlikely.&lt;/li&gt;
&lt;li&gt;CXMT&amp;rsquo;s public LPDDR5X specifications are sufficient to enter qualification. Its 8,533Mbps and 9,600Mbps products are in mass production, while the 10,667Mbps version remains in customer sampling. Apple-grade power, thermals, package thickness, defect rates, long-term reliability, and high-volume yield remain undisclosed.&lt;/li&gt;
&lt;li&gt;A reported 5% to 10% average selling-price discount does not settle the economics. That estimate covers CXMT&amp;rsquo;s mixed DRAM portfolio, not an Apple-specific LPDDR5X quote. Dual bills of materials, inventory separation, qualification, and regulatory replacement risk reduce the net savings.&lt;/li&gt;
&lt;li&gt;Negotiating leverage may matter more than direct units. A fourth supplier capable of filling a commercial order could pressure Samsung, SK Hynix, and Micron across Apple&amp;rsquo;s much larger DRAM procurement book.&lt;/li&gt;
&lt;li&gt;CXMT&amp;rsquo;s near-term competitive pressure is concentrated in mobile and client DRAM, not leading-edge HBM. Treating the report as a blanket negative for every memory stock ignores product mix.&lt;/li&gt;
&lt;li&gt;Event-only ratings are Watchlist for SK Hynix, Wait for Samsung Electronics, and Wait for Apple. The evidence currently supports testing and lobbying, not a disclosed contract, product, volume, or duration.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="1-what-is-confirmed-and-what-is-not"&gt;1. What Is Confirmed and What Is Not
&lt;/h2&gt;&lt;h3 id="11-testing-is-reported-a-contract-is-not"&gt;1.1 Testing is reported; a contract is not
&lt;/h3&gt;&lt;p&gt;The Financial Times reported that Apple is testing CXMT DRAM for devices sold in China. Earlier reporting said Apple had sought regulatory clarity from the U.S. government. Apple, the White House, and CXMT did not respond to Reuters requests for comment at the time.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;Assessment as of July 14, 2026&lt;/th&gt;
 &lt;th&gt;Evidence level&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Is Apple testing CXMT DRAM?&lt;/td&gt;
 &lt;td&gt;Reported for China-market devices&lt;/td&gt;
 &lt;td&gt;Credible secondary reporting&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Has a supply contract been signed?&lt;/td&gt;
 &lt;td&gt;Not confirmed&lt;/td&gt;
 &lt;td&gt;Blocked&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Which product is involved?&lt;/td&gt;
 &lt;td&gt;iPhone, iPad, or Mac cannot be identified&lt;/td&gt;
 &lt;td&gt;Blocked&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Will the chips be used outside China?&lt;/td&gt;
 &lt;td&gt;No supporting evidence&lt;/td&gt;
 &lt;td&gt;Do not infer&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Has the U.S. government approved use?&lt;/td&gt;
 &lt;td&gt;Regulatory certainty was reportedly requested&lt;/td&gt;
 &lt;td&gt;No official decision&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Have incumbent allocations been reduced?&lt;/td&gt;
 &lt;td&gt;Not confirmed&lt;/td&gt;
 &lt;td&gt;Blocked&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The accurate wording is that Apple is qualifying CXMT as a potential supplier, not that Apple has already adopted it. &lt;a class="link" href="https://www.marketscreener.com/news/apple-seeks-approval-to-buy-chips-from-blacklisted-chinese-company-ft-reports-ce7f5fd9d08df72d" target="_blank" rel="noopener"
 &gt;Reuters coverage&lt;/a&gt;, &lt;a class="link" href="https://www.macrumors.com/2026/07/08/apple-begins-testing-controversial-chinese-memory/" target="_blank" rel="noopener"
 &gt;testing report summary&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="12-cxmt-is-no-longer-an-experimental-producer"&gt;1.2 CXMT is no longer an experimental producer
&lt;/h3&gt;&lt;p&gt;Counterpoint Research estimated CXMT&amp;rsquo;s global DRAM revenue share at roughly 8% in 1Q26, placing it fourth behind Samsung, SK Hynix, and Micron. Market-share figures differ depending on whether the denominator is revenue, bit shipments, or wafer input, but the direction is clear: CXMT is no longer a minor laboratory-scale supplier. &lt;a class="link" href="https://counterpointresearch.com/en/insights/global-dram-revenue-surges-to-near-dollar-100-billion-mark-in-q1-2026" target="_blank" rel="noopener"
 &gt;Counterpoint Research&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Scale does not prove Apple-grade maturity. Supplying commodity DRAM to large Chinese customers does not establish the power, reliability, and yield required in Apple&amp;rsquo;s premium devices.&lt;/p&gt;
&lt;h2 id="2-performance-qualification-is-harder-than-reaching-a-headline-speed"&gt;2. Performance: Qualification Is Harder Than Reaching a Headline Speed
&lt;/h2&gt;&lt;p&gt;CXMT says its LPDDR5X portfolio includes 12Gb and 16Gb dies, 12GB, 16GB, and 24GB packages, and data rates up to 10,667Mbps. Its 8,533Mbps and 9,600Mbps products entered mass production in May 2025; the 10,667Mbps version remains in customer sampling. &lt;a class="link" href="https://www.cxmt.com/en/news/info_19.html" target="_blank" rel="noopener"
 &gt;CXMT announcement&lt;/a&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Confirmed status&lt;/th&gt;
 &lt;th&gt;Investment interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Die density&lt;/td&gt;
 &lt;td&gt;12Gb and 16Gb&lt;/td&gt;
 &lt;td&gt;Meets an entry-level mobile and PC requirement&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Package capacity&lt;/td&gt;
 &lt;td&gt;12GB, 16GB, and 24GB&lt;/td&gt;
 &lt;td&gt;Supports standard device configurations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Data rate&lt;/td&gt;
 &lt;td&gt;8,533, 9,600, and 10,667Mbps&lt;/td&gt;
 &lt;td&gt;The highest-speed part is still sampling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power&lt;/td&gt;
 &lt;td&gt;CXMT claims 30% savings versus its LPDDR5&lt;/td&gt;
 &lt;td&gt;Not a like-for-like comparison with the big three&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple qualification&lt;/td&gt;
 &lt;td&gt;Undisclosed&lt;/td&gt;
 &lt;td&gt;Commercial adoption cannot be established&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High-volume yield&lt;/td&gt;
 &lt;td&gt;Undisclosed&lt;/td&gt;
 &lt;td&gt;Tens-of-millions-unit supply cannot be established&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Public specifications refute the claim that Chinese DRAM is automatically too slow for Apple. The real bottlenecks are standby and refresh power, thermal stability, package thickness, defect rates, lot-to-lot variation, long-term reliability, high-volume yield, and integration with Apple&amp;rsquo;s processors and power-management firmware.&lt;/p&gt;
&lt;p&gt;CXMT has earned a seat in qualification. Public evidence does not show that it can maintain Apple-grade quality at scale.&lt;/p&gt;
&lt;h2 id="3-price-a-5-to-10-discount-does-not-settle-the-economics"&gt;3. Price: A 5% to 10% Discount Does Not Settle the Economics
&lt;/h2&gt;&lt;p&gt;SemiAnalysis estimated that CXMT&amp;rsquo;s blended DRAM average selling price in 1Q26 was roughly 5% to 10% below the incumbent trio, while its DDR5 cost per bit could be more than 30% higher. These are company-wide mixed-product estimates, not Apple-specific LPDDR5X bids. &lt;a class="link" href="https://newsletter.semianalysis.com/p/chinas-cxmt-is-set-to-challenge-dram" target="_blank" rel="noopener"
 &gt;SemiAnalysis&lt;/a&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Additional cost&lt;/th&gt;
 &lt;th&gt;Burden on Apple&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;China and global supply-chain separation&lt;/td&gt;
 &lt;td&gt;Separate part numbers, assembly planning, and logistics&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Dual inventory&lt;/td&gt;
 &lt;td&gt;Regional demand mismatch and slower turns&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Qualification engineering&lt;/td&gt;
 &lt;td&gt;Repeated processor, firmware, and thermal validation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Regulatory replacement risk&lt;/td&gt;
 &lt;td&gt;Emergency supplier switch and requalification&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Quality failure&lt;/td&gt;
 &lt;td&gt;Warranty, recall, and brand damage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A nominal discount shrinks after these costs. CXMT&amp;rsquo;s pricing also should not be read as proof of structural cost leadership. State support, lower margin requirements, product mix, and domestic-market strategy may contribute.&lt;/p&gt;
&lt;h2 id="4-negotiating-leverage-may-matter-more-than-units"&gt;4. Negotiating Leverage May Matter More Than Units
&lt;/h2&gt;&lt;p&gt;Apple&amp;rsquo;s economic benefit is not limited to the discount on chips purchased from CXMT. Qualifying a fourth supplier capable of filling a real commercial order could influence the price of Apple&amp;rsquo;s entire DRAM procurement book.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Maintain CXMT qualification
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ enable a small production order
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ demonstrate credible switching capacity
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ limit contract-price increases from the incumbent trio
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;This effect can appear before CXMT gains visible market share. A small allocation removes the incumbents&amp;rsquo; assumption that Apple has no alternative.&lt;/p&gt;
&lt;p&gt;Samples alone are not enough. The option becomes credible only if CXMT can deliver a minimum commercial volume into a shipping product. Without a contract and a willingness to switch, the incumbents have little reason to change their pricing.&lt;/p&gt;
&lt;h2 id="5-regulation-legal-purchasability-and-multi-year-supply-security-are-different"&gt;5. Regulation: Legal Purchasability and Multi-Year Supply Security Are Different
&lt;/h2&gt;&lt;h3 id="51-a-section-1260h-designation-does-not-automatically-ban-ordinary-commercial-purchases"&gt;5.1 A Section 1260H designation does not automatically ban ordinary commercial purchases
&lt;/h3&gt;&lt;p&gt;The U.S. Department of Defense included CXMT on its June 8, 2026 Section 1260H list of Chinese military companies. The designation creates government-contracting and national-security risk, but it is not a blanket prohibition on every private commercial transaction. &lt;a class="link" href="https://www.war.gov/News/Releases/Release/Article/4511232/dow-releases-list-of-chinese-military-companies-in-accordance-with-section-1260/" target="_blank" rel="noopener"
 &gt;U.S. Department of Defense release&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="52-entity-list-risk-matters-most-through-production-continuity"&gt;5.2 Entity List risk matters most through production continuity
&lt;/h3&gt;&lt;p&gt;Reuters reported that an interagency committee had approved CXMT for Entity List addition, but publication remained on hold as of mid-June 2026. An Entity List designation would impose severe licensing restrictions on exports, reexports, or transfers of U.S. equipment, software, technology, and support to CXMT. Buying a finished DRAM chip is legally distinct from exporting U.S. technology to the producer.&lt;/p&gt;
&lt;p&gt;The commercial consequence is still large. Restricted spare parts, software updates, and engineering support could impair yield, capacity, and process migration. Apple needs multi-year supply continuity, not merely the ability to execute one purchase today. &lt;a class="link" href="https://www.investing.com/news/stock-market-news/us-holds-off-on-adding-deepseek-cxmt-to-trade-blacklist-reuters-reports-4746250" target="_blank" rel="noopener"
 &gt;Reuters report on the delay&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="53-the-2027-federal-procurement-restriction-is-not-a-consumer-ban"&gt;5.3 The 2027 federal procurement restriction is not a consumer ban
&lt;/h3&gt;&lt;p&gt;A proposed Federal Acquisition Regulation would implement restrictions from December 23, 2027 on federal-agency procurement of electronic products and services containing covered chips from CXMT, YMTC, and SMIC. It does not ban ordinary consumers from buying iPhones. It could, however, force Apple to maintain separate non-CXMT configurations for government and some enterprise channels. The proposal includes an exception through December 23, 2028 for commercial products without alternatives. &lt;a class="link" href="https://public-inspection.federalregister.gov/2026-03065.pdf" target="_blank" rel="noopener"
 &gt;Proposed FAR rule&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="54-ymtc-is-the-relevant-political-precedent"&gt;5.4 YMTC is the relevant political precedent
&lt;/h3&gt;&lt;p&gt;Apple explored using YMTC NAND in 2022 but shelved the plan after congressional opposition and tighter restrictions. CXMT is not in the same regulatory position today, but the episode shows that political pressure can interrupt a technically plausible sourcing plan.&lt;/p&gt;
&lt;h2 id="6-the-most-plausible-path-limited-use-in-standard-devices-sold-in-china"&gt;6. The Most Plausible Path: Limited Use in Standard Devices Sold in China
&lt;/h2&gt;&lt;p&gt;China-only standard models offer a compromise among engineering, economics, and politics.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Variable&lt;/th&gt;
 &lt;th&gt;Limited China-market adoption&lt;/th&gt;
 &lt;th&gt;Global primary adoption&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Qualification&lt;/td&gt;
 &lt;td&gt;Standard products can be tested first&lt;/td&gt;
 &lt;td&gt;Long validation across premium products&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Regulatory exposure&lt;/td&gt;
 &lt;td&gt;Can be separated from U.S. government channels&lt;/td&gt;
 &lt;td&gt;Conflicts across government and enterprise sales&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Supply-chain operations&lt;/td&gt;
 &lt;td&gt;Separate China bill of materials&lt;/td&gt;
 &lt;td&gt;Larger loss of global commonality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Volume&lt;/td&gt;
 &lt;td&gt;Limited&lt;/td&gt;
 &lt;td&gt;Tens of millions of devices&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Political burden&lt;/td&gt;
 &lt;td&gt;Potentially manageable&lt;/td&gt;
 &lt;td&gt;Much greater U.S. opposition and sanction risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current probability&lt;/td&gt;
 &lt;td&gt;Conditional&lt;/td&gt;
 &lt;td&gt;Low&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Four conditions must hold: Apple reliability qualification, sufficient Chinese yield and volume, regulatory continuity over the contract, and net benefits larger than the cost of dual sourcing.&lt;/p&gt;
&lt;h2 id="7-capital-flow-and-choke-points"&gt;7. Capital Flow and Choke Points
&lt;/h2&gt;&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;AI data-center investment rises
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ HBM and server DRAM receive priority
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ mobile and PC DRAM supply becomes less elastic
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Apple&amp;#39;s memory bill rises
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ the incentive to qualify CXMT increases
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The chain runs from equipment and materials through CXMT wafer production, LPDDR5X packaging and testing, Apple qualification, device assembly, and China distribution. The four choke points are Apple-grade quality and yield, regulatory durability, the cost of separate regional configurations, and CXMT&amp;rsquo;s access to U.S. equipment and technical support.&lt;/p&gt;
&lt;h2 id="8-memory-sector-impact-separate-hbm-from-mobile-dram"&gt;8. Memory-Sector Impact: Separate HBM from Mobile DRAM
&lt;/h2&gt;&lt;p&gt;CXMT&amp;rsquo;s current competitive position is closer to DDR5 and LPDDR than to leading-edge HBM. Extrapolating Apple&amp;rsquo;s test to a collapse in the entire DRAM profit pool ignores product economics.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Direct CXMT exposure&lt;/th&gt;
 &lt;th&gt;More important variable&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Potential LPDDR pricing and share pressure at Apple&lt;/td&gt;
 &lt;td&gt;HBM qualification, commodity mix, foundry losses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Some mobile exposure, but a larger share of the core thesis comes from HBM&lt;/td&gt;
 &lt;td&gt;HBM4 yield, share, ASP, and AI capex&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td&gt;Potential mobile and PC DRAM pressure&lt;/td&gt;
 &lt;td&gt;Server DRAM, HBM ramp, long-term agreements&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple&lt;/td&gt;
 &lt;td&gt;Lower procurement cost and supplier diversification&lt;/td&gt;
 &lt;td&gt;Dual-BOM cost, regulation, total gross margin&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT&lt;/td&gt;
 &lt;td&gt;Apple qualification would improve global credibility&lt;/td&gt;
 &lt;td&gt;Yield, cost, equipment access, regulatory security&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;More CXMT mobile capacity could encourage the incumbent trio to allocate still more capital to HBM and server DRAM. That would pressure commodity pricing without eliminating HBM&amp;rsquo;s qualification, stacking-yield, packaging, and customer-roadmap barriers.&lt;/p&gt;
&lt;p&gt;The long-term risk is real. If CXMT rapidly improves advanced DDR5 and LPDDR5X yield and absorbs export-control shocks through a domestic equipment ecosystem, the incumbents&amp;rsquo; commodity cash flow could weaken. If it later achieves commercial HBM yield, today&amp;rsquo;s product boundary would erode. Public evidence does not show that outcome yet.&lt;/p&gt;
&lt;h2 id="9-three-investment-frames-and-their-red-teams"&gt;9. Three Investment Frames and Their Red Teams
&lt;/h2&gt;&lt;h3 id="frame-1-limited-china-market-adoption"&gt;Frame 1: Limited China-market adoption
&lt;/h3&gt;&lt;p&gt;Classification: idiosyncratic event alpha&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Price: a lower CXMT supply price&lt;/li&gt;
&lt;li&gt;Quantity: selected devices sold in China&lt;/li&gt;
&lt;li&gt;Cost: qualification, dual inventory, regulation, and warranty risk&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Beneficiaries include Apple&amp;rsquo;s procurement organization, CXMT, and China&amp;rsquo;s packaging and materials chain. Incumbent Apple mobile-DRAM pricing would face pressure.&lt;/p&gt;
&lt;p&gt;Red team: A durable U.S.-China agreement, regulatory safe harbor, and premium-device qualification could break the China-only assumption. A failed reliability test would invalidate even limited adoption.&lt;/p&gt;
&lt;h3 id="frame-2-negotiating-impact-appears-before-market-share-impact"&gt;Frame 2: Negotiating impact appears before market-share impact
&lt;/h3&gt;&lt;p&gt;Classification: idiosyncratic event alpha&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Price: Apple contract prices with the incumbent trio&lt;/li&gt;
&lt;li&gt;Quantity: Apple&amp;rsquo;s full DRAM procurement book&lt;/li&gt;
&lt;li&gt;Cost: maintaining CXMT qualification&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Investors can observe supplier share, but the marginal supplier may influence contract pricing long before it wins large share.&lt;/p&gt;
&lt;p&gt;Red team: A memory surplus would force incumbents to cut prices without CXMT. A failed Apple qualification would also make the option non-credible.&lt;/p&gt;
&lt;h3 id="frame-3-cxmt-risk-is-concentrated-in-mobile-and-client-dram"&gt;Frame 3: CXMT risk is concentrated in mobile and client DRAM
&lt;/h3&gt;&lt;p&gt;Classification: product-mix trade&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Price: the ceiling on mobile and PC DRAM pricing&lt;/li&gt;
&lt;li&gt;Quantity: CXMT capacity and domestic shipments&lt;/li&gt;
&lt;li&gt;Cost: advanced-node yield and U.S. equipment restrictions&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Mobile DRAM buyers and HBM-heavy suppliers are relative beneficiaries. Suppliers dependent on Apple mobile DRAM or commodity price increases face more risk.&lt;/p&gt;
&lt;p&gt;Red team: Rapid progress in advanced-node yield and commercial HBM would weaken the product separation. A collapse in AI data-center investment would also remove HBM&amp;rsquo;s relative defense.&lt;/p&gt;
&lt;h2 id="10-stock-specific-monitoring-gates"&gt;10. Stock-Specific Monitoring Gates
&lt;/h2&gt;&lt;p&gt;These are conditional event assessments, not stand-alone buy recommendations.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Security&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Displayed P/E on July 14, 2026&lt;/th&gt;
 &lt;th&gt;Event interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.97x&lt;/td&gt;
 &lt;td&gt;HBM4 and AI investment matter more than direct CXMT risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20.99x&lt;/td&gt;
 &lt;td&gt;Mobile pricing pressure must be weighed against HBM recovery&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td style="text-align: right"&gt;38.38x&lt;/td&gt;
 &lt;td&gt;iPhone, services, and total gross margin matter more than unproven CXMT savings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;P/E figures are displayed values from &lt;a class="link" href="https://www.google.com/finance/quote/005930:KRX" target="_blank" rel="noopener"
 &gt;Google Finance for Samsung&lt;/a&gt;, &lt;a class="link" href="https://www.google.com/finance/quote/000660:KRX" target="_blank" rel="noopener"
 &gt;SK Hynix&lt;/a&gt;, and &lt;a class="link" href="https://www.google.com/finance/quote/AAPL:NASDAQ" target="_blank" rel="noopener"
 &gt;Apple&lt;/a&gt;. Trailing P/E can overstate cheapness for cyclical memory companies near a profit peak.&lt;/p&gt;
&lt;h3 id="sk-hynix"&gt;SK Hynix
&lt;/h3&gt;&lt;p&gt;Confirmation gates: HBM4 qualification and production timing remain intact, HBM yield and ASP estimates do not fall, and CXMT-related share-price weakness occurs without an HBM earnings downgrade.&lt;/p&gt;
&lt;p&gt;Invalidation: HBM4 qualification delay, customer diversification that reduces share, or CXMT achieving commercial yield in advanced server DRAM or HBM.&lt;/p&gt;
&lt;h3 id="samsung-electronics"&gt;Samsung Electronics
&lt;/h3&gt;&lt;p&gt;Confirmation gates: primary-supplier status in the next Apple generation, CXMT use limited to China-market volume, and visible HBM yield and customer-qualification improvement.&lt;/p&gt;
&lt;p&gt;Invalidation: a material multi-year China-device allocation to CXMT, simultaneous loss of LPDDR share and pricing, or an HBM recovery too slow to offset mobile pressure.&lt;/p&gt;
&lt;h3 id="apple"&gt;Apple
&lt;/h3&gt;&lt;p&gt;Confirmation gates: gross-margin guidance holds despite memory inflation, regulation provides durable clarity, and iPhone and services estimates rise independently of supplier diversification.&lt;/p&gt;
&lt;p&gt;Invalidation: regulation removes the CXMT option, memory inflation drives both price hikes and demand weakness, or slower iPhone and services growth fails to support the valuation.&lt;/p&gt;
&lt;h2 id="11-evidence-that-would-change-the-thesis"&gt;11. Evidence That Would Change the Thesis
&lt;/h2&gt;&lt;p&gt;Upside evidence includes an official Apple supplier confirmation, teardown proof of CXMT DRAM, qualification beyond China-market products, a durable U.S. safe harbor, or high-volume yield for CXMT&amp;rsquo;s 10,667Mbps and next-generation products.&lt;/p&gt;
&lt;p&gt;Downside evidence includes failed Apple reliability testing, an Entity List action that interrupts equipment support, the qualification ending without a contract, dual-sourcing costs exceeding savings, or a DRAM surplus that removes Apple&amp;rsquo;s need for a fourth supplier.&lt;/p&gt;
&lt;h2 id="12-evidence-classification"&gt;12. Evidence Classification
&lt;/h2&gt;&lt;h3 id="fact"&gt;Fact
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Apple is reportedly testing CXMT DRAM for China-market devices.&lt;/li&gt;
&lt;li&gt;Apple reportedly sought U.S. regulatory clarity regarding CXMT.&lt;/li&gt;
&lt;li&gt;CXMT mass-produces 8,533Mbps and 9,600Mbps LPDDR5X and samples a 10,667Mbps product.&lt;/li&gt;
&lt;li&gt;Counterpoint estimated CXMT&amp;rsquo;s 1Q26 global DRAM revenue share at about 8%.&lt;/li&gt;
&lt;li&gt;CXMT appears on the June 2026 Section 1260H list.&lt;/li&gt;
&lt;li&gt;Section 1260H does not by itself automatically prohibit ordinary private purchases.&lt;/li&gt;
&lt;li&gt;CXMT had not been formally added to the Entity List as of mid-June 2026.&lt;/li&gt;
&lt;li&gt;The proposed FAR rule implements a December 23, 2027 federal procurement restriction for covered semiconductors.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="inference"&gt;Inference
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Initial commercial use is most likely limited to standard devices sold in China.&lt;/li&gt;
&lt;li&gt;Apple&amp;rsquo;s largest economic benefit may be incumbent negotiation leverage rather than direct chip discounts.&lt;/li&gt;
&lt;li&gt;CXMT&amp;rsquo;s near-term threat is concentrated in mobile and client DRAM rather than HBM.&lt;/li&gt;
&lt;li&gt;Separate China and global configurations are more plausible than a full global switch.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="speculation"&gt;Speculation
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Whether the test involves iPhone, iPad, or Mac&lt;/li&gt;
&lt;li&gt;Commercial launch timing&lt;/li&gt;
&lt;li&gt;Initial volume and contract duration&lt;/li&gt;
&lt;li&gt;Whether Washington will provide a durable safe harbor&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="blocked"&gt;Blocked
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Apple&amp;rsquo;s internal qualification results&lt;/li&gt;
&lt;li&gt;Apple-specific pricing and the discount versus incumbents&lt;/li&gt;
&lt;li&gt;Apple-standard power, package thickness, defect rates, and reliability&lt;/li&gt;
&lt;li&gt;Apple-volume yield and monthly CXMT capacity&lt;/li&gt;
&lt;li&gt;Contract duration and minimum volume&lt;/li&gt;
&lt;li&gt;Exact product-level incumbent share and pricing at Apple&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;Apple can use CXMT, but the only reported path today is testing for devices sold in China. Public specifications justify qualification, while Apple-grade quality, high-volume yield, and multi-year regulatory security remain unproven.&lt;/p&gt;
&lt;p&gt;The first market variable to watch is not CXMT&amp;rsquo;s Apple share. It is whether a commercially credible fourth supplier changes mobile-DRAM contract pricing for the incumbent trio. That pressure is more likely to appear in LPDDR and client DRAM before HBM.&lt;/p&gt;
&lt;p&gt;There is no reason to treat every Korean memory stock identically. Samsung must balance mobile pricing pressure against HBM recovery. For SK Hynix, HBM4 yield and the durability of AI investment remain more important than CXMT. The next decision should follow official supplier confirmation, product teardowns, an Entity List decision, and disclosed commercial volume.&lt;/p&gt;</description></item><item><title>What Remains from 2,893 Messages: Two Weeks of Memory Pricing, AI Infrastructure Bottlenecks, and Leveraged Flow</title><link>https://koreainvestinsights.com/post/telegram-two-week-market-deep-dive-memory-ai-infra-leverage-2026-07-14/</link><pubDate>Tue, 14 Jul 2026 18:15:55 +0900</pubDate><guid>https://koreainvestinsights.com/post/telegram-two-week-market-deep-dive-memory-ai-infra-leverage-2026-07-14/</guid><description>&lt;p&gt;Between July 1 and July 14, 2026, 2,893 messages were collected from 18 Telegram channels. Of these, 1,890 were classified as high-information-density, and 308 were original source documents after stripping forwards and reposts. The most frequently recurring topics were: semiconductors and AI infrastructure (352 mentions), interest rates, currencies, and macro (272), China, policy, and demand (77), biotech (74), and power, nuclear, and energy (69).&lt;/p&gt;
&lt;p&gt;These numbers should not be read directly as a measure of market conviction. When the same article or brokerage memo circulates across multiple channels, the message count rises without adding a single new fact. Conversely, a single official disclosure may not be widely shared yet can carry far greater impact on company valuations. Telegram is a &lt;strong&gt;map of attention&lt;/strong&gt; — it shows what the market is watching. It is not a ledger of confirmed facts.&lt;/p&gt;
&lt;p&gt;This post re-selects from those 308 source documents the signals most likely to alter market structure. Nanya Technology&amp;rsquo;s quarterly results, monthly revenue from TSMC, Wiwynn, Delta, CCL makers, and copper foil producers, Meta&amp;rsquo;s 5 GW data center plan, Brookfield and Bloom Energy&amp;rsquo;s $25 billion power financing, Fermi&amp;rsquo;s convertible notes, Micron&amp;rsquo;s U.S. capacity expansion, Korea&amp;rsquo;s single-stock leveraged products, and SK Hynix ADR discussions were read in cross-reference against one another.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Connected context: This post is a follow-up that re-links &lt;a class="link" href="https://koreainvestinsights.com/ko/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;the H1 2026 AI infrastructure and narrow market postmortem&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/ko/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;the Micron FY3Q26 earnings analysis&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/ko/post/kospi-9pct-selloff-ai-hardware-derating-korea-leverage-2026-07-13/" &gt;the KOSPI sell-off and leverage amplification analysis&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/ko/post/ai-token-value-memory-value-added-2026-07-09/" &gt;the AI token and memory value-added analysis&lt;/a&gt; through the flow of source documents from the first half of July.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;Memory earnings are still being driven by price, not volume. Nanya Technology&amp;rsquo;s Q2 DRAM average selling price rose more than 60% quarter-on-quarter while shipment volume barely changed. Current excess profits are powerful, but earnings sensitivity to any price reversal is equally high.&lt;/li&gt;
&lt;li&gt;AI investment has spread beyond GPUs and HBM to server ODMs, power supply units, liquid cooling, advanced copper foil, and CCL. Taiwan&amp;rsquo;s supply chain June revenue confirms this shift.&lt;/li&gt;
&lt;li&gt;The next bottleneck is power, and after power comes financing. Meta&amp;rsquo;s 5 GW plan and Brookfield&amp;rsquo;s $25 billion financing commitment show the scale of demand. Fermi&amp;rsquo;s $375 million convertible notes show that the same investments carry dilution and refinancing risk.&lt;/li&gt;
&lt;li&gt;Supply shortage is sowing the seeds of future supply. From 2028, when Nanya&amp;rsquo;s new fab, Micron&amp;rsquo;s U.S. investments, and the three major memory makers&amp;rsquo; capacity expansions converge, demand qualification cycles and depreciation will matter more than demand growth alone.&lt;/li&gt;
&lt;li&gt;In the Korean market, fundamentals and trading structure must be separated. Single-stock leveraged products can amplify both gains and losses, and ADR premiums may be the price of arbitrage friction rather than an increase in enterprise value.&lt;/li&gt;
&lt;li&gt;The claim that more agents means infinite memory demand remains a narrative. Total inference volume and context storage are growing, but compression, cache reuse, tiered storage, and hardware-software co-design are lowering cost per byte.&lt;/li&gt;
&lt;li&gt;The most important final metric is not AI capex. It is &lt;strong&gt;how much the end customer pays per unit of useful work completed, and how that money is divided among power, server, networking, and memory suppliers&lt;/strong&gt;.&lt;/li&gt;
&lt;/ol&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;One-Line Conclusion&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The messages from the first half of July read less as a signal that AI investment is over and more as a signal that &lt;strong&gt;the bottleneck capturing returns is shifting from chips to memory, networking, power, and financing&lt;/strong&gt;. That said, price-driven earnings, large-scale borrowing, and leveraged positioning have all grown simultaneously, widening the gap between a good industry and a good entry price.
 &lt;/div&gt;
&lt;/div&gt;
&lt;p&gt;&lt;img alt="Telegram market signal map for the first half of July 2026" loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/images/posts/telegram-two-week-market-grid-2026-07-14.png"&gt;
&lt;/p&gt;
&lt;h2 id="1-how-to-read-two-weeks-evidence-quality-over-message-count"&gt;1. How to Read Two Weeks: Evidence Quality Over Message Count
&lt;/h2&gt;&lt;h3 id="11-2893-messages-are-not-2893-events"&gt;1.1 2,893 Messages Are Not 2,893 Events
&lt;/h3&gt;&lt;p&gt;The most frequently captured topic over the two weeks was semiconductors and AI infrastructure. But the number 352 does not mean 352 independent positive catalysts. When a single brokerage report passes through an original channel, a repost channel, and a summary channel, it gets counted three or more times. The same article with only a changed headline also repeats.&lt;/p&gt;
&lt;p&gt;This analysis therefore divided messages into four tiers.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Tier&lt;/th&gt;
 &lt;th&gt;Basis&lt;/th&gt;
 &lt;th&gt;How to Use&lt;/th&gt;
 &lt;th&gt;Examples&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;A&lt;/td&gt;
 &lt;td&gt;Company IR, regulatory filings, official announcements&lt;/td&gt;
 &lt;td&gt;Factual reference point&lt;/td&gt;
 &lt;td&gt;Nanya results, TSMC monthly revenue, Meta capex, SEC filings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;B&lt;/td&gt;
 &lt;td&gt;Brokerage and industry data with figures and sources&lt;/td&gt;
 &lt;td&gt;Cross-checked against primary materials&lt;/td&gt;
 &lt;td&gt;Taiwan component IR summaries, industry supply/demand forecasts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C&lt;/td&gt;
 &lt;td&gt;Anonymous sourcing, Telegram interpretation, market memos&lt;/td&gt;
 &lt;td&gt;Signal of possibility and sentiment&lt;/td&gt;
 &lt;td&gt;Meta surplus compute sales, Samsung HBM yield, ADR premium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;D&lt;/td&gt;
 &lt;td&gt;Long-term narratives and extreme forecasts&lt;/td&gt;
 &lt;td&gt;Material for forming questions&lt;/td&gt;
 &lt;td&gt;10 billion agents, infinite memory demand growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Even if the same claim appears across multiple channels, if the original source is one, the evidence is one. Conversely, a single Tier A document can matter more than ten Tier C posts. Collection on July 14 may have partial gaps due to Telegram web domain connection errors, so that final day&amp;rsquo;s frequency is best excluded from comparisons.&lt;/p&gt;
&lt;h3 id="12-repetition-reveals-positioning-not-truth"&gt;1.2 Repetition Reveals Positioning, Not Truth
&lt;/h3&gt;&lt;p&gt;Repetition itself has its uses. What it reveals, however, is market participants&amp;rsquo; attention and positioning, not the probability that something is true.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;When multiple channels share the same HBM price increase, it means the memory bull thesis is already widely held.&lt;/li&gt;
&lt;li&gt;When posts about leveraged ETF side effects repeat, it means supply/demand anxiety has become investors&amp;rsquo; primary explanatory frame, independent of actual product scale.&lt;/li&gt;
&lt;li&gt;When power and cooling articles multiply, it means the market is searching for bottlenecks beyond GPUs.&lt;/li&gt;
&lt;li&gt;When ADR premium posts repeat, it means U.S.-listing structure has become a short-term price catalyst, overshadowing domestic earnings fundamentals.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The correct sequence for using Telegram in investment is therefore: &lt;code&gt;detect repetition → verify source document → cross-check against official data → connect to earnings and cash flow → check price reflection&lt;/code&gt;. Jumping from repetition detection directly to buying means purchasing the most crowded narrative at the latest possible entry.&lt;/p&gt;
&lt;h2 id="2-mapping-two-weeks-of-markets-onto-a-single-grid"&gt;2. Mapping Two Weeks of Markets onto a Single Grid
&lt;/h2&gt;&lt;p&gt;Placing the two weeks of source documents on four axes — price, volume, cost, and financing — makes the pattern clear.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Axis&lt;/th&gt;
 &lt;th&gt;Signal Confirmed in First Half of July&lt;/th&gt;
 &lt;th&gt;Investment Question&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Price (P)&lt;/td&gt;
 &lt;td&gt;Nanya DRAM ASP up more than 60% QoQ, memory supply shortage expected&lt;/td&gt;
 &lt;td&gt;How long does the price increase persist?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Volume (Q)&lt;/td&gt;
 &lt;td&gt;Taiwan server, power, CCL, copper foil monthly revenue up; Meta 5 GW plan&lt;/td&gt;
 &lt;td&gt;Are actual shipments and utilization rates growing as fast as the narrative?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cost (C)&lt;/td&gt;
 &lt;td&gt;Grid power, cooling, new fabs, depreciation, U.S. production costs&lt;/td&gt;
 &lt;td&gt;Does revenue growth translate into margin and free cash flow?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Financing (F)&lt;/td&gt;
 &lt;td&gt;$25B power financing, $375M convertible notes, long-term contracts and prepayments&lt;/td&gt;
 &lt;td&gt;Who provides the capital, and who bears dilution and refinancing risk?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Market narratives typically emphasize P and Q — prices are rising and demand is growing. But the durability of stock performance is determined by C and F. If the cost of building fabs and securing power surges, or if debt maturities arrive quickly, strong demand may not translate into strong shareholder returns.&lt;/p&gt;
&lt;p&gt;Within this grid, the key themes of the first half of July are:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;P is very strong. Memory prices are explosively expanding producer earnings.&lt;/li&gt;
&lt;li&gt;Q is also diffusing beyond chips. Server ODMs, power supply, cooling, and substrate raw materials all grew together.&lt;/li&gt;
&lt;li&gt;C is rising. Power, fab construction, cooling, and U.S. production costs are the next constraints on investment.&lt;/li&gt;
&lt;li&gt;The gap in F has widened. The divergence is growing between companies that can secure long-term capital cheaply and those that depend on dilutive funding.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="3-what-nanya-technology-revealed-about-the-nature-of-memory-earnings"&gt;3. What Nanya Technology Revealed About the Nature of Memory Earnings
&lt;/h2&gt;&lt;p&gt;If one had to pick the single most important source document from the two weeks, it would be the &lt;a class="link" href="https://t.me/Yeouido_Lab/36066" target="_blank" rel="noopener"
 &gt;Nanya Technology FY2Q26 earnings summary&lt;/a&gt;. But more important than the Telegram post is the separation of price and volume in &lt;a class="link" href="https://www.nanya.com/tw/IR/14/%E6%B4%BB%E5%8B%95%E8%A1%8C%E4%BA%8B%E6%9B%86?IRId=4295" target="_blank" rel="noopener"
 &gt;Nanya Technology&amp;rsquo;s official earnings release&lt;/a&gt;.&lt;/p&gt;
&lt;h3 id="31-the-numbers"&gt;3.1 The Numbers
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;QoQ&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YoY&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$82.549B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+684.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;79.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.6pp&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Large improvement&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$50.192B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Increase&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Expanded profit&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$14.66&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Increase&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Increase&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating cash flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$55.013B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Increase&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Increase&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net capex&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$4.046B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Limited&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Limited&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted free cash flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$50.967B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Surge&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Surge&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A 79.5% gross margin looks like a software company, yet the business is DRAM. The reason is price, not volume. According to company disclosures, DRAM average selling prices rose more than 60% quarter-on-quarter while shipments were nearly flat.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Revenue change ≈ Price change × Volume change
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Nanya FY2Q26 = Price surge × Flat volume
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;This structure simultaneously explains the advantage and the risk of the current memory bull cycle.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Advantage: When prices rise without additional wafers, a large portion of the revenue increase flows directly to operating income and cash flow.&lt;/li&gt;
&lt;li&gt;Risk: When price appreciation slows, volume growth cannot fill the gap, and earnings sensitivity works in reverse.&lt;/li&gt;
&lt;li&gt;Optical illusion: Applying a high P/E multiple to a peak-cycle gross margin means being bullish on both price and multiple simultaneously.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="32-why-this-cycle-may-last-longer-than-previous-ones"&gt;3.2 Why This Cycle May Last Longer Than Previous Ones
&lt;/h3&gt;&lt;p&gt;Nanya reported low inventory, AI infrastructure and server revenue exceeding 20% of the mix, and a supply shortage expected to continue for multiple quarters. Long-term supply contracts are also increasing. As HBM and server DRAM absorb wafer capacity and leading-edge process nodes, commodity DRAM supply has tightened as well.&lt;/p&gt;
&lt;p&gt;Micron similarly emphasized long-term customer contracts, price floors and ceilings, prepayments, and supply commitments in FY3Q26. As confirmed in the &lt;a class="link" href="https://koreainvestinsights.com/ko/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron earnings analysis&lt;/a&gt;, long-term contracts can extend the duration of memory earnings.&lt;/p&gt;
&lt;p&gt;But long-term contracts do not eliminate the cycle.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Contracts transform price risk into repricing-at-renewal risk and customer credit risk.&lt;/li&gt;
&lt;li&gt;Price ceilings protect downside but can cap additional upside during rapid price appreciation.&lt;/li&gt;
&lt;li&gt;Non-HBM DRAM and NAND remain exposed to spot and contract pricing cycles.&lt;/li&gt;
&lt;li&gt;When customers absorb excess capacity, they renegotiate volumes and prices in the next contract cycle.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="33-supply-shortage-is-seeding-2028-supply-expansion"&gt;3.3 Supply Shortage Is Seeding 2028 Supply Expansion
&lt;/h3&gt;&lt;p&gt;Nanya has set a FY2026 capex cap of NT$52B, and phase one of its new fab targets monthly output of 30,000 wafers by 2028, with estimated total investment of $16 billion for full 45,000 wafer-per-month capacity. Micron also announced plans to invest more than $250 billion in the United States through 2035, targeting 40% of global DRAM production on U.S. soil, per &lt;a class="link" href="https://investors.micron.com/news-releases/news-release-details/micron-accelerates-us-investments-pours-first-concrete-new-york" target="_blank" rel="noopener"
 &gt;Micron&amp;rsquo;s official announcement&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;The current supply shortage is real. At the same time, that shortage is building new supply and creating depreciation loads for 2028 and beyond. Investors must therefore hold two sentences simultaneously:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Memory prices and earnings in 2026–2027 may be strong. Beginning in 2028, the pace at which incremental capacity passes customer qualification and the weight of accumulated depreciation must be recalculated.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="4-samsung-electronics-and-hbm4e-customer-shipments-matter-more-than-yield-numbers"&gt;4. Samsung Electronics and HBM4E: Customer Shipments Matter More Than Yield Numbers
&lt;/h2&gt;&lt;p&gt;In early July, a widely circulated Telegram post reported that Samsung Electronics&amp;rsquo; HBM4E trial yield had exceeded 70%. The &lt;a class="link" href="https://t.me/corevalue/45816" target="_blank" rel="noopener"
 &gt;source post&lt;/a&gt; is a secondary document relaying news coverage and internal commentary. It is not a number confirmed by Samsung&amp;rsquo;s official IR or customer qualification results.&lt;/p&gt;
&lt;p&gt;Even if the 70% yield figure is accurate, three additional steps are required before it informs investment judgment.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;What process, stack configuration, and test stage does the yield refer to?&lt;/li&gt;
&lt;li&gt;Has it passed customer reliability testing and system-level qualification?&lt;/li&gt;
&lt;li&gt;Is commercial volume being shipped repeatedly and recognized as revenue?&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;In HBM, test yield, mass production yield, and customer qualification yield are not the same number. The most common error the market makes is mapping a strong internal test result directly to customer revenue.&lt;/p&gt;
&lt;p&gt;Samsung Electronics earnings forecasts circulating over the same period also require careful categorization. The &lt;a class="link" href="https://t.me/corevalue/45934" target="_blank" rel="noopener"
 &gt;Samsung memory-related post&lt;/a&gt; relays brokerage estimates and industry interpretation. Q2 operating income, bonus accruals, NAND pricing, enterprise SSD and KV-cache demand, and foundry customer discussions are all mixed into a single post. Items that are not official earnings releases or contract disclosures should remain as estimates.&lt;/p&gt;
&lt;p&gt;Nonetheless, some directional conclusions are valid.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;HBM shortages are reshuffling DRAM wafer allocation.&lt;/li&gt;
&lt;li&gt;As inference workloads grow, enterprise SSDs and NAND may become more critical at the lower tiers of the memory hierarchy.&lt;/li&gt;
&lt;li&gt;The investment thesis for Samsung Electronics is not solely about HBM share recovery, but about the optionality of combining DRAM, NAND, foundry, and packaging within a single company.&lt;/li&gt;
&lt;li&gt;The value of that optionality rises when customer qualification and recurring revenue are confirmed.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="5-taiwans-june-revenue-data-shows-ai-investment-diffusion"&gt;5. Taiwan&amp;rsquo;s June Revenue Data Shows AI Investment Diffusion
&lt;/h2&gt;&lt;p&gt;The strongest cross-verification in these two weeks came from the monthly revenue of Taiwan&amp;rsquo;s supply chain. It was not a single company&amp;rsquo;s guidance — rather, revenue at different nodes of the value chain moved in the same direction.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company / Segment&lt;/th&gt;
 &lt;th style="text-align: right"&gt;June 2026 Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;MoM&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YoY&lt;/th&gt;
 &lt;th&gt;What to Note&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;TSMC / Foundry&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$442.680B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+67.9%&lt;/td&gt;
 &lt;td&gt;Sustained demand for AI compute chips and leading-edge processes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Wiwynn / Server ODM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$111.371B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+32.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+29.8%&lt;/td&gt;
 &lt;td&gt;AI server and general server shipment expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Delta / Power &amp;amp; Cooling&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$65.603B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55.4%&lt;/td&gt;
 &lt;td&gt;Power supply unit and liquid cooling demand diffusing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EMC / Advanced CCL&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$17.733B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Increase per data&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+120.7%&lt;/td&gt;
 &lt;td&gt;Growth in advanced materials for AI servers and 800G switches&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;TUC / Advanced CCL&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$4.895B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+106.3%&lt;/td&gt;
 &lt;td&gt;Higher share of M7/M8 high-value products&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Co-Tech / HVLP Copper Foil&lt;/td&gt;
 &lt;td style="text-align: right"&gt;NT$0.962B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+43.4%&lt;/td&gt;
 &lt;td&gt;Growing bottleneck in low-roughness copper foil for high-speed signals&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;TSMC figures can be verified at the &lt;a class="link" href="https://investor.tsmc.com/english/monthly-revenue/2026" target="_blank" rel="noopener"
 &gt;official monthly revenue page&lt;/a&gt;. Wiwynn, Delta, EMC, TUC, and Co-Tech figures were cross-referenced against Telegram source posts relaying each company&amp;rsquo;s IR. The &lt;a class="link" href="https://t.me/merITz_tech/17574" target="_blank" rel="noopener"
 &gt;Wiwynn&lt;/a&gt;, &lt;a class="link" href="https://t.me/merITz_tech/17602" target="_blank" rel="noopener"
 &gt;Delta&lt;/a&gt;, and &lt;a class="link" href="https://t.me/merITz_tech/17598" target="_blank" rel="noopener"
 &gt;Co-Tech&lt;/a&gt; posts are the primary references.&lt;/p&gt;
&lt;h3 id="51-the-physical-bill-beyond-the-gpu"&gt;5.1 The Physical Bill Beyond the GPU
&lt;/h3&gt;&lt;p&gt;Adding one AI server requires more than a GPU.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Compute chip
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → HBM and DRAM
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → Server assembly and rack
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → Power supply unit and power conversion
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → Liquid cooling
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → High-speed PCB, CCL, HVLP copper foil
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → Optical modules and switches
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → Transformers, generation, grid interconnection
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;In 2024–2025, compute chips were the scarcest item. June&amp;rsquo;s monthly revenue data shows the bottleneck migrating to the periphery. The growth in CCL and HVLP copper foil in particular illustrates that materials reducing signal loss at higher data transfer rates gain value as speeds increase.&lt;/p&gt;
&lt;p&gt;Year-over-year growth rates alone should not be used to confirm the duration of supply shortages.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The prior-year base may be low.&lt;/li&gt;
&lt;li&gt;Price increases and volume growth must be separated.&lt;/li&gt;
&lt;li&gt;If new line ramp-ups drove revenue, the comparison base rises in subsequent quarters.&lt;/li&gt;
&lt;li&gt;Customer inventory building may be running ahead of end demand.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The correct verification sequence is therefore: &lt;code&gt;revenue → shipment volume → average selling price → inventory → operating cash flow&lt;/code&gt;. If revenue rises while inventory and accounts receivable grow faster, orders may be running ahead of final consumption.&lt;/p&gt;
&lt;h2 id="6-optical-networking-and-cpo-direction-is-right-numbers-need-more-verification"&gt;6. Optical Networking and CPO: Direction Is Right, Numbers Need More Verification
&lt;/h2&gt;&lt;p&gt;A &lt;a class="link" href="https://t.me/Yeouido_Lab/36085" target="_blank" rel="noopener"
 &gt;post on optical networking and CPO&lt;/a&gt; presented a brokerage forecast that the global optical networking market could grow from $15 billion in 2026 to $154 billion in 2028, with CPO accounting for $91 billion. The same source projected that network value and optical module counts per next-generation rack would increase substantially, and that silicon photonics adoption rates would surge.&lt;/p&gt;
&lt;p&gt;The directional case is persuasive. As GPU counts rise and data movement within and between racks grows, bandwidth and power constraints become increasingly difficult to resolve with electrical interconnects alone. Moving from 400G to 800G, 1.6T, and 3.2T makes light sources, lasers, silicon photonics, packaging, and precision alignment increasingly critical.&lt;/p&gt;
&lt;p&gt;However, the market size figures could not be independently verified against the original report and methodology. It is therefore safer to read them as follows:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;th&gt;Judgment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bandwidth becomes a bottleneck in AI infrastructure&lt;/td&gt;
 &lt;td&gt;High confidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Optical module and silicon photonics share increases&lt;/td&gt;
 &lt;td&gt;Medium-high confidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CPO market reaches $91B by 2028&lt;/td&gt;
 &lt;td&gt;Secondary source; further verification needed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Specific Korean companies immediately capture large revenue&lt;/td&gt;
 &lt;td&gt;Unconfirmed until customer, component, and mass production disclosures&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;In Korean equities, as covered in the &lt;a class="link" href="https://koreainvestinsights.com/ko/post/korea-optical-cpo-value-chain-seven-companies-2026-05-09/" &gt;optical networking and CPO value chain analysis&lt;/a&gt;, actual products and customer qualifications matter more than thematic labels. A company whose name includes &amp;ldquo;optical&amp;rdquo; is not the same company as one supplying components into 1.6T optical links in an AI cluster.&lt;/p&gt;
&lt;h2 id="7-the-power-bottleneck-is-real-metas-5-gw-and-brookfields-25-billion"&gt;7. The Power Bottleneck Is Real: Meta&amp;rsquo;s 5 GW and Brookfield&amp;rsquo;s $25 Billion
&lt;/h2&gt;&lt;p&gt;Even with strong memory and server revenues, a data center generates no revenue if it cannot connect power. The clearest official data on this point from the first half of July came from Meta and Brookfield.&lt;/p&gt;
&lt;h3 id="71-meta-hyperion"&gt;7.1 Meta Hyperion
&lt;/h3&gt;&lt;p&gt;Meta&amp;rsquo;s Louisiana data center expansion plan outlined &lt;a class="link" href="https://about.fb.com/news/2026/07/teachers-local-businesses-win-as-meta-expands-louisiana-data-center/" target="_blank" rel="noopener"
 &gt;up to 5 GW of compute capacity and more than $50 billion in investment&lt;/a&gt;, including more than $1 billion in local infrastructure, seven new gas power plants, three grid battery storage systems, nuclear output uprates, and power purchase agreements — with Meta bearing the costs of power, water, and infrastructure.&lt;/p&gt;
&lt;p&gt;The importance of this announcement is not merely the scale of the numbers. It reflects the fact that AI data centers are evolving from semiconductor procurement projects into &lt;strong&gt;regional power system redesign projects&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;Meta also announced in 2025 a joint venture with Blue Owl for Hyperion development and a structure to raise approximately $27 billion in development financing. The &lt;a class="link" href="https://about.fb.com/news/2025/10/meta-blue-owl-capital-develop-hyperion-data-center/" target="_blank" rel="noopener"
 &gt;official announcement&lt;/a&gt; shows that even big tech prefers to share capital and risk with external investors rather than putting all facilities directly on its own balance sheet.&lt;/p&gt;
&lt;h3 id="72-brookfield-and-bloom-energy"&gt;7.2 Brookfield and Bloom Energy
&lt;/h3&gt;&lt;p&gt;Brookfield and Bloom Energy expanded their AI infrastructure power project financing framework from an existing $5 billion to &lt;a class="link" href="https://investor.bloomenergy.com/press-releases/press-release-details/2026/Brookfield-and-Bloom-Energy-Expand-AI-Infrastructure-Partnership-to-25-Billion-Fivefold-Increase-to-Build-and-Finance-Rapid-Power-for-AI-Infrastructure/default.aspx" target="_blank" rel="noopener"
 &gt;$25 billion&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;This $25 billion should not be read as confirmed Bloom Energy revenue. A financing framework shows the ceiling and commitment of capital that can be deployed into projects. Actual revenue requires land, customers, power purchase agreements, fuel supply, equipment orders, and groundbreaking.&lt;/p&gt;
&lt;p&gt;Nevertheless, the signal quality is high.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Data centers find it difficult to wait for grid interconnection and are seeking on-site power solutions.&lt;/li&gt;
&lt;li&gt;Equipment vendor orders alone are insufficient — long-term financing must accompany them.&lt;/li&gt;
&lt;li&gt;Capital cost and contract structure may determine project cadence more than equipment availability.&lt;/li&gt;
&lt;li&gt;Profit sharing among big tech, infrastructure funds, and power operators becomes increasingly important.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="8-same-power-theme-different-shareholder-outcomes-fermis-convertible-notes"&gt;8. Same Power Theme, Different Shareholder Outcomes: Fermi&amp;rsquo;s Convertible Notes
&lt;/h2&gt;&lt;p&gt;Strong power and data center demand does not mean all related equities perform well. The &lt;a class="link" href="https://t.me/Yeouido_Lab/36081" target="_blank" rel="noopener"
 &gt;Fermi-related Telegram post&lt;/a&gt; illustrated this counterexample clearly.&lt;/p&gt;
&lt;p&gt;Fermi priced $375 million of 5% convertible senior notes due 2031, with a potential over-allotment bringing the total to $431.25 million. Reading the &lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2071778/000121390026076788/ea0297535-8k_fermi.htm" target="_blank" rel="noopener"
 &gt;SEC filing&lt;/a&gt; and &lt;a class="link" href="https://www.nasdaq.com/press-release/fermi-inc-prices-upsized-offering-375-million-convertible-senior-notes-dilution" target="_blank" rel="noopener"
 &gt;final pricing release&lt;/a&gt; together, it is clear that the company is at the stage of funding large-scale data center ambitions through capital markets before those ambitions generate revenue.&lt;/p&gt;
&lt;p&gt;When evaluating power themes, announced GW targets and confirmed cash flows must be distinguished.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stage&lt;/th&gt;
 &lt;th&gt;Evidence to Verify&lt;/th&gt;
 &lt;th&gt;Meaning for Shareholders&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Concept&lt;/td&gt;
 &lt;td&gt;Site, target GW, development plan&lt;/td&gt;
 &lt;td&gt;Optionality exists but no revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Contract&lt;/td&gt;
 &lt;td&gt;Customer, PPA, minimum usage commitments&lt;/td&gt;
 &lt;td&gt;Revenue visibility begins&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Financing&lt;/td&gt;
 &lt;td&gt;Interest rate, collateral, conversion price, maturity&lt;/td&gt;
 &lt;td&gt;Dilution and refinancing risk determined&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Construction&lt;/td&gt;
 &lt;td&gt;EPC, equipment orders, permits&lt;/td&gt;
 &lt;td&gt;Cost and schedule risk materializes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operations&lt;/td&gt;
 &lt;td&gt;Utilization, power cost, load factor&lt;/td&gt;
 &lt;td&gt;Operating cash flow confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Brookfield&amp;rsquo;s $25 billion and Fermi&amp;rsquo;s $375 million are both AI power investments. The former shows the strength of long-term infrastructure financing; the latter shows the dilution cost borne by shareholders of a development-stage company with no revenue yet. Even within the same theme, the quality of capital determines shareholder returns.&lt;/p&gt;
&lt;h2 id="9-metas-excess-compute-sales-evidence-of-oversupply-or-an-option"&gt;9. Meta&amp;rsquo;s Excess Compute Sales: Evidence of Oversupply or an Option?
&lt;/h2&gt;&lt;p&gt;A &lt;a class="link" href="https://t.me/eqmirae/3050" target="_blank" rel="noopener"
 &gt;Meta cloud-related post&lt;/a&gt; covered reports that Meta is exploring selling AI computing capacity beyond its own internal use to external customers. This plan is based on a &lt;a class="link" href="https://news.bloomberglaw.com/tech-and-telecom-law/meta-is-building-a-cloud-business-to-sell-excess-ai-compute-1" target="_blank" rel="noopener"
 &gt;Bloomberg report&lt;/a&gt; citing anonymous sources at the planning stage. No confirmed product, pricing, or contract scale has been disclosed.&lt;/p&gt;
&lt;p&gt;This news can be read two ways.&lt;/p&gt;
&lt;h3 id="bearish-reading"&gt;Bearish Reading
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Meta has secured more capacity than it needs.&lt;/li&gt;
&lt;li&gt;The pace of monetization from internal AI services has not kept up with capacity expansion.&lt;/li&gt;
&lt;li&gt;Selling surplus compute signals a coming slowdown in capital expenditure.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="bullish-reading"&gt;Bullish Reading
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Training and inference demand varies by time-of-day and project.&lt;/li&gt;
&lt;li&gt;Selling unused capacity to external customers improves asset turnover and return on investment.&lt;/li&gt;
&lt;li&gt;It creates an incremental revenue stream beyond advertising.&lt;/li&gt;
&lt;li&gt;It lowers the downside of large-scale preemptive investment, creating an embedded option.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Current data alone cannot determine which reading is correct. The necessary indicators are:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;External sales volume and pricing&lt;/li&gt;
&lt;li&gt;GPU utilization rate for internal workloads&lt;/li&gt;
&lt;li&gt;Contribution margin inclusive of depreciation and power costs&lt;/li&gt;
&lt;li&gt;Additional capex guidance&lt;/li&gt;
&lt;li&gt;Minimum usage commitments from customers and contract duration&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The ability to sell surplus capacity reveals the possibility of overcapacity but is not itself evidence of overcapacity. In the best case, it is monetization of underutilized assets; in the worst case, it is a post-hoc response to overestimated internal demand.&lt;/p&gt;
&lt;h2 id="10-the-agent-era-and-memory-demand-translating-narrative-into-physical-quantities"&gt;10. The Agent Era and Memory Demand: Translating Narrative into Physical Quantities
&lt;/h2&gt;&lt;p&gt;A &lt;a class="link" href="https://t.me/Yeouido_Lab/35634" target="_blank" rel="noopener"
 &gt;long-term outlook post on agents and memory&lt;/a&gt; presents a vision where AI advances through infrastructure construction, operations, an agent economy, and inter-agent transactions, making memory the foundation of civilization. Numbers like 10 billion agents are not validated forecasts — they are long-term narrative.&lt;/p&gt;
&lt;p&gt;But the question the post raises is valid. Agents use longer state and more tool calls than a chatbot that answers and exits. Breaking work into multiple steps, retaining prior results, and exchanging information with other agents can increase total inference volume and stored state.&lt;/p&gt;
&lt;p&gt;To translate this into memory demand, five variables must be tracked separately.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Total memory demand
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;≈ Number of useful tasks
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;× Inference steps per task
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;× Active context per step
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;× Fraction of context residing in memory
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;÷ Compression, cache, quantization, and routing efficiency
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The bull case emphasizes the first four variables. The counter-argument is that the efficiency denominator is improving rapidly.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;KV cache compression and reuse&lt;/li&gt;
&lt;li&gt;Speculative decoding&lt;/li&gt;
&lt;li&gt;Routing between small and large models&lt;/li&gt;
&lt;li&gt;Mixture-of-experts selective activation&lt;/li&gt;
&lt;li&gt;Tiered storage across HBM, DRAM, CXL, and SSD&lt;/li&gt;
&lt;li&gt;Prompt and state summarization&lt;/li&gt;
&lt;li&gt;Hardware-software co-design&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Therefore, &lt;code&gt;agents growing = HBM demand growing at the same rate&lt;/code&gt; is not a law of physics. A better benchmark is &lt;strong&gt;memory bytes-seconds per useful task completed, multiplied by total task volume&lt;/strong&gt;. If efficiency improves tenfold but total task volume grows a hundredfold, demand still increases. Conversely, if agent counts rise but most run on short, low-cost models, HBM demand may fall well short of the narrative.&lt;/p&gt;
&lt;h2 id="11-who-is-the-ultimate-payer-ai-software-profitability"&gt;11. Who Is the Ultimate Payer: AI Software Profitability
&lt;/h2&gt;&lt;p&gt;The infrastructure investment cycle can be sustained only if end customers pay the cost. In early July, a &lt;a class="link" href="https://t.me/gaoshoukorea/62002" target="_blank" rel="noopener"
 &gt;Telegram post&lt;/a&gt; shared that Anthropic is profitable in Q2 on a stock-based compensation-excluded basis, and that token-level gross margins on some models exceed 80%.&lt;/p&gt;
&lt;p&gt;The sources are a &lt;a class="link" href="https://sequoiacap.com/podcast/dylan-patel-of-semianalysis-why-hardware-software-co-design-is-ais-real-100x/" target="_blank" rel="noopener"
 &gt;Sequoia interview with Dylan Patel&lt;/a&gt; and &lt;a class="link" href="https://newsletter.semianalysis.com/p/anthropic-growth-and-bedrock-mix" target="_blank" rel="noopener"
 &gt;SemiAnalysis estimates&lt;/a&gt;. These are not Anthropic&amp;rsquo;s audited financial statements. They should be read as interview statements and model-based estimates.&lt;/p&gt;
&lt;p&gt;In particular, API list-price token margins and company-wide profitability are different things.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Included in Token-Level Margin&lt;/th&gt;
 &lt;th&gt;What Is Additionally Required for Company-Wide Profit&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Inference revenue&lt;/td&gt;
 &lt;td&gt;Included&lt;/td&gt;
 &lt;td&gt;Reflect channel discounts and commitments&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Inference compute&lt;/td&gt;
 &lt;td&gt;Included&lt;/td&gt;
 &lt;td&gt;Reflect idle capacity and peak costs&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud distribution&lt;/td&gt;
 &lt;td&gt;Partial estimate&lt;/td&gt;
 &lt;td&gt;Bedrock/Vertex revenue sharing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;R&amp;amp;D&lt;/td&gt;
 &lt;td&gt;May be excluded&lt;/td&gt;
 &lt;td&gt;Large-scale training and headcount&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stock-based compensation&lt;/td&gt;
 &lt;td&gt;May be excluded&lt;/td&gt;
 &lt;td&gt;Fully diluted cost&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sales &amp;amp; marketing&lt;/td&gt;
 &lt;td&gt;May be excluded&lt;/td&gt;
 &lt;td&gt;Enterprise sales and customer support&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Even so, the relevance of the software profitability discussion is clear. If end services generate high contribution margins, the capex on GPUs, memory, networking, and power can be sustained. Conversely, if token price competition intensifies and revenue per task falls, it becomes harder for infrastructure providers to maintain high pricing for long.&lt;/p&gt;
&lt;p&gt;The true end of the AI supply chain is not HBM shipments.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Amount paid by users
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → AI service gross margin
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → Cloud and model provider capex capacity
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; → GPU, memory, networking, and power orders
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;If cash runs short at any link in this chain, the backlog upstream gets revised later.&lt;/p&gt;
&lt;h2 id="12-koreas-separate-variable-single-stock-leveraged-products"&gt;12. Korea&amp;rsquo;s Separate Variable: Single-Stock Leveraged Products
&lt;/h2&gt;&lt;p&gt;Source documents on the Korean equity market repeatedly claimed that single-stock leveraged products tied to Samsung Electronics and SK Hynix have been absorbing positioning flows and amplifying volatility. A &lt;a class="link" href="https://t.me/corevalue/46025" target="_blank" rel="noopener"
 &gt;representative post&lt;/a&gt; cited cumulative retail buying of ₩14.2 trillion since product launch. This specific cumulative figure and its measurement scope could not be independently verified.&lt;/p&gt;
&lt;p&gt;Separately, media reporting noted that during the four trading days around the June sell-off, retail investors net-bought ₩14.3 trillion in KOSPI and deployed ₩10.7 trillion into Samsung Electronics and SK Hynix. The &lt;a class="link" href="https://v.daum.net/v/20260625060155479" target="_blank" rel="noopener"
 &gt;relevant report&lt;/a&gt; is the reference. This figure covers a different period and scope than the &amp;ldquo;₩14.2 trillion cumulative since product launch&amp;rdquo; figure. The two should not be treated as the same statistic.&lt;/p&gt;
&lt;p&gt;What is officially confirmed is the existence of the products and the regulatory risk warnings.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;The Financial Supervisory Service warned on May 27, prior to launch, about the &lt;a class="link" href="https://dart.fss.or.kr/dsaa003/selectBodoMain.ax?seqno=28181" target="_blank" rel="noopener"
 &gt;loss amplification and negative compounding effects of single-stock leveraged and inverse products&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;KRX listed the &lt;a class="link" href="https://kind.krx.co.kr/external/2026/05/22/000241/20260522000586/68152.htm" target="_blank" rel="noopener"
 &gt;Samsung Electronics single-stock leveraged ETF&lt;/a&gt; and an &lt;a class="link" href="https://kind.krx.co.kr/external/2026/05/22/000553/20260522001367/68342.htm" target="_blank" rel="noopener"
 &gt;SK Hynix leveraged ETN&lt;/a&gt;.&lt;/li&gt;
&lt;li&gt;KRX issued an &lt;a class="link" href="https://kind.krx.co.kr/external/2026/06/08/000981/20260608001993/99355.htm" target="_blank" rel="noopener"
 &gt;investor caution&lt;/a&gt; regarding the premium/discount to NAV in an SK Hynix futures-leveraged ETF.&lt;/li&gt;
&lt;li&gt;Creation and redemption disclosures for related ETFs appeared in July, indicating that product flows can move in either direction rather than representing fixed buying demand.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="121-why-volatility-increases"&gt;12.1 Why Volatility Increases
&lt;/h3&gt;&lt;p&gt;Leveraged products must rebalance daily to maintain their target multiple. When the underlying rises, they may need to add exposure; when it falls sharply, they may need to reduce it. The exact timing of transactions depends on the mix of futures, swaps, and equity hedges, but the structural tendency is to amplify both upside and downside.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Rising market
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Leveraged product NAV increases
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Exposure expansion needed to maintain target multiple
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Potential additional demand for underlying
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Falling market
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ NAV decline and redemptions
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Hedge reduction and potential futures selling
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Amplified underlying volatility
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;This structure does not turn good earnings into bad ones. But it can steepen the price path even on days when fundamentals have not changed. Analyzing Samsung Electronics and SK Hynix therefore requires monitoring ETF NAV, creation and redemption flows, futures basis, and program trading flows alongside earnings forecasts.&lt;/p&gt;
&lt;h2 id="13-sk-hynix-adr-the-premium-may-be-the-price-of-friction-not-of-enterprise-value"&gt;13. SK Hynix ADR: The Premium May Be the Price of Friction, Not of Enterprise Value
&lt;/h2&gt;&lt;p&gt;A &lt;a class="link" href="https://t.me/Yeouido_Lab/36093" target="_blank" rel="noopener"
 &gt;post on SK Hynix ADR&lt;/a&gt; argued that when real-time arbitrage between the U.S. ADR and the Korean ordinary shares is constrained by settlement lag, taxes, and regulations, an initial ADR premium can emerge. The post also cited a TSMC-like premium range for reference.&lt;/p&gt;
&lt;p&gt;This interpretation is plausible but not confirmed. ADR pricing can be framed as follows:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;ADR price
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;≈ Ordinary share price × conversion ratio × FX rate
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ U.S. investor accessibility premium
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ borrow, settlement, tax, and conversion friction
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;- new supply and arbitrage pressure
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Initially, if U.S. investor demand is fast and the supply of ordinaries converted to ADRs is slow, a premium can emerge. But that premium does not mean SK Hynix&amp;rsquo;s operating income has increased. It is the market structure&amp;rsquo;s price for friction — one that is likely to compress as conversion and arbitrage normalize.&lt;/p&gt;
&lt;p&gt;For an ADR to create long-term value, at least one of the following is needed:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;An expanded U.S. investor base lowers the required rate of return.&lt;/li&gt;
&lt;li&gt;Higher trading liquidity and analyst coverage add a monitoring premium.&lt;/li&gt;
&lt;li&gt;U.S. capital markets provide more favorable financing for capex.&lt;/li&gt;
&lt;li&gt;Index inclusion or the lifting of institutional investment restrictions creates real incremental demand.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Conversely, looking only at the initial premium and double-counting enterprise value across ordinary shares and the ADR would be an error. As discussed in the &lt;a class="link" href="https://koreainvestinsights.com/ko/post/sk-hynix-adr-000660-price-path-leverage-etf-plumbing-2026-07-04/" &gt;SK Hynix ADR and leverage structure analysis&lt;/a&gt;, the ADR is plumbing layered on top of fundamentals. Plumbing changes price discovery but does not automatically generate cash flow.&lt;/p&gt;
&lt;h2 id="14-when-four-signals-collide"&gt;14. When Four Signals Collide
&lt;/h2&gt;&lt;p&gt;The source documents from the first half of July do not all point in the same direction. The four key conflicts are what matter most.&lt;/p&gt;
&lt;h3 id="conflict-1-memory-prices-are-strong-but-volume-is-still-weak"&gt;Conflict 1: Memory Prices Are Strong But Volume Is Still Weak
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Nanya&amp;rsquo;s prices rose more than 60%.&lt;/li&gt;
&lt;li&gt;Shipment volume barely increased.&lt;/li&gt;
&lt;li&gt;Earnings exploded but peak-price risk also increased.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Judgment: In the next quarter, watch bit shipments and customer inventory more than the rate of price appreciation.&lt;/p&gt;
&lt;h3 id="conflict-2-ai-component-revenue-is-diffusing-but-power-and-financing-costs-are-rising-too"&gt;Conflict 2: AI Component Revenue Is Diffusing, But Power and Financing Costs Are Rising Too
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Server ODM, power supply, CCL, and copper foil revenues are strong.&lt;/li&gt;
&lt;li&gt;Meta plans 5 GW and more than $50 billion in investment.&lt;/li&gt;
&lt;li&gt;Power financing has expanded to $25 billion.&lt;/li&gt;
&lt;li&gt;Development-stage companies rely on convertible notes and dilutive capital.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Judgment: Watch contract counterparties, cost of capital, and free cash flow more than revenue growth rates.&lt;/p&gt;
&lt;h3 id="conflict-3-agent-demand-is-growing-but-per-unit-memory-cost-is-falling"&gt;Conflict 3: Agent Demand Is Growing, But Per-Unit Memory Cost Is Falling
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Long-horizon state, tool use, and inter-agent communication increase total inference volume.&lt;/li&gt;
&lt;li&gt;Compression, caching, tiered storage, and small-model routing lower cost per byte.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Judgment: Watch cost per useful task and total task volume together, not just agent count.&lt;/p&gt;
&lt;h3 id="conflict-4-korean-memory-companies-earnings-are-strong-but-the-price-plumbing-is-unstable"&gt;Conflict 4: Korean Memory Companies&amp;rsquo; Earnings Are Strong, But the Price Plumbing Is Unstable
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Memory prices and cash flows are strong.&lt;/li&gt;
&lt;li&gt;Single-stock leverage products and ADR expectations amplify the price path.&lt;/li&gt;
&lt;li&gt;Even strong companies can experience sharp drawdowns from redemptions, hedge unwinds, and premium normalization.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Judgment: Earnings estimates and position sizing must be kept separate. Even with high conviction in the business, an unstable positioning structure requires a different approach to entry and sizing.&lt;/p&gt;
&lt;h2 id="15-scenarios-from-h2-2026-through-2028"&gt;15. Scenarios from H2 2026 Through 2028
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;H2 2026&lt;/th&gt;
 &lt;th&gt;2027&lt;/th&gt;
 &lt;th&gt;2028&lt;/th&gt;
 &lt;th&gt;Beneficiary&lt;/th&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;A. Orderly excess demand&lt;/td&gt;
 &lt;td&gt;Memory prices and Taiwan revenue stay strong&lt;/td&gt;
 &lt;td&gt;Growth continues via long-term contracts and power expansion&lt;/td&gt;
 &lt;td&gt;New supply absorbed gradually&lt;/td&gt;
 &lt;td&gt;HBM, eSSD, power supply, cooling, advanced materials&lt;/td&gt;
 &lt;td&gt;Overvalued option-like developers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;B. Strong prices, flat volume&lt;/td&gt;
 &lt;td&gt;ASP drives earnings&lt;/td&gt;
 &lt;td&gt;Customer inventory and negotiating leverage weigh&lt;/td&gt;
 &lt;td&gt;Price normalization causes earnings plunge&lt;/td&gt;
 &lt;td&gt;Low-cost producers, cash-rich companies&lt;/td&gt;
 &lt;td&gt;Producers with high price sensitivity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C. Power and financing delays&lt;/td&gt;
 &lt;td&gt;Chip orders hold but site delays accumulate&lt;/td&gt;
 &lt;td&gt;Some projects cancelled or renegotiated&lt;/td&gt;
 &lt;td&gt;Demand reconcentrates on large operators&lt;/td&gt;
 &lt;td&gt;Companies with long-term PPAs and strong capital&lt;/td&gt;
 &lt;td&gt;Developers without pre-contracted revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;D. Simultaneous supply and efficiency improvement&lt;/td&gt;
 &lt;td&gt;Bull narrative persists&lt;/td&gt;
 &lt;td&gt;Efficiency gains partially offset demand growth&lt;/td&gt;
 &lt;td&gt;New fabs and efficiency gains overlap&lt;/td&gt;
 &lt;td&gt;Low-cost platforms, software&lt;/td&gt;
 &lt;td&gt;Commodity memory and high-fixed-cost new fabs&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;E. Leverage unwind&lt;/td&gt;
 &lt;td&gt;Repeated sharp declines disconnected from earnings&lt;/td&gt;
 &lt;td&gt;Possible product structure adjustment&lt;/td&gt;
 &lt;td&gt;Prices re-converge to fundamentals&lt;/td&gt;
 &lt;td&gt;Companies with FCF and buyback capacity&lt;/td&gt;
 &lt;td&gt;High-leverage chase positions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The most probable path at present is a mix of A and B. Memory prices and AI component revenues are likely to remain strong in the near term, but whether volume growth and final-service monetization keep pace with price appreciation must be confirmed. 2028 is the first major verification window, when new fabs, power project startups, AI efficiency improvements, and long-term contract renewals all converge.&lt;/p&gt;
&lt;h2 id="16-indicators-to-watch-a-checklist-for-converting-narrative-into-numbers"&gt;16. Indicators to Watch: A Checklist for Converting Narrative into Numbers
&lt;/h2&gt;&lt;h3 id="memory"&gt;Memory
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Indicator&lt;/th&gt;
 &lt;th&gt;Bull Confirmation&lt;/th&gt;
 &lt;th&gt;Caution Signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM ASP and shipment volume&lt;/td&gt;
 &lt;td&gt;Price and bit shipments both rising&lt;/td&gt;
 &lt;td&gt;Price rising alone, shipments stagnant&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer inventory&lt;/td&gt;
 &lt;td&gt;Low inventory and expanding long-term contracts&lt;/td&gt;
 &lt;td&gt;Front-loading orders followed by inventory surge&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM qualification&lt;/td&gt;
 &lt;td&gt;Repeated commercial shipments and expanding customer base&lt;/td&gt;
 &lt;td&gt;Only test yields disclosed, revenue delayed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND and eSSD&lt;/td&gt;
 &lt;td&gt;Price, volume, and margin improving together&lt;/td&gt;
 &lt;td&gt;Only HBM narrative strong while NAND inventory builds&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Capex&lt;/td&gt;
 &lt;td&gt;Phased expansion matched to demand&lt;/td&gt;
 &lt;td&gt;Depreciation growing faster than revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="ai-hardware-supply-chain"&gt;AI Hardware Supply Chain
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Indicator&lt;/th&gt;
 &lt;th&gt;Bull Confirmation&lt;/th&gt;
 &lt;th&gt;Caution Signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Taiwan monthly revenue&lt;/td&gt;
 &lt;td&gt;Multiple nodes growing together for 3+ months&lt;/td&gt;
 &lt;td&gt;Only one or two companies growing, others slowing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Advanced materials&lt;/td&gt;
 &lt;td&gt;Volume and high-value mix both rising&lt;/td&gt;
 &lt;td&gt;Revenue growth driven only by price increases&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Inventory and receivables&lt;/td&gt;
 &lt;td&gt;Growing at roughly the same pace as revenue&lt;/td&gt;
 &lt;td&gt;Growing faster than revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Optical networking&lt;/td&gt;
 &lt;td&gt;800G and 1.6T volume production confirmed by customers&lt;/td&gt;
 &lt;td&gt;Only market size forecasts expanding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="power-and-financing"&gt;Power and Financing
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Indicator&lt;/th&gt;
 &lt;th&gt;Bull Confirmation&lt;/th&gt;
 &lt;th&gt;Caution Signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Data centers&lt;/td&gt;
 &lt;td&gt;PPA, named customers, minimum usage commitments&lt;/td&gt;
 &lt;td&gt;Only target GW announced, customers undisclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Funding&lt;/td&gt;
 &lt;td&gt;Long-term low-cost capital, risk sharing&lt;/td&gt;
 &lt;td&gt;High-yield debt, convertibles, short maturities&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cash flow&lt;/td&gt;
 &lt;td&gt;FCF conversion after commissioning&lt;/td&gt;
 &lt;td&gt;Capex consistently overwhelming FCF&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power costs&lt;/td&gt;
 &lt;td&gt;Locked in via long-term contracts, predictable&lt;/td&gt;
 &lt;td&gt;Spot fuel prices and grid costs surging&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="korean-market-structure"&gt;Korean Market Structure
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Indicator&lt;/th&gt;
 &lt;th&gt;Bull Confirmation&lt;/th&gt;
 &lt;th&gt;Caution Signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Leveraged products&lt;/td&gt;
 &lt;td&gt;NAV premium stable, creation/redemption gradual&lt;/td&gt;
 &lt;td&gt;Premium widening, sharp redemptions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Futures and program flow&lt;/td&gt;
 &lt;td&gt;Balanced underlying and futures supply/demand&lt;/td&gt;
 &lt;td&gt;Concentrated end-of-day hedge trading&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADR&lt;/td&gt;
 &lt;td&gt;Smooth conversion, premium stable&lt;/td&gt;
 &lt;td&gt;Initial premium spikes then new supply floods in&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Retail positioning&lt;/td&gt;
 &lt;td&gt;Cash-based diversified buying&lt;/td&gt;
 &lt;td&gt;Concentrated margin and leverage use&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="17-what-to-believe-and-what-to-wait-for"&gt;17. What to Believe and What to Wait For
&lt;/h2&gt;&lt;h3 id="high-confidence-conclusions"&gt;High-Confidence Conclusions
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;Memory price appreciation is materially lifting producer quarterly earnings.&lt;/li&gt;
&lt;li&gt;AI investment has diffused into servers, power supplies, cooling, and advanced PCB materials.&lt;/li&gt;
&lt;li&gt;Power and long-term capital have become the critical bottlenecks in data center construction.&lt;/li&gt;
&lt;li&gt;Single-stock leveraged products can amplify the daily volatility of the underlying asset.&lt;/li&gt;
&lt;li&gt;New memory supply coming online around 2028 is the direct consequence of today&amp;rsquo;s shortage.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="directionally-sound-but-requiring-further-verification"&gt;Directionally Sound But Requiring Further Verification
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;Samsung Electronics HBM4E trial yield of 70% and per-customer qualification status&lt;/li&gt;
&lt;li&gt;Scale and profitability of Meta&amp;rsquo;s external AI compute sales&lt;/li&gt;
&lt;li&gt;CPO market size in 2028 and Korean companies&amp;rsquo; actual share&lt;/li&gt;
&lt;li&gt;Initial ADR premium range for SK Hynix and its expected duration&lt;/li&gt;
&lt;li&gt;How much the agent economy will independently grow HBM, DRAM, and SSD demand, respectively&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="assertions-that-cannot-be-supported-by-current-evidence"&gt;Assertions That Cannot Be Supported by Current Evidence
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;Inferring that something is true because it appeared frequently in Telegram&lt;/li&gt;
&lt;li&gt;Interpreting the $25 billion financing framework as $25 billion in confirmed revenue for a specific equipment maker&lt;/li&gt;
&lt;li&gt;Assuming all announced data center GW will be commissioned&lt;/li&gt;
&lt;li&gt;Interpreting ADR premiums as evidence of increased enterprise value&lt;/li&gt;
&lt;li&gt;Claiming that agent count and HBM demand grow at the same rate&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="18-final-assessment"&gt;18. Final Assessment
&lt;/h2&gt;&lt;p&gt;Two weeks of Telegram data cannot be reduced to a single bull or bear thesis. Instead, the data reveals that the market is simultaneously pricing four different time horizons.&lt;/p&gt;
&lt;p&gt;First, this is the time of price. As Nanya&amp;rsquo;s results show, memory prices are driving earnings higher faster than volume can. In this window, producers&amp;rsquo; operating leverage is powerful.&lt;/p&gt;
&lt;p&gt;Second, next comes the time of volume diffusion. Monthly revenue from TSMC, Wiwynn, Delta, CCL makers, and HVLP copper foil producers shows that the AI investment invoice is spreading beyond GPUs. But inventory and cash flows must follow.&lt;/p&gt;
&lt;p&gt;Third, after that comes the time of power and financing. Meta&amp;rsquo;s 5 GW and Brookfield&amp;rsquo;s $25 billion show the scale. Fermi&amp;rsquo;s convertible notes ask how much of that scale comes back to shareholders as dilution cost.&lt;/p&gt;
&lt;p&gt;Fourth, 2028 is the time of supply and efficiency. The fabs being built today by high prices, the generation assets being built today by high power demand, and the efficiency gains being made today by high token costs will all enter the market simultaneously.&lt;/p&gt;
&lt;p&gt;What investors should do is not abandon AI infrastructure or pursue it unconditionally. It is to verify where each company sits in the bottleneck structure, whether its growth is driven by price or volume, and who bears the cost and financing burden.&lt;/p&gt;
&lt;p&gt;The best companies own the bottleneck, have low cost of capital, generate recurring revenue and cash flow, and can expand capacity in stages. The most dangerous companies announce enormous target capacities but have weak customer contracts and thin financing, and convert stock price appreciation back into dilutive capital.&lt;/p&gt;
&lt;p&gt;Telegram does not reveal this distinction in advance. Telegram shows where people are running. From that point, official disclosures, income statements, cash flow statements, and the terms of contracts and financing must be read directly.&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Two-Week Deep Dive Conclusion&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Demand for AI infrastructure remains strong. But the excess returns flow not from the "AI" label itself, but to &lt;strong&gt;memory producers earning today's prices, component makers whose actual volumes are growing, and operators who provide both power and capital&lt;/strong&gt;. Watch the shortage of 2026–2027, but also watch simultaneously for the supply and depreciation that shortage is building toward 2028.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="key-source-documents-and-official-references"&gt;Key Source Documents and Official References
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://t.me/Yeouido_Lab/36066" target="_blank" rel="noopener"
 &gt;Nanya Technology FY2Q26 Telegram source&lt;/a&gt;, &lt;a class="link" href="https://www.nanya.com/tw/IR/14/%E6%B4%BB%E5%8B%95%E8%A1%8C%E4%BA%8B%E6%9B%86?IRId=4295" target="_blank" rel="noopener"
 &gt;Nanya Technology official IR&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://investor.tsmc.com/english/monthly-revenue/2026" target="_blank" rel="noopener"
 &gt;TSMC 2026 monthly revenue&lt;/a&gt;, &lt;a class="link" href="https://investor.tsmc.com/english/financial-calendar" target="_blank" rel="noopener"
 &gt;TSMC financial calendar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://t.me/merITz_tech/17574" target="_blank" rel="noopener"
 &gt;Wiwynn June revenue source&lt;/a&gt;, &lt;a class="link" href="https://t.me/merITz_tech/17602" target="_blank" rel="noopener"
 &gt;Delta June revenue source&lt;/a&gt;, &lt;a class="link" href="https://t.me/merITz_tech/17598" target="_blank" rel="noopener"
 &gt;Co-Tech June revenue source&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://about.fb.com/news/2026/07/teachers-local-businesses-win-as-meta-expands-louisiana-data-center/" target="_blank" rel="noopener"
 &gt;Meta Louisiana data center official announcement&lt;/a&gt;, &lt;a class="link" href="https://about.fb.com/news/2025/10/meta-blue-owl-capital-develop-hyperion-data-center/" target="_blank" rel="noopener"
 &gt;Meta · Blue Owl joint venture official announcement&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://investor.bloomenergy.com/press-releases/press-release-details/2026/Brookfield-and-Bloom-Energy-Expand-AI-Infrastructure-Partnership-to-25-Billion-Fivefold-Increase-to-Build-and-Finance-Rapid-Power-for-AI-Infrastructure/default.aspx" target="_blank" rel="noopener"
 &gt;Brookfield and Bloom Energy $25 billion financing framework&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2071778/000121390026076788/ea0297535-8k_fermi.htm" target="_blank" rel="noopener"
 &gt;Fermi SEC filing&lt;/a&gt;, &lt;a class="link" href="https://www.nasdaq.com/press-release/fermi-inc-prices-upsized-offering-375-million-convertible-senior-notes-dilution" target="_blank" rel="noopener"
 &gt;Convertible notes final pricing&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://investors.micron.com/news-releases/news-release-details/micron-accelerates-us-investments-pours-first-concrete-new-york" target="_blank" rel="noopener"
 &gt;Micron U.S. investment official announcement&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://dart.fss.or.kr/dsaa003/selectBodoMain.ax?seqno=28181" target="_blank" rel="noopener"
 &gt;FSS single-stock leveraged and inverse product warning&lt;/a&gt;, &lt;a class="link" href="https://kind.krx.co.kr/external/2026/06/08/000981/20260608001993/99355.htm" target="_blank" rel="noopener"
 &gt;KRX NAV premium/discount investor caution&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://news.bloomberglaw.com/tech-and-telecom-law/meta-is-building-a-cloud-business-to-sell-excess-ai-compute-1" target="_blank" rel="noopener"
 &gt;Meta excess compute sales report&lt;/a&gt;, &lt;a class="link" href="https://t.me/eqmirae/3050" target="_blank" rel="noopener"
 &gt;Related Telegram interpretation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://sequoiacap.com/podcast/dylan-patel-of-semianalysis-why-hardware-software-co-design-is-ais-real-100x/" target="_blank" rel="noopener"
 &gt;AI software profitability interview&lt;/a&gt;, &lt;a class="link" href="https://newsletter.semianalysis.com/p/anthropic-growth-and-bedrock-mix" target="_blank" rel="noopener"
 &gt;SemiAnalysis estimates&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Data as of July 1–14, 2026. Channel and message counts are based on collected public Telegram posts; July 14 may have partial gaps due to connection errors. Company earnings and investment plans prioritize official sources; brokerage estimates, anonymous reports, and Telegram interpretations are separately identified. This post is a market analysis for informational purposes and does not constitute a recommendation to buy or sell any specific security.&lt;/p&gt;

 &lt;/blockquote&gt;</description></item><item><title>2028 Matters More Than the 2027 Boom: Integrated Scenarios for Samsung Electronics &amp; SK Hynix</title><link>https://koreainvestinsights.com/post/samsung-sk-hynix-2027-2028-integrated-scenarios-risk-adjusted-valuation-2026-07-13/</link><pubDate>Mon, 13 Jul 2026 20:39:17 +0900</pubDate><guid>https://koreainvestinsights.com/post/samsung-sk-hynix-2027-2028-integrated-scenarios-risk-adjusted-valuation-2026-07-13/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context: This post is a follow-up to the &lt;a class="link" href="https://koreainvestinsights.com/post/hbm-2030-supply-demand-267eb-demand-model-crosscheck-2026-07-13/" &gt;HBM 2030 Supply-Demand Model Cross-Check&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-hynix-2028e-profit-valuation-cycle-scenarios-2026-07-09/" &gt;Samsung Electronics &amp;amp; SK Hynix Valuation Through 2028E Earnings&lt;/a&gt;, and the &lt;a class="link" href="https://koreainvestinsights.com/post/kospi-9pct-selloff-ai-hardware-derating-korea-leverage-2026-07-13/" &gt;July 13 AI Hardware Selloff Analysis&lt;/a&gt;. Accepting 2027&amp;rsquo;s boom as the baseline, it quantifies the probability that supply, efficiency, and financing conditions shift simultaneously beginning in 2028.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;HBM supply/demand is likely to remain tight through 2027. However, the projection of &lt;code&gt;2030 demand 26.7EB vs. supply 10.6EB, 2.52× deficit&lt;/code&gt; is not a base-case forecast — it requires multiple bullish assumptions to hold simultaneously.&lt;/li&gt;
&lt;li&gt;The most important year is not 2027 but &lt;strong&gt;2028&lt;/strong&gt;. Supply expansion from Samsung Electronics, SK Hynix, and Micron, inference efficiency gains, and refinancing verification across AI data centers all converge in the same window.&lt;/li&gt;
&lt;li&gt;HBM long-term supply contracts extend earnings duration but do not eliminate the cycle. They convert price risk into contract renewal risk, customer credit risk, product-mix risk, and repricing risk.&lt;/li&gt;
&lt;li&gt;Conditional probabilities: P1 sustained excess demand 35%, P2 localized credit stress with hyperscaler re-concentration 40%, P3 efficiency gains and supply normalization 15%, P4 systemic credit seizure 10%.&lt;/li&gt;
&lt;li&gt;Probability-weighted EPS for 2027: Samsung Electronics 60,750원, SK Hynix 393,000원. For 2028: Samsung 49,900원, SK Hynix 316,000원.&lt;/li&gt;
&lt;li&gt;Present value of 2028 terminal values discounted at 11% annually: Samsung Electronics 317,246원, SK Hynix 1,956,316원 — implying 24.7% and 6.0% upside respectively to the July 13 KRX closing prices. Once 2028 normalization risk is incorporated, Samsung&amp;rsquo;s margin of safety is wider; SK Hynix is more sensitive to whether P1 holds.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom Line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The 2027 boom is relatively visible. The harder question is &lt;strong&gt;whether high HBM pricing and earnings persist through 2028 and beyond&lt;/strong&gt;. Current prices discount the duration of earnings well beyond the 2027 peak.
 &lt;/div&gt;
&lt;/div&gt;
&lt;p&gt;&lt;img alt="Integrated 2027–2028 Scenario Map for Samsung Electronics and SK Hynix" loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/images/posts/samsung-hynix-2027-2028-scenario-map-2026-07-13.png"&gt;
&lt;/p&gt;
&lt;h2 id="1-analytical-questions-and-evidence-quality"&gt;1. Analytical Questions and Evidence Quality
&lt;/h2&gt;&lt;p&gt;This analysis ties three questions together in a single framework.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;How much confidence can we place in the view that HBM supply/demand will remain tight through 2027?&lt;/li&gt;
&lt;li&gt;If supply, efficiency, and financing conditions change simultaneously in 2028, how do Samsung Electronics&amp;rsquo; and SK Hynix&amp;rsquo;s earnings differ across scenarios?&lt;/li&gt;
&lt;li&gt;Given those changes, at what implied earnings level does the current share price trade?&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The numbers in this report must be read by category.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Category&lt;/th&gt;
 &lt;th&gt;Definition&lt;/th&gt;
 &lt;th&gt;Example in This Report&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fact&lt;/td&gt;
 &lt;td&gt;Verified from public filings or market data&lt;/td&gt;
 &lt;td&gt;July 13 KRX closing prices, Oracle FCF, CoreWeave borrowings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Consensus&lt;/td&gt;
 &lt;td&gt;Aggregated market estimates&lt;/td&gt;
 &lt;td&gt;2027–2028 EPS and net income&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Broker estimate&lt;/td&gt;
 &lt;td&gt;Individual broker forecast&lt;/td&gt;
 &lt;td&gt;Korea Investment &amp;amp; Securities and BNK Investment &amp;amp; Securities SK Hynix EPS&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Model output&lt;/td&gt;
 &lt;td&gt;Derived from disclosed formulas and assumptions&lt;/td&gt;
 &lt;td&gt;P1–P4 probability-weighted EPS and present values&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Analyst judgment&lt;/td&gt;
 &lt;td&gt;Non-statistical analytical judgment&lt;/td&gt;
 &lt;td&gt;Scenario probabilities, fair PER, 11% required return&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Blocked&lt;/td&gt;
 &lt;td&gt;Cannot be determined from public sources alone&lt;/td&gt;
 &lt;td&gt;HBM contract-level pricing, customer-by-customer volumes, actual production yield&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Core formulas are as follows.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Scenario Terminal Value = Scenario EPS × Scenario Fair PER
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Probability-Weighted EPS = Σ(Scenario Probability × Scenario EPS)
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Probability-Weighted Terminal Value = Σ(Scenario Probability × Scenario Terminal Value)
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Present Value = Terminal Value ÷ (1 + 11%)^Remaining Years
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;HBM Need/Fleet = Installed and active HBM demand stock ÷ supply capacity that can be served by installed fleet
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Scenario probabilities are not derived from historical frequency distributions. They are conditional judgments reflecting currently observable demand, supply, efficiency, and financing conditions. Fair PER values are likewise not mechanically observed from the market but are analytical estimates incorporating earnings durability, business concentration, financial structure, and long-term risk.&lt;/p&gt;
&lt;h2 id="2-current-price-and-market-consensus"&gt;2. Current Price and Market Consensus
&lt;/h2&gt;&lt;p&gt;July 13 KRX closing prices and consensus estimates aggregated on the same date are used throughout.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Samsung Electronics&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-13 KRX Closing Price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;254,500원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,845,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027 Consensus EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65,802원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;444,359원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028 Consensus EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65,907원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;433,625원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027 Consensus Net Income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;44.3조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;32.4조원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028 Consensus Net Income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;44.1조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;31.9조원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current Price / 2027 Consensus EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.87×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.15×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current Price / 2028 Consensus EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.86×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.25×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;On the surface, both companies look remarkably cheap. But these low PER multiples do not mean the market is unaware of 2027 earnings — they are more likely a signal that the market doubts the durability of earnings beyond 2028.&lt;/p&gt;
&lt;p&gt;The dispersion among SK Hynix estimates illustrates this clearly.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Source&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2027 EPS&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2028 EPS&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Market Consensus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;444,359원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;433,625원&lt;/td&gt;
 &lt;td&gt;Moderate 2028 normalization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Investment &amp;amp; Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;415,254원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;495,696원&lt;/td&gt;
 &lt;td&gt;Earnings growth continues through 2028&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;BNK Investment &amp;amp; Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;214,642원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;66,989원&lt;/td&gt;
 &lt;td&gt;Rapid peak passage after 2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For the same company, 2028 EPS estimates range from 66,989원 to 495,696원. The divergence stems not from 2026 demand forecasts but from differing assumptions on &lt;strong&gt;2028 HBM average selling prices, the pace of supply expansion, and the duration of AI capex&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;Recalculating current PER using this report&amp;rsquo;s probability-weighted EPS:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Year&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Samsung Electronics&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.19×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.69×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.10×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.84×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Applying the probability-weighted fair PER in reverse to the current share price, the implied EPS the market is pricing in is approximately 31,700원 for Samsung and approximately 245,000원 for SK Hynix — roughly 52% and 45% below the respective 2027 consensus. The market is discounting not 2027 earnings per se, but how long those earnings will last.&lt;/p&gt;
&lt;h2 id="3-what-has-changed-in-the-memory-cycle--and-what-hasnt"&gt;3. What Has Changed in the Memory Cycle — and What Hasn&amp;rsquo;t
&lt;/h2&gt;&lt;h3 id="what-has-changed"&gt;What Has Changed
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;AI servers, HBM, and enterprise SSDs have raised the share of enterprise and data-center demand in the total mix.&lt;/li&gt;
&lt;li&gt;HBM requires customer qualification, advanced packaging, yield ramps, and long-term contracts, making short-run supply response slower than for commodity DRAM.&lt;/li&gt;
&lt;li&gt;As HBM absorbs a growing share of wafer starts, commodity DRAM output from the same fabs declines, which can support commodity DRAM pricing.&lt;/li&gt;
&lt;li&gt;Financing by large cloud providers and AI data center operators front-loads actual memory orders, acting as an accelerant.&lt;/li&gt;
&lt;li&gt;Conversely, when credit problems emerge, demand does not fade gradually — it can step down abruptly through contract renegotiations and investment deferrals.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="what-has-not-changed"&gt;What Has Not Changed
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;High prices ultimately invite new fabs, yield improvement, process-node transitions, and competitor capacity additions.&lt;/li&gt;
&lt;li&gt;High HBM pricing and margins increase the incentive for a second-source supplier to enter and for customers to apply cost-reduction pressure.&lt;/li&gt;
&lt;li&gt;Long-term contracts do not eliminate price risk — they convert it into renewal pricing and customer credit risk.&lt;/li&gt;
&lt;li&gt;Simultaneously applying peak EPS to a high PER is a double-optimism error that counts both earnings and multiples at their most favorable.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The claim that HBM has made memory a better business than legacy DRAM, and the claim that the memory cycle has been abolished, are not the same assertion.&lt;/p&gt;
&lt;h2 id="4-hbm-supply-demand-model-and-the-2028-inflection"&gt;4. HBM Supply-Demand Model and the 2028 Inflection
&lt;/h2&gt;&lt;p&gt;In the &lt;a class="link" href="https://koreainvestinsights.com/post/hbm-2030-supply-demand-267eb-demand-model-crosscheck-2026-07-13/" &gt;HBM 2030 Supply-Demand Model Cross-Check&lt;/a&gt;, we reproduced the original model&amp;rsquo;s &lt;code&gt;demand 26.7EB, supply 10.6EB, 2.52× deficit&lt;/code&gt; in consistent units. The formula itself does not incorrectly compare a demand flow against a supply stock. The issue is that long-horizon assumptions multiply together in series.&lt;/p&gt;
&lt;p&gt;These include:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Monthly token consumption growth&lt;/li&gt;
&lt;li&gt;Model size, context length, and agent state expansion&lt;/li&gt;
&lt;li&gt;KV cache and memory residency ratios&lt;/li&gt;
&lt;li&gt;Throughput, quantization, and routing efficiency&lt;/li&gt;
&lt;li&gt;HBM service life and migration of inference workloads to other memory types&lt;/li&gt;
&lt;li&gt;Achievable production volumes and packaging yield across the three major memory makers&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;For this reason, the 2030 2.52× deficit is best treated as a bullish stress scenario rather than a base case. Re-deriving central assumptions aligned to each of P1–P4 gives the following:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Path&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Token Multiplier&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Model Scale&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Throughput Efficiency&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KV Efficiency&lt;/th&gt;
 &lt;th style="text-align: right"&gt;HBM Residency&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Demand&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Need/Fleet&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P1 Orderly Excess Demand&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.1EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.52×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P2 Localized Credit Stress &amp;amp; Re-concentration&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;60%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12.8EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.21×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P3 Efficiency Gains &amp;amp; Supply Expansion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;14×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.5EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.62×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P4 Systemic Credit Seizure&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;14×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.6EB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.25×&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For comparability, supply capacity is held fixed at 10.6EB. Actual HBM wafer starts, product mix, good-die yield, and packaging capacity are non-public, so absolute values cannot be confirmed.&lt;/p&gt;
&lt;p&gt;Looking along the time axis, the judgment simplifies considerably.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;th&gt;Confidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026–2027&lt;/td&gt;
 &lt;td&gt;HBM4 transition and AI demand outpace new-fab supply additions&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;SK Hynix Yongin Line 1, Micron ID1/Tongluo, and Samsung HBM expansion begin easing supply&lt;/td&gt;
 &lt;td&gt;Medium-High&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2029–2030&lt;/td&gt;
 &lt;td&gt;Remaining deficit is more likely to narrow than to deepen&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2030 2.5× deficit&lt;/td&gt;
 &lt;td&gt;Requires multiple bullish assumptions to hold simultaneously&lt;/td&gt;
 &lt;td&gt;Low&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The investment decision does not therefore hinge on whether the 2027 boom occurs. The pivot is whether the additional supply coming online in 2028 clears customer qualifications, and whether that timing coincides with inference efficiency gains and tighter AI data center financing.&lt;/p&gt;
&lt;h2 id="5-sk-hynix-long-term-contracts-and-the-earnings-curve"&gt;5. SK Hynix Long-Term Contracts and the Earnings Curve
&lt;/h2&gt;&lt;p&gt;Korea Investment &amp;amp; Securities&amp;rsquo; July 13 SK Hynix report attributes the Q2 consensus miss not to demand collapse but to a timing lag in how HBM product-mix shifts and long-term supply contract pricing flow through to reported results.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026F&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2027F&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2028F&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;32.83조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;48.77조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;58.53조원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating Profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;24.51조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;37.45조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;44.70조원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;292,068원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;415,254원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;495,696원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ROE&lt;/td&gt;
 &lt;td style="text-align: right"&gt;94.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;46.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;HBM assumptions are also aggressive.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;HBM Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026F&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2027F&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bit Growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+31.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ASP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.60/Gb&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.82/Gb&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ASP Growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+76.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.71조원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.30조원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue Growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+123.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Key estimate adjustments within the report include:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;HBM4 sales begin in Q3 2026, but the largest earnings inflection is modeled in Q4 when HBM ASP rises &lt;code&gt;+52% QoQ&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Compared to the May estimates, non-HBM DRAM ASP for 2026–2027 was trimmed by approximately 16%.&lt;/li&gt;
&lt;li&gt;2026 HBM revenue was raised by 17%.&lt;/li&gt;
&lt;li&gt;2027 HBM shipment volume was held flat while ASP was cut by 5%.&lt;/li&gt;
&lt;li&gt;NAND shipment volumes and pricing were raised.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;This report is not a retraction of the long-term HBM growth thesis. It is a revision that lowers short-run commodity DRAM price elasticity and pushes back the point at which HBM4 economics enter reported results.&lt;/p&gt;
&lt;h3 id="why-long-term-contracts-should-not-be-taken-at-face-value"&gt;Why Long-Term Contracts Should Not Be Taken at Face Value
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;Contract-level price floors, ceilings, minimum purchase obligations, and prepayment terms are not publicly disclosed.&lt;/li&gt;
&lt;li&gt;Reports that contracts carry &amp;ldquo;no price ceiling&amp;rdquo; cannot by themselves confirm that prices rise every year — one must know whether pricing is linked to spot rates and how often it is reset.&lt;/li&gt;
&lt;li&gt;Long-term contracts lock in volumes and customer relationships but do not eliminate the risk of customer renegotiation, default, or credit deterioration.&lt;/li&gt;
&lt;li&gt;Non-HBM DRAM and NAND outside those contracts remain fully exposed to the commodity price cycle.&lt;/li&gt;
&lt;li&gt;If HBM4 yield is low, SK Hynix&amp;rsquo;s shipments and incremental earnings may still fall short even if the market remains undersupplied.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The largest benefit of long-term contracts is not cycle elimination but &lt;strong&gt;extending earnings duration and converting the form of risk&lt;/strong&gt;.&lt;/p&gt;
&lt;h2 id="6-how-ai-infrastructure-financing-flows-into-memory-orders"&gt;6. How AI Infrastructure Financing Flows Into Memory Orders
&lt;/h2&gt;&lt;p&gt;AI memory demand is not driven purely by technology pull. Data center operators&amp;rsquo; financing activity creates front-loaded orders, and when financial conditions tighten, order gaps can emerge sharply.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Rising AI Expectations
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Cloud/AI data center investment and long-term contract expansion
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Front-loaded capex in GPU, HBM, power, and data centers
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Asset, revenue, and enterprise value appreciation
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Additional debt and equity financing
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Larger capex programs
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Reverse direction
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Monetization delay or higher refinancing costs
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Contract renegotiation / expansion delays
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Memory order gaps
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; ASP and share price decline
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Collateral value and financing conditions deteriorate
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Further capex reduction
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The numbers confirmed from public disclosures are substantial.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Confirmed Value&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle FY2026 Free Cash Flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-$23.7B&lt;/td&gt;
 &lt;td&gt;Large-scale AI capex exceeds cash generation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle Debt Financing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$43B&lt;/td&gt;
 &lt;td&gt;Heavy external funding for AI investment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle Equity Financing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5B&lt;/td&gt;
 &lt;td&gt;Debt alone is insufficient to cover investment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle RPO&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$638B&lt;/td&gt;
 &lt;td&gt;Long-term contracts provide high revenue visibility&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle Customer Prepayments &amp;amp; Direct GPU Provision&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$75B&lt;/td&gt;
 &lt;td&gt;End customers share investment risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CoreWeave Total Debt (End 2025)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$21.6B&lt;/td&gt;
 &lt;td&gt;High financial leverage in AI data centers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CoreWeave New DDTL 4.0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$8.5B&lt;/td&gt;
 &lt;td&gt;Additional financing to continue expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CoreWeave DSCR Threshold&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Minimum 1.15×&lt;/td&gt;
 &lt;td&gt;Full refinancing verification begins no later than after June 30, 2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;CoreWeave&amp;rsquo;s end-2025 detailed borrowings comprised DDTL 2.0 approximately $5.04B, DDTL 2.1 approximately $2.74B, DDTL 1.0 approximately $1.55B, DDTL 3.0 approximately $0.34B, and equipment/software finance leases approximately $4.17B. DDTL 4.0, signed in March 2026, totals $8.5B.&lt;/p&gt;
&lt;p&gt;These figures do not imply that AI demand is fictitious. The contracts, GPUs, power infrastructure, and data centers are real. However, if cash flow fails to keep pace with financing costs in 2027–2028, weaker operators are likely to sell assets to be absorbed by large cloud providers.&lt;/p&gt;
&lt;p&gt;This is why the probability assigned to &lt;strong&gt;P2 — localized stress followed by asset re-concentration among hyperscalers&lt;/strong&gt; — exceeds that of P4 systemic collapse. In P2, memory orders do not disappear but a gap of one or more quarters is plausible.&lt;/p&gt;
&lt;h2 id="7-p1p4-conditional-probability-tree"&gt;7. P1–P4 Conditional Probability Tree
&lt;/h2&gt;&lt;p&gt;Final probabilities are derived by sequencing &lt;code&gt;real demand → financial verification → supply/efficiency → asset absorption&lt;/code&gt;, not from historical memory cycle frequencies.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;2028 AI Monetization and Refinancing Verification
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; Pass 50%
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ P1 Sustained excess demand 70% = 35%
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; └─ P3 Efficiency/supply normalization 30% = 15%
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; Stress 50%
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ├─ P2 Localized stress/re-concentration 80% = 40%
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; └─ P4 Systemic credit seizure 20% = 10%
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The 50/50 first branch is not a statistical prior. It is a neutral starting point reflecting the fact that real demand and contracts are strong but financial verification is not yet complete. Calculating probabilities to single percentage-point precision from current data would produce false precision — adjustments are made only in 5-percentage-point increments.&lt;/p&gt;
&lt;h3 id="evidence-supporting-verification-pass-vs-financial-stress"&gt;Evidence Supporting Verification Pass vs. Financial Stress
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Evidence Supporting Verification Pass&lt;/th&gt;
 &lt;th&gt;Evidence Supporting Financial Stress&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM mass-production ramp timing lag through 2027&lt;/td&gt;
 &lt;td&gt;Oracle FY2026 FCF -$23.7B, debt financing $43B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle RPO $638B, customer prepayments and direct GPU provision $75B&lt;/td&gt;
 &lt;td&gt;CoreWeave total debt $21.6B, new DDTL $8.5B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM long-term contracts and prepayments raise order visibility&lt;/td&gt;
 &lt;td&gt;CoreWeave DSCR 1.15× verification after mid-2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Large cloud providers&amp;rsquo; actual data center investment&lt;/td&gt;
 &lt;td&gt;Rising risk premium on long-dated AI bonds&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The 70/30 split between P1 and P3 on the verification-pass branch reflects the fact that through 2027, qualification, yield, and packaging bottlenecks remain, while memory-side MoE, MLA, KV compression, routing, and offloading can all scale simultaneously from 2028 onward.&lt;/p&gt;
&lt;p&gt;The 80/20 split between P2 and P4 on the stress branch is also clear. GPUs and data centers are not assets that disappear — hyperscalers with strong cash flow can acquire and redeploy them. Because simultaneous capex pullbacks, multiple covenant breaches, and HBM long-term contract impairments have not yet materialized together, P4 is treated as a tail risk at 10%.&lt;/p&gt;
&lt;h3 id="evidence-ledger"&gt;Evidence Ledger
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Evidence&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P1&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P2&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P3&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P4&lt;/th&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM supply lag through 2027&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&amp;ndash;&lt;/td&gt;
 &lt;td&gt;Shortage persists&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Top-3 memory supply expansion post-2028&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td&gt;P3 independent pathway&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT client DDR5/LPDDR expansion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td&gt;Caps commodity price ceiling before HBM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;YMTC high-layer NAND ramp and customer adoption&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td&gt;Pressure on Samsung NAND and enterprise SSD mix first&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China HBM/server memory customer qualification&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&amp;ndash;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td&gt;If confirmed, shifts from P1 to P3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle RPO &amp;amp; customer-provided GPU&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td&gt;Demand is real; buffers contagion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle negative FCF &amp;amp; large financing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+&lt;/td&gt;
 &lt;td&gt;Localized credit stress risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CoreWeave debt &amp;amp; DSCR verification&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+&lt;/td&gt;
 &lt;td&gt;2027–2028 financial inflection&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Rising AI long-dated bond risk premium&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&amp;ndash;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+&lt;/td&gt;
 &lt;td&gt;Shifts 5pp from P1 to P2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Inference efficiency gains &amp;amp; offloading&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td&gt;Still insufficient production-level evidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyperscalers&amp;rsquo; asset absorption capacity&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;++&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&amp;ndash;&lt;/td&gt;
 &lt;td&gt;Why P2 outweighs P4&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;&lt;code&gt;++&lt;/code&gt; and &lt;code&gt;--&lt;/code&gt; denote direction only and are not statistical likelihood ratios.&lt;/p&gt;
&lt;h3 id="probability-revision-history"&gt;Probability Revision History
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Step&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P1&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P2&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P3&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P4&lt;/th&gt;
 &lt;th&gt;Rationale&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Initial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Physical supply shortage dominance, localized financial risk, supply/efficiency path&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI long-duration bond weakness reflected&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;P1 to P2 shift of 5pp; no contagion evidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyperscaler capex review&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Demand execution and asset concentration effects offset&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI long-term scenario review&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;2027–2028 unchanged; only post-2030 P3 probability expands&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China supply path review&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Commodity supply risk reflected; HBM customer qualification evidence lacking&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="probability-uncertainty-ranges-and-revision-rules"&gt;Probability Uncertainty Ranges and Revision Rules
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Path&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Central&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Uncertainty Range&lt;/th&gt;
 &lt;th&gt;Strong Signals That Would Shift the Central Estimate&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25–45%&lt;/td&gt;
 &lt;td&gt;Two or more hyperscalers raise capex guidance; financing costs stabilize; long-term contracts strengthened&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30–50%&lt;/td&gt;
 &lt;td&gt;Single weak-hand operator stress event with hyperscaler asset absorption&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10–25%&lt;/td&gt;
 &lt;td&gt;Top-3 or China supply expansion confirmed in volume; per-token HBM consumption declines&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5–15%&lt;/td&gt;
 &lt;td&gt;Multiple operator defaults, two or more capex cuts, and long-term contract impairments occur simultaneously&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;ul&gt;
&lt;li&gt;A single news article or anecdote warrants only a 0–2pp adjustment.&lt;/li&gt;
&lt;li&gt;A single strong signal from a company filing or earnings report warrants at most a 5pp adjustment.&lt;/li&gt;
&lt;li&gt;Two or more independent strong signals in the same direction warrant a 5–10pp shift.&lt;/li&gt;
&lt;li&gt;Localized credit stress moves probability from P1 to P2; contagion evidence moves it from P2 to P4; supply/efficiency evidence moves it from P1 to P3.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="8-final-scenarios-and-hbm-price-pressure"&gt;8. Final Scenarios and HBM Price Pressure
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Path&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Prob.&lt;/th&gt;
 &lt;th&gt;2028 State&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2030 Need/Fleet&lt;/th&gt;
 &lt;th style="text-align: right"&gt;HBM Price Pressure&lt;/th&gt;
 &lt;th&gt;Key Outcome&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P1 Orderly Excess Demand&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td&gt;Tight or balanced&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.3–1.8×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Flat to -10%&lt;/td&gt;
 &lt;td&gt;Contract price floor holds; HBM earnings plateau&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P2 Localized Credit Stress &amp;amp; Re-concentration&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td&gt;Order gap, modest supply slack&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.9–1.3×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-15 to -30%&lt;/td&gt;
 &lt;td&gt;Weak operators exit; hyperscalers absorb assets&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P3 Efficiency Gains &amp;amp; Supply Expansion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td&gt;Growing supply surplus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.6–0.9×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-20 to -35%&lt;/td&gt;
 &lt;td&gt;Token usage grows but HBM premium compresses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P4 Systemic Credit Seizure&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Capex cuts and inventory correction&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.3–0.6×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-40 to -55%&lt;/td&gt;
 &lt;td&gt;LT contract renegotiation, output cuts, capex cancellations, policy intervention&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Holding all other variables constant, shifting 10 percentage points out of P2 changes the 2028 terminal values as follows:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Probability Shift&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Samsung Terminal Value Change&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Hynix Terminal Value Change&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;10pp from P2 to P1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21,800원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+227,000원&lt;/td&gt;
 &lt;td&gt;Higher probability of sustained undersupply&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;10pp from P2 to P3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7,500원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-70,000원&lt;/td&gt;
 &lt;td&gt;AI economy grows but HBM premium compresses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;10pp from P2 to P4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-16,800원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-132,000원&lt;/td&gt;
 &lt;td&gt;Localized stress contagion to systemic risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;SK Hynix is more sensitive to every probability shift. Even using the same central probabilities, SK Hynix&amp;rsquo;s valuation error bars and share price volatility are larger due to its higher HBM concentration.&lt;/p&gt;
&lt;h2 id="9-china-supply-expansion-is-a-variable-across-all-paths-not-a-separate-scenario"&gt;9. China Supply Expansion Is a Variable Across All Paths, Not a Separate Scenario
&lt;/h2&gt;&lt;p&gt;CXMT and YMTC capacity growth is not a fifth scenario independent of P1–P4. It is a supply variable that alters product-level pricing and market share within every path.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;China Progress Stage&lt;/th&gt;
 &lt;th&gt;Market Affected First&lt;/th&gt;
 &lt;th&gt;Samsung Electronics&lt;/th&gt;
 &lt;th&gt;SK Hynix&lt;/th&gt;
 &lt;th&gt;Scenario Mapping&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT domestic DDR4/DDR5/LPDDR expansion&lt;/td&gt;
 &lt;td&gt;Commodity DRAM&lt;/td&gt;
 &lt;td&gt;ASP and product mix pressure&lt;/td&gt;
 &lt;td&gt;Relatively limited direct impact&lt;/td&gt;
 &lt;td&gt;EPS trimmed across all paths; probabilities unchanged&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT global PC customer qualification&lt;/td&gt;
 &lt;td&gt;Global client DDR5/LPDDR&lt;/td&gt;
 &lt;td&gt;ASP ceiling and share pressure simultaneously&lt;/td&gt;
 &lt;td&gt;Client exposure at risk&lt;/td&gt;
 &lt;td&gt;P3 upside risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China low-end server RDIMM penetration&lt;/td&gt;
 &lt;td&gt;Standard server DRAM&lt;/td&gt;
 &lt;td&gt;Partial server mix pressure&lt;/td&gt;
 &lt;td&gt;DRAM outside AI servers at risk&lt;/td&gt;
 &lt;td&gt;P3 upside review&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;YMTC high-layer NAND ramp and customer adoption&lt;/td&gt;
 &lt;td&gt;Consumer NAND/enterprise SSD&lt;/td&gt;
 &lt;td&gt;Largest NAND share and margin pressure&lt;/td&gt;
 &lt;td&gt;Solidigm and enterprise SSD at risk&lt;/td&gt;
 &lt;td&gt;P3 EPS and multiple reduction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China HBM mass production and AI customer qualification&lt;/td&gt;
 &lt;td&gt;HBM&lt;/td&gt;
 &lt;td&gt;Partial HBM catch-up offset&lt;/td&gt;
 &lt;td&gt;Large scarcity premium and share risk&lt;/td&gt;
 &lt;td&gt;5–10pp shift from P1 to P3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China ramp and global AI capex slowdown concurrent&lt;/td&gt;
 &lt;td&gt;All memory&lt;/td&gt;
 &lt;td&gt;Commodity/NAND/HBM combined pressure&lt;/td&gt;
 &lt;td&gt;HBM/DRAM combined pressure&lt;/td&gt;
 &lt;td&gt;Monitor for P3-to-P4 contagion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;In 2026–2027, CXMT is more likely to cap the client DDR5 and LPDDR price ceiling than to directly undermine HBM pricing. YMTC may similarly intensify price competition in consumer NAND and some enterprise SSDs first.&lt;/p&gt;
&lt;p&gt;For China to emerge as a direct HBM threat, DRAM die production alone is insufficient. TSV bonding, stacking, base-die integration, advanced packaging, and repeated mass-production qualifications by AI customers are all required.&lt;/p&gt;
&lt;p&gt;The following signals must be confirmed before Chinese capacity expansion is incorporated numerically in meaningful quantities:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;CXMT DDR5/LPDDR is adopted at volume by global PC makers at prices more than 20% below those of the top-3 memory suppliers.&lt;/li&gt;
&lt;li&gt;CXMT secures qualification beyond consumer PCs into enterprise PCs and server RDIMM.&lt;/li&gt;
&lt;li&gt;YMTC high-layer NAND is confirmed by actual shipment volumes and enterprise SSD customer wins — not just stated targets.&lt;/li&gt;
&lt;li&gt;China HBM moves beyond sample qualification to repeated mass-production qualification by AI accelerator customers.&lt;/li&gt;
&lt;li&gt;Supply growth and commodity DRAM/NAND price declines among the top-3 suppliers are observed together for two or more consecutive quarters.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="10-samsung-electronics-eps-and-fair-value-by-scenario"&gt;10. Samsung Electronics: EPS and Fair Value by Scenario
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Year&lt;/th&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;EPS&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Fair PER&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Terminal Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;552,500원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;62,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;496,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;58,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;435,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;315,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;68,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;578,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;360,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;38,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;285,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;24,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;192,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A base PER above 9× is not assigned to Samsung because foundry loss reduction has not yet been confirmed. An 8.5× P1 multiple becomes achievable only if HBM share recovery, commodity DRAM/NAND pricing, and foundry normalization all operate together.&lt;/p&gt;
&lt;p&gt;An 8.0× multiple in P4 is applied to avoid the double-discount error of multiplying a trough EPS by an also-depressed PER. In an actual recession scenario, normalized EPS, net cash, and book value must all be considered together.&lt;/p&gt;
&lt;h2 id="11-sk-hynix-eps-and-fair-value-by-scenario"&gt;11. SK Hynix: EPS and Fair Value by Scenario
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Year&lt;/th&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;EPS&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Fair PER&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Terminal Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;430,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,870,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;405,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,835,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;370,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,220,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;P4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;250,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.5×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,375,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;470,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4,230,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;280,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,960,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;210,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,260,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;P4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;80,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.0×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;640,000원&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The 9× P1 multiple presupposes HBM market share, long-term contract terms, high ROE, and a structural rerating of SK Hynix as a strategic AI asset. Maintaining this multiple through 2028 requires confirmed HBM4E share defense, long-term contract repricing, and shareholder return activity.&lt;/p&gt;
&lt;p&gt;Korea Investment &amp;amp; Securities&amp;rsquo; target price of 3,800,000원 applies a 6.0× PBR to 12-month forward BPS — above the historical band peak of 5.0× cited in the same report. A 3,800,000원 target therefore requires not just earnings delivery but a structural rerating of the memory business.&lt;/p&gt;
&lt;p&gt;In P2 and P3, HBM concentration shifts from an advantage to a vulnerability. When ASP falls, earnings and multiples can compress simultaneously.&lt;/p&gt;
&lt;h2 id="12-probability-weighted-results-and-risk-adjusted-comparison"&gt;12. Probability-Weighted Results and Risk-Adjusted Comparison
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Year&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Prob.-Weighted EPS&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Blended Fair PER&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Prob.-Weighted Terminal Value&lt;/th&gt;
 &lt;th style="text-align: right"&gt;PV at 11% Annual Discount&lt;/th&gt;
 &lt;th style="text-align: right"&gt;vs. Current Price&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;60,750원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.04×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;488,525원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;421,385원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;393,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.53×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,959,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,552,333원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+38.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;49,900원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.18×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;408,250원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;317,246원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2028&lt;/td&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;316,000원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.97×&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,517,500원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,956,316원&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Four observations follow from these results:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Viewed through 2027 alone, both companies show risk-adjusted values above current prices.&lt;/li&gt;
&lt;li&gt;Once 2028 normalization probability is discounted, Samsung&amp;rsquo;s margin of safety is wider.&lt;/li&gt;
&lt;li&gt;SK Hynix&amp;rsquo;s 2028 present value is only 6% above the current share price. A large portion of expected return depends on P1 holding.&lt;/li&gt;
&lt;li&gt;Samsung&amp;rsquo;s commodity memory, foundry, and net cash position provide a partial downside buffer in P2 and P3.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;What is already well known to the market is HBM undersupply and higher memory pricing. What may be underpriced in two directions:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;If Samsung&amp;rsquo;s HBM catch-up coincides with commodity DRAM/NAND price recovery, both earnings and multiples could improve together.&lt;/li&gt;
&lt;li&gt;If SK Hynix&amp;rsquo;s long-term contracts and HBM4E defend price floors and share even after 2028 supply additions, the current normalization discount may be excessive.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;Conversely, the market may prove correct if supply, efficiency, and credit conditions all improve simultaneously in 2028, cutting peak earnings short.&lt;/p&gt;
&lt;h2 id="13-how-to-think-about-ai-long-term-demand-beyond-2030"&gt;13. How to Think About AI Long-Term Demand Beyond 2030
&lt;/h2&gt;&lt;p&gt;&lt;a class="link" href="https://ai-2040.com/?choices=plan-a-root" target="_blank" rel="noopener"
 &gt;AI 2040 Plan A&lt;/a&gt; presents a path from approximately 20 million H100e in 2026 to 60 billion H100e by 2034, with AI compute power consumption reaching 5TW by 2034. However, the document itself describes this as a policy scenario, not a best-estimate forecast, and notes that certain variables were adjusted alongside the narrative.&lt;/p&gt;
&lt;p&gt;Long-term scenarios are therefore not fed directly into 2027–2028 EPS estimates.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P1 Excess Demand&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P2 Re-concentration&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P3 Volume Growth / Price Normalization&lt;/th&gt;
 &lt;th style="text-align: right"&gt;P4 Failure&lt;/th&gt;
 &lt;th&gt;Status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026–2028&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Current analysis probabilities&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Post-2030 conditional baseline&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10%&lt;/td&gt;
 &lt;td&gt;Long-term judgment; not a statistical probability&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;If AI and robotics raise the efficiency of chip design, fab construction, and production, total memory bit demand may grow while the scarcity premium on HBM specifically narrows. &lt;strong&gt;HBM market size expansion and long-term PER expansion must therefore be evaluated separately.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;As long-term memory demand grows, the opportunity set expands not only for HBM but also for commodity DRAM, NAND, enterprise SSDs, controllers, CXL, power, cooling, and interconnects. This is a relative advantage for Samsung, with its broader product portfolio.&lt;/p&gt;
&lt;h2 id="14-what-needs-to-materialize-for-this-analysis-to-hold"&gt;14. What Needs to Materialize for This Analysis to Hold
&lt;/h2&gt;&lt;h3 id="samsung-electronics"&gt;Samsung Electronics
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;HBM4 and HBM4E customer qualifications must convert into actual mass-production volumes and revenue share.&lt;/li&gt;
&lt;li&gt;Commodity DRAM and NAND contract price recovery must offset HBM investment costs, bonuses, and provisions.&lt;/li&gt;
&lt;li&gt;Foundry losses must narrow and begin making a meaningful contribution to 2027–2028 earnings.&lt;/li&gt;
&lt;li&gt;Net cash and shareholder return capacity must be maintained despite heavy capex.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="sk-hynix"&gt;SK Hynix
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;HBM4E share and long-term contract price floors must hold even as Samsung and Micron expand supply.&lt;/li&gt;
&lt;li&gt;HBM repricing beginning in Q4 2026 must track toward the 2027 average of $2.82/Gb.&lt;/li&gt;
&lt;li&gt;TSV and stacking yield weakness must prove temporary and attributable to product transition costs.&lt;/li&gt;
&lt;li&gt;High free cash flow must flow toward shareholder returns and capital policy, not just capacity expansion.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="15-risk-signals-and-leading-indicators"&gt;15. Risk Signals and Leading Indicators
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;th&gt;Leading Indicator&lt;/th&gt;
 &lt;th&gt;Samsung Impact&lt;/th&gt;
 &lt;th&gt;SK Hynix Impact&lt;/th&gt;
 &lt;th&gt;Analysis Adjustment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM long-term contract renegotiation&lt;/td&gt;
 &lt;td&gt;Price floor, prepayment, and minimum purchase weakening&lt;/td&gt;
 &lt;td&gt;Moderate&lt;/td&gt;
 &lt;td&gt;Very large&lt;/td&gt;
 &lt;td&gt;P2/P4 up&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyperscaler capex guidance cut&lt;/td&gt;
 &lt;td&gt;Two or more simultaneous guidance cuts&lt;/td&gt;
 &lt;td&gt;Large&lt;/td&gt;
 &lt;td&gt;Very large&lt;/td&gt;
 &lt;td&gt;P2/P4 up&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI data center refinancing event&lt;/td&gt;
 &lt;td&gt;CoreWeave covenant terms/financing cost, Oracle funding&lt;/td&gt;
 &lt;td&gt;Moderate&lt;/td&gt;
 &lt;td&gt;Large&lt;/td&gt;
 &lt;td&gt;P2 up; P4 up if contagion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Accelerated supply expansion&lt;/td&gt;
 &lt;td&gt;Samsung/Micron volume-capable HBM output rises&lt;/td&gt;
 &lt;td&gt;Positive for HBM catch-up; negative for pricing&lt;/td&gt;
 &lt;td&gt;Very negative&lt;/td&gt;
 &lt;td&gt;P3 up&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sharp inference efficiency improvement&lt;/td&gt;
 &lt;td&gt;Per-token HBM bits decline&lt;/td&gt;
 &lt;td&gt;Broad product portfolio provides partial buffer&lt;/td&gt;
 &lt;td&gt;Negative&lt;/td&gt;
 &lt;td&gt;HBM multiple lower&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China commodity supply expansion&lt;/td&gt;
 &lt;td&gt;CXMT/YMTC pricing, share, global customer qualifications&lt;/td&gt;
 &lt;td&gt;More negative for commodity DRAM/NAND&lt;/td&gt;
 &lt;td&gt;Client DRAM and Solidigm under pressure&lt;/td&gt;
 &lt;td&gt;Product-level EPS reduction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China HBM challenge&lt;/td&gt;
 &lt;td&gt;TSV/stacking/base die/packaging/customer qualification&lt;/td&gt;
 &lt;td&gt;Partial offset of HBM catch-up value&lt;/td&gt;
 &lt;td&gt;Very negative for scarcity premium&lt;/td&gt;
 &lt;td&gt;P1 to P3 shift&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry normalization failure&lt;/td&gt;
 &lt;td&gt;Continued losses, no yield improvement&lt;/td&gt;
 &lt;td&gt;Samsung-specific risk&lt;/td&gt;
 &lt;td&gt;Limited impact&lt;/td&gt;
 &lt;td&gt;Samsung P1 multiple reduction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="upside-signals"&gt;Upside Signals
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;2028 HBM long-term contracts with higher volumes and prices and broader customer diversification&lt;/li&gt;
&lt;li&gt;Early stabilization of HBM4E yield and mass-production volumes&lt;/li&gt;
&lt;li&gt;Simultaneous confirmation of Samsung&amp;rsquo;s rising HBM share and narrowing foundry losses&lt;/li&gt;
&lt;li&gt;AI service revenue and cash flow growing faster than compute resource contracts&lt;/li&gt;
&lt;li&gt;HBM lead times and allocations hold even as supply expansion lags&lt;/li&gt;
&lt;li&gt;SK Hynix shareholder return plan becomes concrete&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="downside-signals"&gt;Downside Signals
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;HBM ASP decline and shipment volume drop occur simultaneously&lt;/li&gt;
&lt;li&gt;Two or more major hyperscalers lower capex growth rates&lt;/li&gt;
&lt;li&gt;Long-term contract renegotiation, breach, or prepayment impairment&lt;/li&gt;
&lt;li&gt;Memory inventory build and spot-to-contract price inversion&lt;/li&gt;
&lt;li&gt;Actual AI service per-token HBM consumption falls 20%+ for two consecutive quarters&lt;/li&gt;
&lt;li&gt;Samsung/Micron volume supply increase coincides with HBM premium decline&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;On a monthly basis, the indicators to track together are: HBM4/4E contract prices and prepayments, top-3 memory producer volumes and yields, advanced packaging lead times, hyperscaler capex and cash flow, Oracle and CoreWeave financing costs, Taiwan server ODM revenue, and DRAM/NAND contract prices and inventory.&lt;/p&gt;
&lt;h2 id="16-relative-investment-interpretation"&gt;16. Relative Investment Interpretation
&lt;/h2&gt;&lt;p&gt;The two names are not equivalent ways to buy the same memory upcycle.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;Samsung Electronics&lt;/th&gt;
 &lt;th&gt;SK Hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P1 Strengthening&lt;/td&gt;
 &lt;td&gt;HBM catch-up and commodity memory tailwind&lt;/td&gt;
 &lt;td&gt;Largest earnings leverage from HBM share and LT contracts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P2 Warning&lt;/td&gt;
 &lt;td&gt;Business diversification and net cash provide partial defense&lt;/td&gt;
 &lt;td&gt;More sensitive to order gaps and multiple compression&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P3 Strengthening&lt;/td&gt;
 &lt;td&gt;Broad memory product range and foundry execution provide partial buffer&lt;/td&gt;
 &lt;td&gt;Direct exposure to HBM premium compression&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;P4 Warning&lt;/td&gt;
 &lt;td&gt;Absolute decline unavoidable but relative defense possible&lt;/td&gt;
 &lt;td&gt;Greatest risk of simultaneous earnings and multiple decline&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;From a risk-adjusted standpoint, Samsung offers a wider and more defensive choice. SK Hynix has greater earnings leverage in P1, but its valuation shifts faster when supply, efficiency, and credit conditions change.&lt;/p&gt;
&lt;p&gt;Equipment and materials names also should not be grouped together indiscriminately. Equipment companies exposed to order gaps must be separated from consumable and parts suppliers whose revenues track utilization rates. Equipment with confirmed purchase orders and components/materials with recurring revenue streams are a better fit for this scenario framework than a pure HBM theme basket.&lt;/p&gt;
&lt;h2 id="17-conclusion"&gt;17. Conclusion
&lt;/h2&gt;&lt;p&gt;The memory boom through 2027 can be maintained as the central path for earnings estimates. But the real test of current share prices is not 2027 EPS — it is the duration of earnings beyond 2028. The claim that the HBM supply deficit has lasted longer than past cycles, and the claim that memory companies deserve a permanently elevated scarcity multiple, are not the same argument.&lt;/p&gt;
&lt;p&gt;Samsung Electronics holds HBM catch-up potential, commodity DRAM/NAND, foundry operations, and net cash simultaneously, offering P1 upside and relative downside defense across P2–P4. SK Hynix&amp;rsquo;s HBM share and long-term contracts provide the highest earnings leverage in P1, but if supply, efficiency, and financing conditions normalize, earnings and multiples can compress together.&lt;/p&gt;
&lt;p&gt;The current probability-weighted results say: &lt;strong&gt;through 2027, both companies offer meaningful upside; looking through to 2028, Samsung&amp;rsquo;s risk-adjusted margin of safety is wider.&lt;/strong&gt; For SK Hynix to see further rerating, confirmation is needed that HBM4E yield, 2028 long-term contract terms, and AI infrastructure financing continue to support P1.&lt;/p&gt;
&lt;h2 id="items-that-cannot-yet-be-confirmed-from-public-sources"&gt;Items That Cannot Yet Be Confirmed from Public Sources
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;Samsung Electronics&amp;rsquo; segment-level EPS for 2027–2028 and foundry normalization sensitivity&lt;/li&gt;
&lt;li&gt;SK Hynix long-term contract customer weighting, repricing formula, and prepayment/minimum purchase terms&lt;/li&gt;
&lt;li&gt;Top-3 memory makers&amp;rsquo; HBM-dedicated wafer input, good-die and stacking yield, and customer-specific producible volumes&lt;/li&gt;
&lt;li&gt;Composition of 2028 hyperscaler capex between internal cash flow, corporate bonds, and project finance&lt;/li&gt;
&lt;li&gt;CoreWeave and Oracle refinancing schedules and covenant headroom&lt;/li&gt;
&lt;li&gt;Per-generation HBM4E capacity, bandwidth, and memory residency rate changes per accelerator&lt;/li&gt;
&lt;li&gt;CXMT/YMTC actual shipment volumes, global customer qualifications, and product-level earnings sensitivity&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="key-sources"&gt;Key Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://vo.la/fd9udR9" target="_blank" rel="noopener"
 &gt;Korea Investment &amp;amp; Securities, SK Hynix 2Q26 Preview, 2026-07-13&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://vo.la/jBPy7zV" target="_blank" rel="noopener"
 &gt;Korea Investment &amp;amp; Securities, &amp;ldquo;The Start Begins Now,&amp;rdquo; 2026-05-20&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://koreainvestinsights.com/post/hbm-2030-supply-demand-267eb-demand-model-crosscheck-2026-07-13/" &gt;Korea Invest Insights, HBM 2030 Supply-Demand Model Cross-Check&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://ai-2040.com/?choices=plan-a-root" target="_blank" rel="noopener"
 &gt;AI 2040, Plan A&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://ai-2040.com/supplements/compute-supplement" target="_blank" rel="noopener"
 &gt;AI 2040, Compute Supplement&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://ai-2040.com/supplements/economics-of-plan-a" target="_blank" rel="noopener"
 &gt;AI 2040, Economics of Plan A&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.oracle.com/news/announcement/q4fy26-earnings-release-2026-06-10/" target="_blank" rel="noopener"
 &gt;Oracle FY2026 Earnings&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1769628/000176962826000104/crwv-20251231.htm" target="_blank" rel="noopener"
 &gt;CoreWeave 2025 Form 10-K&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1769628/000176962826000129/crwv-20260330.htm" target="_blank" rel="noopener"
 &gt;CoreWeave DDTL 4.0 Form 8-K&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The scenario probabilities and fair values in this report are analytical estimates based on publicly available information and do not constitute personalized investment advice. Actual investment decisions must separately account for price volatility, investment horizon, taxation, currency risk, and individual loss tolerance.&lt;/p&gt;</description></item><item><title>Inside KOSPI's 9.05% Selloff: AI Hardware Derating and Korea's Leverage Amplifier</title><link>https://koreainvestinsights.com/post/kospi-9pct-selloff-ai-hardware-derating-korea-leverage-2026-07-13/</link><pubDate>Mon, 13 Jul 2026 18:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/kospi-9pct-selloff-ai-hardware-derating-korea-leverage-2026-07-13/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context: this article follows our work on &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;H1 2026 AI infrastructure bottlenecks and narrow leadership&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-earnings-selloff-nvidia-q4fy26-rebound-trigger-2026-07-08/" &gt;Samsung Electronics&amp;rsquo; post-earnings selloff and the NVIDIA precedent&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-adr-000660-price-path-leverage-etf-plumbing-2026-07-04/" &gt;SK hynix ADR and leveraged-ETF plumbing&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/" &gt;late-July big-tech earnings scenarios for the memory thesis&lt;/a&gt;. The purpose is to separate long-term AI demand from short-term price mechanics.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;On July 13, 2026, KOSPI closed at &lt;strong&gt;6,799.24, down 9.05%&lt;/strong&gt;. KOSDAQ fell 4.43% to 800.35. Samsung Electronics lost 9.21% to KRW 258,750 and SK hynix fell 13.35% to KRW 1,889,000.&lt;/li&gt;
&lt;li&gt;A Korea-only market-plumbing accident does not explain the full move. Japanese NAND, MLCC and FC-BGA names also sold off sharply. The best causal frame is that &lt;strong&gt;a global unwind in crowded AI and electronics hardware set the direction, while Korea&amp;rsquo;s leveraged ETFs, margin exposure and program trading magnified the decline&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;The evidence does not establish a collapse in physical AI demand. TSMC reported record June revenue, while early-July memory exports and long-term HBM supply evidence still point to strong demand. Strong demand, however, does not automatically protect an elevated earnings path or valuation multiple.&lt;/li&gt;
&lt;li&gt;The neutral reading is that deleveraging has started but completion is unconfirmed. A durable bottom requires fewer new lows, weaker program selling, a decline in leveraged-ETF and margin exposure, and improving relative strength in U.S. semiconductors over the next one to three trading days.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The selloff looks more like a simultaneous repricing of &lt;strong&gt;AI hardware profit duration and crowded positioning&lt;/strong&gt; than proof that AI demand has disappeared. Long-term business quality and short-term equity timing are different questions.
 &lt;/div&gt;
&lt;/div&gt;
&lt;p&gt;&lt;img alt="Korean market selloff and causal structure on July 13, 2026" loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/images/posts/kospi-9pct-selloff-ai-hardware-derating-2026-07-13.png"&gt;
&lt;/p&gt;
&lt;h2 id="1-the-closing-tape"&gt;1. The closing tape
&lt;/h2&gt;&lt;p&gt;Intraday screens showed KOSPI near 6,803, Samsung Electronics at KRW 254,500 and SK hynix at KRW 1,847,000. A market report should start with closing data rather than intraday lows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Asset&lt;/th&gt;
 &lt;th style="text-align: right"&gt;July 13 close&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Daily return&lt;/th&gt;
 &lt;th&gt;Data type&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6,799.24&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.05%&lt;/td&gt;
 &lt;td&gt;Naver Finance 15:30 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;800.35&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.43%&lt;/td&gt;
 &lt;td&gt;Naver Finance 15:30 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 258,750&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.21%&lt;/td&gt;
 &lt;td&gt;Kiwoom REST daily bar&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,889,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.35%&lt;/td&gt;
 &lt;td&gt;Kiwoom REST daily bar&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The index closes can be checked on the &lt;a class="link" href="https://finance.naver.com/sise/sise_index.naver?code=KOSPI" target="_blank" rel="noopener"
 &gt;KOSPI&lt;/a&gt; and &lt;a class="link" href="https://finance.naver.com/sise/sise_index.naver?code=KOSDAQ" target="_blank" rel="noopener"
 &gt;KOSDAQ&lt;/a&gt; pages.&lt;/p&gt;
&lt;p&gt;Program trading played a direct role. The Kiwoom market proxy recorded approximately KRW 2.14 trillion of net program selling. Arbitrage flows were about KRW 172 billion net positive, but non-arbitrage flows were approximately KRW 2.31 trillion net negative. The dominant force was therefore broad basket reduction rather than a simple basis trade.&lt;/p&gt;
&lt;p&gt;These are Kiwoom REST market proxies, not final KRX official statistics. They are useful for direction and scale but should be superseded when official exchange data are available.&lt;/p&gt;
&lt;h2 id="2-a-multi-stage-causal-chain"&gt;2. A multi-stage causal chain
&lt;/h2&gt;&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Middle East, oil, rates and AI-capex funding concerns
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; global reduction of crowded AI and electronics-hardware exposure
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; lower SK hynix 2026-2027 earnings estimates become Korea&amp;#39;s direct catalyst
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; selling spreads into Japanese NAND, MLCC and FC-BGA baskets
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; Korea&amp;#39;s index concentration, margin exposure, single-stock leveraged ETFs and program selling magnify losses
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; positive physical-demand evidence fails to lift prices, triggering further liquidation
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Factor&lt;/th&gt;
 &lt;th&gt;Role&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Confidence&lt;/th&gt;
 &lt;th&gt;Neutral interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Global AI and electronics hardware unwind&lt;/td&gt;
 &lt;td&gt;Base direction&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;Japanese NAND, MLCC and FC-BGA names fell together, indicating a factor-level move beyond Korea.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Semiconductor leverage liquidation&lt;/td&gt;
 &lt;td&gt;Korean amplifier&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;Large-cap program selling and leveraged-ETF rebalancing fit the observed nonlinear decline.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix earnings revision&lt;/td&gt;
 &lt;td&gt;Direct catalyst&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;Long-term demand remained constructive, but near-term earnings and the 2026-2027 trajectory were reduced.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oil, rates and geopolitical risk&lt;/td&gt;
 &lt;td&gt;Macro trigger&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;These explain higher discount rates and risk aversion but not Korea&amp;rsquo;s full relative loss.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI capex funding cost&lt;/td&gt;
 &lt;td&gt;Multiple pressure&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium&lt;/td&gt;
 &lt;td&gt;Capital costs may separate cash-rich investors from more leveraged infrastructure spenders.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Physical AI-demand collapse&lt;/td&gt;
 &lt;td&gt;Structural bear case&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Low&lt;/td&gt;
 &lt;td&gt;This conflicts with TSMC sales, memory exports, long-term contracts and upstream capacity planning.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The distinction between cause and amplifier matters. Korea&amp;rsquo;s leverage structure alone cannot easily explain the synchronized fall in Japanese hardware names. Conversely, the regional unwind alone does not explain why Korea and its two largest memory stocks suffered such extreme losses.&lt;/p&gt;
&lt;h2 id="3-a-regional-shock-with-a-korean-multiplier"&gt;3. A regional shock with a Korean multiplier
&lt;/h2&gt;&lt;p&gt;Around early afternoon, KOSPI was down roughly 6.8% to 7.2%, while Taiwan was slightly positive, Japan was down around 1.8%, China about 1.5%, and Hong Kong only modestly lower. These were intraday observations, not final closes.&lt;/p&gt;
&lt;p&gt;Later, Japanese hardware losses deepened. Intraday screens showed Taiyo Yuden down 18.87%, Yaskawa Electric 14.34%, Kioxia 10.77%, Murata 8.06% and Ibiden 7.39%. The figures are directional observations only.&lt;/p&gt;
&lt;p&gt;The common move does not mean every company faced the same fundamental shock. Yaskawa had company-specific execution issues. Kioxia also had concentrated domestic margin exposure. Murata and Taiyo Yuden have significant smartphone and auto exposure, making their declines closer to high-beta electronics selling. Ibiden&amp;rsquo;s FC-BGA exposure gives it stronger evidentiary value for an AI-hardware derating.&lt;/p&gt;
&lt;p&gt;The most coherent decomposition is therefore &lt;strong&gt;a global hardware crowding unwind plus a Korean leverage amplifier&lt;/strong&gt;.&lt;/p&gt;
&lt;h2 id="4-why-korea-moved-more-than-taiwan"&gt;4. Why Korea moved more than Taiwan
&lt;/h2&gt;&lt;p&gt;Both Korea and Taiwan are semiconductor-heavy, but corporate concentration is not the full story.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Market&lt;/th&gt;
 &lt;th&gt;Main H1 driver&lt;/th&gt;
 &lt;th&gt;Nature of concentration&lt;/th&gt;
 &lt;th&gt;Vulnerability in a correction&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea&lt;/td&gt;
 &lt;td&gt;Memory earnings upgrades, domestic retail and financial-investment flows, margin and leveraged exposure&lt;/td&gt;
 &lt;td&gt;Corporate, index, derivatives and credit concentration occurred together&lt;/td&gt;
 &lt;td&gt;Forced selling can become self-reinforcing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Taiwan&lt;/td&gt;
 &lt;td&gt;TSMC earnings credibility and domestic absorption of foreign selling&lt;/td&gt;
 &lt;td&gt;High corporate concentration but weaker single-stock leverage plumbing&lt;/td&gt;
 &lt;td&gt;Vulnerable if TSMC earnings confidence breaks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Japan&lt;/td&gt;
 &lt;td&gt;Foreign cash-equity demand, governance reform and multi-sector participation&lt;/td&gt;
 &lt;td&gt;Leadership spread across semiconductors, banks, trading houses, automation and defense&lt;/td&gt;
 &lt;td&gt;Yen and global-cycle sensitivity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Cited industry data indicated that KOSPI margin financing increased from roughly KRW 22.56 trillion to KRW 29.23 trillion during the second quarter, while KOSDAQ margin financing declined from about KRW 10.36 trillion to KRW 8.09 trillion. If confirmed on a consistent basis, this helps explain why KOSPI moved more violently than KOSDAQ. July 13 end-of-day margin balances and forced-liquidation amounts were not yet available.&lt;/p&gt;
&lt;h2 id="5-evidence-against-a-simple-ai-demand-collapse-thesis"&gt;5. Evidence against a simple AI-demand-collapse thesis
&lt;/h2&gt;&lt;h3 id="51-tsmc-june-revenue"&gt;5.1 TSMC June revenue
&lt;/h3&gt;&lt;p&gt;&lt;a class="link" href="https://investor.tsmc.com/english/monthly-revenue/2026" target="_blank" rel="noopener"
 &gt;TSMC&amp;rsquo;s official monthly revenue data&lt;/a&gt; showed June sales of NT$442.68 billion, up 6.2% month over month and 67.9% year over year. Second-quarter revenue was approximately NT$1.270 trillion, up about 12.0% sequentially and 36.1% year over year.&lt;/p&gt;
&lt;p&gt;The June comparison benefited from a weak June 2025 base. First-quarter growth of about 35.1% and second-quarter growth of about 36.1% do not show a doubling of the underlying rate. Yet the absolute monthly record and positive monthly and quarterly growth cannot be dismissed as a base effect. This is strong evidence that GPU, ASIC and leading-edge logic demand is converting into revenue. It is only indirect evidence for SK hynix HBM pricing, LTA margins and yields.&lt;/p&gt;
&lt;h3 id="52-early-july-memory-exports"&gt;5.2 Early-July memory exports
&lt;/h3&gt;&lt;p&gt;A brokerage interpretation of July 1-10 preliminary trade data showed DRAM export value up roughly 29.4% from the comparable June period, dollar-per-kilogram pricing up 15.7%, and implied volume up about 11.8%. The direction suggests both price and volume improved.&lt;/p&gt;
&lt;p&gt;Ten-day preliminary data do not guarantee the full month. Dollars per kilogram are not bit ASP, and the figures cannot be mapped directly to company revenue or operating profit.&lt;/p&gt;
&lt;h3 id="53-sk-hynix-ltas-and-estimate-revisions"&gt;5.3 SK hynix LTAs and estimate revisions
&lt;/h3&gt;&lt;p&gt;A &lt;a class="link" href="https://biz.chosun.com/stock/stock_general/2026/07/13/O2R4UKPBY5G2RJXLUIVOIGNWVU/" target="_blank" rel="noopener"
 &gt;Korea Investment &amp;amp; Securities estimate cited in the press&lt;/a&gt; placed second-quarter 2026 operating profit at KRW 60.4 trillion, roughly 8% below a KRW 65 trillion consensus, and cut 2026 and 2027 operating-profit estimates by about 9% and 11%. The same analysis remained constructive on DRAM and NAND pricing and the durability benefits of long-term agreements.&lt;/p&gt;
&lt;p&gt;The balanced interpretation is that near-term earnings altitude and revision momentum weakened, while LTAs may support volume visibility and profit floors. Without public price floors, cost pass-through, take-or-pay and renegotiation terms, it is not possible to conclude that LTAs eliminate the memory cycle.&lt;/p&gt;
&lt;h3 id="54-cpu-wafers-and-long-term-fab-plans"&gt;5.4 CPU, wafers and long-term fab plans
&lt;/h3&gt;&lt;p&gt;Server CPU growth reinforces the idea that AI servers require more than GPUs, broadening bottlenecks into substrates, sockets, BMCs and conventional DRAM. Forecasts such as a US$100 billion server-CPU market by 2028 and a 53% three-year CAGR rely on aggressive assumptions for agentic-AI demand and CPU pricing and should remain scenarios until orders and supplier results confirm them.&lt;/p&gt;
&lt;p&gt;Micron and GlobalWafers&amp;rsquo; long-term U.S. 300mm wafer agreement, Samsung&amp;rsquo;s reported Yongin schedule acceleration, and SK&amp;rsquo;s M15X and Yongin plans show that memory demand is entering upstream procurement and long-term capacity plans. Those fabs ramp mainly from 2027 through 2030, supporting near-term tightness while increasing the risk of margin normalization after 2029.&lt;/p&gt;
&lt;h3 id="55-hbm-through-2030"&gt;5.5 HBM through 2030
&lt;/h3&gt;&lt;p&gt;Some industry work projects HBM demand rising from roughly 4.8EB today to 26.7EB in 2030 versus supply of 10.6EB. The supply forecast appears to incorporate about 2.65 times current capacity, or roughly 21.5% annual growth, rather than assuming no expansion.&lt;/p&gt;
&lt;p&gt;The directional case is credible because stacking, leading-edge-node migration, yield, TSV, packaging and inspection constraints prevent wafer input from translating one-for-one into effective bit supply. The exact shortage is highly sensitive to token usage, KV-cache efficiency, mixture-of-experts, quantization and offloading assumptions.&lt;/p&gt;
&lt;p&gt;A practical base path is very tight supply through 2027, a growing capacity contribution from 2028, and some broad shortage relief in 2029-2030 while HBM remains structurally tighter than commodity memory.&lt;/p&gt;
&lt;h2 id="6-why-strong-demand-does-not-guarantee-a-strong-stock"&gt;6. Why strong demand does not guarantee a strong stock
&lt;/h2&gt;&lt;p&gt;The market is no longer asking only whether AI demand exists. It is asking how long excess profits can last.&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Funding costs:&lt;/strong&gt; larger AI infrastructure commitments make debt costs and cash-generation differences more important. This implies dispersion among spenders, not necessarily demand disappearance.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;AI inflation:&lt;/strong&gt; higher memory, power and infrastructure prices help suppliers but can lift inflation, bond yields and customer return hurdles. Earnings can improve while equity multiples contract.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Profit-duration debate:&lt;/strong&gt; the key question is whether current margins survive competition, capacity additions and technology shifts.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Emerging-market index concentration:&lt;/strong&gt; heavy TSMC, Samsung Electronics and SK hynix weights create benchmark-flow risk independent of corporate earnings. Global investors can stay long AI while diversifying into India, China, energy or utilities.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="7-separating-ltas-from-a-6x-pb-thesis"&gt;7. Separating LTAs from a 6x P/B thesis
&lt;/h2&gt;&lt;p&gt;If LTAs protect volume and price downside, reduce earnings volatility and extend high ROE, higher valuation multiples are theoretically justified. The market, however, observes estimate cuts immediately while treating cycle elimination as an unproven assumption.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Observable&lt;/th&gt;
 &lt;th&gt;Still unverified&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Second-quarter and 2026-2027 estimate changes&lt;/td&gt;
 &lt;td&gt;LTA floors, cost pass-through, penalties and renegotiation clauses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Recent forecast errors and ASP expectations&lt;/td&gt;
 &lt;td&gt;HBM4 yields, packaging cost and realized premium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;A target-price formula using forward book value&lt;/td&gt;
 &lt;td&gt;Long-duration exceptional ROE and the disappearance of downcycles&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Applying a 12-month forward book value of KRW 643,124 mechanically gives about KRW 1.93 million at 3x P/B and KRW 3.86 million at 6x. The difference reflects profit duration and risk more than a few trillion won of quarterly earnings.&lt;/p&gt;
&lt;p&gt;At zero growth and a 10% cost of equity, a simplified residual-income identity links 3x and 6x P/B to sustainable ROEs of roughly 30% and 60%. With 4% growth, the implied ROEs fall to about 22% and 40%. This is an explanatory sensitivity, not a precise valuation. A 6x multiple requires large excess profits, long duration and structurally lower risk at the same time.&lt;/p&gt;
&lt;h2 id="8-single-stock-leveraged-etfs-and-short-gamma-plumbing"&gt;8. Single-stock leveraged ETFs and short-gamma plumbing
&lt;/h2&gt;&lt;p&gt;Market estimates suggested that single-stock leveraged-ETF assets fell from about KRW 17.4 trillion on June 25 to KRW 11.6 trillion on July 8. A modeled additional 5% shock implied sensitivities of roughly KRW 2.65 trillion for SK hynix and KRW 1.49 trillion for Samsung Electronics. These were modeled exposures, not actual sell orders.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Share-price decline
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; lower leveraged-ETF NAV
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; underlying sales to restore target leverage
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; further price decline
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; more program selling and margin liquidation
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; -&amp;gt; another round of underlying sales
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The July 13 shock may mark the beginning of normalization because it mechanically reduced leverage. It does not prove completion because end-of-day ETF assets, margin balances and forced-liquidation data were unavailable. Japan&amp;rsquo;s hardware selloff also shows that leverage was an amplifier rather than the sole root cause.&lt;/p&gt;
&lt;p&gt;&lt;a class="link" href="https://n.news.naver.com/mnews/article/008/0005385145?sid=101" target="_blank" rel="noopener"
 &gt;Reported regulatory discussions&lt;/a&gt; indicate recognition of the structural risk. Restricting new products could reduce medium-term fragility, but forced conversion, leverage cuts or liquidation of existing products could create a second round of short-term selling. New-product policy and legacy-product treatment must be separated.&lt;/p&gt;
&lt;h2 id="9-what-strengthened-weakened-and-remains-unknown"&gt;9. What strengthened, weakened and remains unknown
&lt;/h2&gt;&lt;h3 id="strengthened"&gt;Strengthened
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;AI infrastructure demand is broadening from GPUs and HBM into CPUs, conventional DRAM, wafers, substrates and power networks.&lt;/li&gt;
&lt;li&gt;Limited 2027 commodity-memory capacity and possible LTA prioritization support memory pricing and profit duration.&lt;/li&gt;
&lt;li&gt;Early-July DRAM data suggest simultaneous improvement in price and shipments.&lt;/li&gt;
&lt;li&gt;Long-term fab schedules and upstream contracts indicate demand is being incorporated into plans beyond 2029.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="weakened"&gt;Weakened
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;The assumption that good earnings and capex news immediately lift share prices.&lt;/li&gt;
&lt;li&gt;The claim that a low forward P/E automatically defines a bottom.&lt;/li&gt;
&lt;li&gt;The view that an SK hynix ADR premium creates a risk-free rerating of the Korean ordinary share.&lt;/li&gt;
&lt;li&gt;Confidence in any specific short-term price floor.&lt;/li&gt;
&lt;li&gt;The claim that LTAs alone eliminate the memory cycle and immediately justify 6x P/B.&lt;/li&gt;
&lt;li&gt;The idea that the selloff was an isolated Korean leveraged-ETF accident.&lt;/li&gt;
&lt;/ol&gt;
&lt;h3 id="unknown"&gt;Unknown
&lt;/h3&gt;&lt;ol&gt;
&lt;li&gt;Whether July 13 was the final leverage-liquidation low.&lt;/li&gt;
&lt;li&gt;How much LTAs change actual margin floors and pricing upside.&lt;/li&gt;
&lt;li&gt;Whether late-July big-tech AI spending is maintained or raised.&lt;/li&gt;
&lt;li&gt;How strongly post-2029 capacity normalizes excess margins.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="10-scenario-map"&gt;10. Scenario map
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;Confirmation conditions&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bull: flow-driven capitulation&lt;/td&gt;
 &lt;td&gt;New lows stop, program selling collapses, ETF assets and margin exposure fall, foreign cash buying returns&lt;/td&gt;
 &lt;td&gt;Deleveraging gives way to a higher-quality rebound&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base: volatile bottom-building&lt;/td&gt;
 &lt;td&gt;Program selling slows but cash buying remains weak, new lows alternate with rebounds, markets await CPI, ASML and TSMC&lt;/td&gt;
 &lt;td&gt;Liquidation continues and the low is unconfirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bear: structural repricing&lt;/td&gt;
 &lt;td&gt;New lows despite constructive CPI, ASML and TSMC; big-tech capex slows; estimates fall in sequence; SK hynix lacks relative strength&lt;/td&gt;
 &lt;td&gt;Both 2027 earnings and justified P/B move lower&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The most important test is whether prices begin to rise on good news. A change in reaction function is more direct bottom evidence than the optimism of the headline itself.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;U.S. semiconductor response&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SOX outperforms Nasdaq 100 and MU, NVDA, AVGO, AMAT and LRCX reclaim lows&lt;/td&gt;
 &lt;td&gt;Asia over-discounted the risk and Korean leverage was a major amplifier&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SOX falls 3% to 5% with memory, equipment and substrates weak together&lt;/td&gt;
 &lt;td&gt;A global AI-hardware multiple correction is confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Leveraged AI-infrastructure spenders fall while cash-rich big tech and semis defend&lt;/td&gt;
 &lt;td&gt;Funding-capacity differentiation, not broad AI-demand collapse&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="11-what-to-monitor-next"&gt;11. What to monitor next
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;The next one to three trading days: new lows, volatility, volume and program-selling intensity.&lt;/li&gt;
&lt;li&gt;Confirmed single-stock ETF assets, margin balances, forced liquidation and foreign cash flows.&lt;/li&gt;
&lt;li&gt;Hormuz traffic, insurance costs, WTI, the U.S. 10-year yield and USD/KRW.&lt;/li&gt;
&lt;li&gt;SOX and the relative strength of MU, NVDA, AVGO, AMAT and LRCX.&lt;/li&gt;
&lt;li&gt;July 14 U.S. CPI composition and the joint response of rates, oil and semiconductors.&lt;/li&gt;
&lt;li&gt;TSMC&amp;rsquo;s July 16 guidance, AI-accelerator demand, CoWoS constraints and capex.&lt;/li&gt;
&lt;li&gt;Late-July AI capex, AI revenue, free cash flow and funding pressure at Microsoft, Meta, Alphabet and Amazon.&lt;/li&gt;
&lt;li&gt;Third-quarter SK hynix HBM4 shipments, yields, ASPs and LTA effects, plus Samsung&amp;rsquo;s HBM qualification and AI5 foundry ramp.&lt;/li&gt;
&lt;li&gt;The distinction between restrictions on new single-stock leveraged ETFs and treatment of existing products.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="12-evidence-boundary"&gt;12. Evidence boundary
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Evidence level&lt;/th&gt;
 &lt;th&gt;Inputs&lt;/th&gt;
 &lt;th&gt;Use&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;td&gt;Closing prices, Kiwoom daily bars, official TSMC monthly sales and company releases&lt;/td&gt;
 &lt;td&gt;Establish facts and direction; do not automatically map into supplier earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;td&gt;Brokerage estimates and major financial-media reporting&lt;/td&gt;
 &lt;td&gt;Forecasts and market interpretation, separated from company facts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Limited&lt;/td&gt;
 &lt;td&gt;Intraday screens, program proxies and ten-day export data&lt;/td&gt;
 &lt;td&gt;Directional signals requiring official close and full-period confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hypothesis&lt;/td&gt;
 &lt;td&gt;Confidential LTA terms, 6x P/B assumptions, bottom and regulatory effects&lt;/td&gt;
 &lt;td&gt;Scenario analysis only; insufficient to change a thesis alone&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;The July 13 decline was neither a Korea-only market-plumbing accident nor proof that physical AI demand disappeared. A global reduction in crowded AI and electronics-hardware exposure set the direction. Lower SK hynix estimates and funding-cost concerns provided direct catalysts, while Korea&amp;rsquo;s leveraged ETFs, margin exposure and program selling magnified the decline.&lt;/p&gt;
&lt;p&gt;Record TSMC monthly revenue, preliminary memory exports and long-term HBM supply evidence indicate that demand remains alive. The market is nevertheless demanding a higher standard of proof: how long profits last, whether LTAs protect margins, whether AI capex can be funded from cash flow, and how much crowded exposure has actually been removed.&lt;/p&gt;
&lt;p&gt;The next conclusion will come from price behavior rather than the number of optimistic headlines. U.S. semiconductor confirmation, the response to CPI and TSMC, a return of foreign cash buying and weaker Korean program selling will determine whether this was a leverage reset or the start of a deeper structural repricing.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Data cutoff: Korea&amp;rsquo;s July 13, 2026 close. Program and selected flow figures are Kiwoom REST proxies pending official KRX confirmation. This article is market-structure research for informational purposes and is not a recommendation to buy or sell any security.&lt;/p&gt;

 &lt;/blockquote&gt;</description></item><item><title>What Rose After Removing Semiconductors? A 320-Stock KOSPI and KOSDAQ YTD Breadth Study</title><link>https://koreainvestinsights.com/post/korea-nonsemiconductor-ytd-leaders-kospi-kosdaq-breadth-2026-07-10/</link><pubDate>Sun, 12 Jul 2026 09:10:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/korea-nonsemiconductor-ytd-leaders-kospi-kosdaq-breadth-2026-07-10/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context
This is a follow-up to our &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;H1 2026 AI infrastructure and narrow-market review&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/kospi-benchmark-hard-to-beat-narrow-market-monte-carlo-2026-06-20/" &gt;Monte Carlo analysis of how hard it was to beat KOSPI&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-market-liquidity-foreign-reallocation-adr-narrow-leadership-2026-06-03/" &gt;Korea liquidity and collapsing breadth study&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/kosdaq-recovery-samsung-hynix-sideways-semiconductor-sme-2026-06-22/" &gt;conditions for a KOSDAQ recovery after Samsung Electronics and SK hynix crowding&lt;/a&gt;. This time, the semiconductor complex and Samsung Electronics/SK hynix equity proxies are removed to identify what actually led the rest of the market. Related hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/korea-daily-market-hub/" &gt;Korea Daily Market Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;As of July 10, 2026, the top KOSPI outperformers outside semiconductors had spread into &lt;strong&gt;power grids, construction, retail, financials and holding companies, and energy&lt;/strong&gt;. The top YTD names were Gaon Cable, Daewoo E&amp;amp;C, Shinsegae, LS ELECTRIC, and Lotte Shopping.&lt;/li&gt;
&lt;li&gt;YTD winners and current leaders are not the same. Among the top KOSPI names, only a much smaller group, including Gaon Cable, Kolmar Korea, GS E&amp;amp;C, S-Oil, KB Financial, GS, Shinhan Financial, and Hana Financial, remained positive over the latest 20 sessions.&lt;/li&gt;
&lt;li&gt;KOSDAQ was far narrower. Only 45 of 138 non-semiconductor names, or 32.6%, were positive YTD. The median YTD return was -12.4% and the median 20-session return was -15.5%, even though the top-ten YTD median was +142.5%.&lt;/li&gt;
&lt;li&gt;The quality of KOSDAQ winners also diverged. FiberPro, Seegene, YG-1, and DKT were supported by earnings or orders. Woori Technology, Solid, and Hyundai Movex combined real business opportunities with thematic premiums. JS Link, Sampyo Cement, Hyundai Bioscience, KMW, and NatureCell were driven more by business-transition, asset-value, clinical, or regulatory options.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Leadership broadened by sector outside semiconductors, but not by stock. The useful screen now is not the YTD ranking. It is the smaller set of names that remain strong over 20 sessions and have earnings or order momentum that can repeat.
 &lt;/div&gt;
&lt;/div&gt;
&lt;p&gt;&lt;img alt="KOSPI and KOSDAQ non-semiconductor breadth comparison" loading="lazy" sizes="(max-width: 767px) calc(100vw - 30px), (max-width: 1023px) 700px, (max-width: 1279px) 950px, 1232px" src="https://koreainvestinsights.com/images/posts/korea-nonsemiconductor-ytd-breadth-2026-07-10.png"&gt;
&lt;/p&gt;
&lt;h2 id="1-universe-and-methodology"&gt;1. Universe and methodology
&lt;/h2&gt;&lt;p&gt;The question is simple: what rose outside semiconductors? A sector-code exclusion alone is not enough because SK Square, Samsung C&amp;amp;T, and Samsung Life remain sensitive to the equity value of Samsung Electronics or SK hynix. This study therefore removes the semiconductor industry, core semiconductor ETF constituents, and major equity-holding proxies.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KOSPI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KOSDAQ&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Starting universe&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Top 200 by market cap&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Top 200 by market cap&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Comparable non-semiconductor companies&lt;/td&gt;
 &lt;td style="text-align: right"&gt;182&lt;/td&gt;
 &lt;td style="text-align: right"&gt;138&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;YTD calculation&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Adjusted close on Dec. 30, 2025 to adjusted close on July 10, 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Same&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Recent trend&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Latest 20-session adjusted return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Same&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Main exclusions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;ETFs, ETNs, preferred shares, KRX semiconductor names, core semiconductor ETF constituents, Samsung Electronics preferred, SK Square, SK Inc., Samsung C&amp;amp;T, Samsung Life&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRX semiconductor names, semiconductor ETF core/multiple constituents, manually classified semiconductor IP, equipment, and materials names&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;No year-end comparison price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;K Bank&lt;/td&gt;
 &lt;td style="text-align: right"&gt;MakinaRocks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Conglomerates such as Doosan and Foosung remain when their primary industry is non-semiconductor, despite partial semiconductor exposure. RFHIC and RF Materials are retained as communications and RF businesses.&lt;/p&gt;
&lt;p&gt;Market-cap ranks use the July 10, 2026 &lt;a class="link" href="https://finance.naver.com/sise/sise_market_sum.naver?sosok=0&amp;amp;page=1" target="_blank" rel="noopener"
 &gt;Naver Finance KOSPI&lt;/a&gt; and &lt;a class="link" href="https://finance.naver.com/sise/sise_market_sum.naver?sosok=1&amp;amp;page=1" target="_blank" rel="noopener"
 &gt;KOSDAQ&lt;/a&gt; screens. Full results are available as &lt;a class="link" href="https://koreainvestinsights.com/data/korea-nonsemiconductor-ytd-2026-07-10/kospi_nonsemi_ytd_20260710.csv" &gt;182-name KOSPI CSV&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/data/korea-nonsemiconductor-ytd-2026-07-10/kosdaq_nonsemi_ytd_20260710.csv" &gt;138-name KOSDAQ CSV&lt;/a&gt;.&lt;/p&gt;
&lt;h2 id="2-kospi-leadership-spread-into-five-non-semiconductor-lanes"&gt;2. KOSPI: leadership spread into five non-semiconductor lanes
&lt;/h2&gt;&lt;p&gt;The top 30 non-semiconductor names include five construction companies, three retailers, and four electrical-equipment or cable names. Financial and holding companies, energy, defense, telecom, and batteries also appear. “Financial and holding companies” is an analytical grouping that combines banks, brokers, insurers, and holding companies rather than one exchange industry code. The full KOSPI non-semiconductor universe was not extremely narrow, with 63.2% positive YTD, but the largest excess returns were concentrated in a handful of sectors.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Market-cap rank&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Latest 20D&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Gaon Cable&lt;/td&gt;
 &lt;td style="text-align: right"&gt;108&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+455.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+81.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;81&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+356.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-20.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Shinsegae&lt;/td&gt;
 &lt;td style="text-align: right"&gt;90&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+154.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;LS ELECTRIC&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+120.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;Lotte Shopping&lt;/td&gt;
 &lt;td style="text-align: right"&gt;107&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+114.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-20.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;SK Eternix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;194&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+107.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;LG Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+98.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-19.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;8&lt;/td&gt;
 &lt;td&gt;Mirae Asset Life Insurance&lt;/td&gt;
 &lt;td style="text-align: right"&gt;148&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+95.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-35.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;9&lt;/td&gt;
 &lt;td&gt;SK Networks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;181&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+89.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-34.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td&gt;Hyundai Department Store&lt;/td&gt;
 &lt;td style="text-align: right"&gt;119&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+88.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;11&lt;/td&gt;
 &lt;td&gt;Samsung E&amp;amp;A&lt;/td&gt;
 &lt;td style="text-align: right"&gt;76&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+83.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;12&lt;/td&gt;
 &lt;td&gt;Foosung&lt;/td&gt;
 &lt;td style="text-align: right"&gt;193&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+82.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-27.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;13&lt;/td&gt;
 &lt;td&gt;Mirae Asset Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;31&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+81.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-19.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;14&lt;/td&gt;
 &lt;td&gt;LIG Defense &amp;amp; Aerospace&lt;/td&gt;
 &lt;td style="text-align: right"&gt;47&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+77.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;15&lt;/td&gt;
 &lt;td&gt;Doosan&lt;/td&gt;
 &lt;td style="text-align: right"&gt;34&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+77.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-27.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;16&lt;/td&gt;
 &lt;td&gt;OCI Holdings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;117&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+71.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-34.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;17&lt;/td&gt;
 &lt;td&gt;Kolmar Korea&lt;/td&gt;
 &lt;td style="text-align: right"&gt;147&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+71.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;18&lt;/td&gt;
 &lt;td&gt;Hyundai G.F. Holdings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;162&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;19&lt;/td&gt;
 &lt;td&gt;SK Telecom&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;20&lt;/td&gt;
 &lt;td&gt;LS Corp.&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+64.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-14.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;21&lt;/td&gt;
 &lt;td&gt;Hyosung Heavy Industries&lt;/td&gt;
 &lt;td style="text-align: right"&gt;27&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+64.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;22&lt;/td&gt;
 &lt;td&gt;Doosan Fuel Cell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;125&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+63.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-35.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;23&lt;/td&gt;
 &lt;td&gt;APR&lt;/td&gt;
 &lt;td style="text-align: right"&gt;55&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+62.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;24&lt;/td&gt;
 &lt;td&gt;GS E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;139&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+61.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;25&lt;/td&gt;
 &lt;td&gt;Samsung SDI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;21&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+61.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-19.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;26&lt;/td&gt;
 &lt;td&gt;S-Oil&lt;/td&gt;
 &lt;td style="text-align: right"&gt;52&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+59.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;27&lt;/td&gt;
 &lt;td&gt;Hyundai Motor&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+54.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-24.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;28&lt;/td&gt;
 &lt;td&gt;DL E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;151&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+53.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-14.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;29&lt;/td&gt;
 &lt;td&gt;Hyundai E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;61&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+52.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-31.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;30&lt;/td&gt;
 &lt;td&gt;NH Investment &amp;amp; Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;62&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+49.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;31&lt;/td&gt;
 &lt;td&gt;Samsung Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+48.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;32&lt;/td&gt;
 &lt;td&gt;KB Financial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+47.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;33&lt;/td&gt;
 &lt;td&gt;Korea Investment Holdings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;59&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+47.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;34&lt;/td&gt;
 &lt;td&gt;DN Automotive&lt;/td&gt;
 &lt;td style="text-align: right"&gt;167&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+46.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;35&lt;/td&gt;
 &lt;td&gt;Sanil Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;92&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+45.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-18.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;36&lt;/td&gt;
 &lt;td&gt;Hanwha Life&lt;/td&gt;
 &lt;td style="text-align: right"&gt;111&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+42.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;37&lt;/td&gt;
 &lt;td&gt;GS Holdings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;79&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+42.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;38&lt;/td&gt;
 &lt;td&gt;Shinhan Financial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;14&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+42.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;39&lt;/td&gt;
 &lt;td&gt;Douzone Bizon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;121&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+40.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;40&lt;/td&gt;
 &lt;td&gt;Hana Financial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+36.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="gaon-cable-requires-an-adjusted-price-caveat"&gt;Gaon Cable requires an adjusted-price caveat
&lt;/h3&gt;&lt;p&gt;Gaon Cable went ex-rights for a bonus issue on June 30, 2026. The company allotted 0.8 new shares per common share, and the exchange set the reference price at KRW190,600. &lt;a class="link" href="https://v.daum.net/v/20260630111702421" target="_blank" rel="noopener"
 &gt;Ex-rights report&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;Its adjusted-price YTD return is +455.8%, while a simple raw-close comparison produces +208.9%. Adjusted prices are appropriate for long-horizon comparisons, but readers should understand why the number differs from a basic chart-screen calculation.&lt;/p&gt;
&lt;h2 id="3-which-kospi-ytd-winners-are-still-leading"&gt;3. Which KOSPI YTD winners are still leading?
&lt;/h2&gt;&lt;p&gt;Only nine of the top 40 remained positive over the latest 20 sessions.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Latest 20D&lt;/th&gt;
 &lt;th&gt;Current read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Gaon Cable&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+455.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+81.7%&lt;/td&gt;
 &lt;td&gt;The strongest grid/cable momentum, with unusually high corporate-action and price volatility&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kolmar Korea&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+71.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23.3%&lt;/td&gt;
 &lt;td&gt;Export consumer and earnings momentum remain intact&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;S-Oil&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+59.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.6%&lt;/td&gt;
 &lt;td&gt;Refining-cycle and margin expectations continue to support returns&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KB Financial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+47.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14.4%&lt;/td&gt;
 &lt;td&gt;Large-cap financial and shareholder-return leadership remains intact&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GS E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+61.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.6%&lt;/td&gt;
 &lt;td&gt;Relatively strong recent price action among construction rerating candidates&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GS Holdings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+42.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.6%&lt;/td&gt;
 &lt;td&gt;Refining, energy asset value, and holding-company discount compression&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shinhan Financial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+42.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.6%&lt;/td&gt;
 &lt;td&gt;Shows that financial leadership is broader than one bank&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hana Financial&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+36.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.2%&lt;/td&gt;
 &lt;td&gt;A key test of the banking rally&amp;rsquo;s durability&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai G.F. Holdings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.6%&lt;/td&gt;
 &lt;td&gt;Retail/holding-company rerating survives, but momentum has slowed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Daewoo E&amp;amp;C, Mirae Asset Life, SK Networks, OCI Holdings, Doosan Fuel Cell, and Hyundai E&amp;amp;C were YTD winners but fell more than 20% over the latest 20 sessions. Their YTD rank increasingly describes past flows rather than current leadership.&lt;/p&gt;
&lt;h2 id="4-kosdaq-bigger-winners-weaker-market"&gt;4. KOSDAQ: bigger winners, weaker market
&lt;/h2&gt;&lt;p&gt;Optical communications, telecom equipment, nuclear and power, electronic components, and healthcare dominated the non-semiconductor KOSDAQ leaders. Yet a small set of spectacular winners coexisted with a broadly weak market.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Market-cap rank&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Latest 20D&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Taihan Fiberoptics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;46&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+491.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-33.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;YG-1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;94&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+324.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+41.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;DKT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;197&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+223.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;Woori Technology&lt;/td&gt;
 &lt;td style="text-align: right"&gt;48&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+208.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-36.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;Samji Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;137&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+145.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;Sung Ho Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;58&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+139.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-52.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;RF Materials&lt;/td&gt;
 &lt;td style="text-align: right"&gt;184&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+132.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-21.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;8&lt;/td&gt;
 &lt;td&gt;JS Link&lt;/td&gt;
 &lt;td style="text-align: right"&gt;67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+118.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;9&lt;/td&gt;
 &lt;td&gt;Seojin System&lt;/td&gt;
 &lt;td style="text-align: right"&gt;28&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+87.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-37.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td&gt;Vitzrocell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;64&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+76.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-33.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;11&lt;/td&gt;
 &lt;td&gt;Sampyo Cement&lt;/td&gt;
 &lt;td style="text-align: right"&gt;85&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-28.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;12&lt;/td&gt;
 &lt;td&gt;InBody&lt;/td&gt;
 &lt;td style="text-align: right"&gt;113&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+60.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;13&lt;/td&gt;
 &lt;td&gt;RFHIC&lt;/td&gt;
 &lt;td style="text-align: right"&gt;65&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+59.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-32.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;14&lt;/td&gt;
 &lt;td&gt;Sphere&lt;/td&gt;
 &lt;td style="text-align: right"&gt;74&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+48.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-40.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;15&lt;/td&gt;
 &lt;td&gt;Hyundai Bioscience&lt;/td&gt;
 &lt;td style="text-align: right"&gt;117&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+46.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-18.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;16&lt;/td&gt;
 &lt;td&gt;Solid&lt;/td&gt;
 &lt;td style="text-align: right"&gt;127&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+44.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-16.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;17&lt;/td&gt;
 &lt;td&gt;Aju IB Investment&lt;/td&gt;
 &lt;td style="text-align: right"&gt;176&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+39.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-50.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;18&lt;/td&gt;
 &lt;td&gt;Mirae Asset Venture Investment&lt;/td&gt;
 &lt;td style="text-align: right"&gt;87&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+33.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-57.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;19&lt;/td&gt;
 &lt;td&gt;Cheryong Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;112&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;20&lt;/td&gt;
 &lt;td&gt;Daejoo Electronic Materials&lt;/td&gt;
 &lt;td style="text-align: right"&gt;68&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-32.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;21&lt;/td&gt;
 &lt;td&gt;Satrec Initiative&lt;/td&gt;
 &lt;td style="text-align: right"&gt;81&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-25.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;22&lt;/td&gt;
 &lt;td&gt;FiberPro&lt;/td&gt;
 &lt;td style="text-align: right"&gt;196&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+25.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;23&lt;/td&gt;
 &lt;td&gt;Cosmecca Korea&lt;/td&gt;
 &lt;td style="text-align: right"&gt;86&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+20.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;24&lt;/td&gt;
 &lt;td&gt;Hyundai Movex&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-40.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;25&lt;/td&gt;
 &lt;td&gt;KMW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;105&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-36.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;26&lt;/td&gt;
 &lt;td&gt;Seegene&lt;/td&gt;
 &lt;td style="text-align: right"&gt;61&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;27&lt;/td&gt;
 &lt;td&gt;Intellian Technologies&lt;/td&gt;
 &lt;td style="text-align: right"&gt;102&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-30.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;28&lt;/td&gt;
 &lt;td&gt;L&amp;amp;C Bio&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+19.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;29&lt;/td&gt;
 &lt;td&gt;EcoPro HN&lt;/td&gt;
 &lt;td style="text-align: right"&gt;139&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;30&lt;/td&gt;
 &lt;td&gt;NatureCell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;57&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-21.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Pearl Abyss ranked 48th in this non-semiconductor group, with a -0.9% YTD return and -10.0% over the latest 20 sessions.&lt;/p&gt;
&lt;h2 id="5-kospi-versus-kosdaq-breadth"&gt;5. KOSPI versus KOSDAQ breadth
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KOSPI non-semiconductor&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KOSDAQ non-semiconductor&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Comparable names&lt;/td&gt;
 &lt;td style="text-align: right"&gt;182&lt;/td&gt;
 &lt;td style="text-align: right"&gt;138&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Median YTD return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;YTD gainers&lt;/td&gt;
 &lt;td style="text-align: right"&gt;115&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;YTD positive share&lt;/td&gt;
 &lt;td style="text-align: right"&gt;63.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;32.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Median latest-20D return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-15.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Latest-20D gainers&lt;/td&gt;
 &lt;td style="text-align: right"&gt;43&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Latest-20D positive share&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Top-ten median YTD return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+110.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+142.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is the central result. KOSDAQ&amp;rsquo;s top ten outperformed KOSPI&amp;rsquo;s top ten, but the KOSDAQ median stock fell 12.4% YTD. Winner magnitude was greater, and the rest of the market was weaker.&lt;/p&gt;
&lt;p&gt;The statement that liquidity rotated beyond semiconductors is partly true for KOSPI, where 63.2% of non-semiconductor names were positive YTD. It is not yet true for KOSDAQ as a whole, where only 32.6% were positive.&lt;/p&gt;
&lt;h2 id="6-why-did-the-selected-kosdaq-winners-rise"&gt;6. Why did the selected KOSDAQ winners rise?
&lt;/h2&gt;&lt;p&gt;The following 13 names are selected cases used to distinguish the quality of each rally, not a fundamental review of every top-30 name. Several high-return companies, including Taihan Fiberoptics, Sung Ho Electronics, and RF Materials, are not classified in this section.&lt;/p&gt;
&lt;h3 id="6-1-earnings-and-orders-supported-the-move"&gt;6-1. Earnings and orders supported the move
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th&gt;Verified change&lt;/th&gt;
 &lt;th&gt;First rejection&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;YG-1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+324.5%&lt;/td&gt;
 &lt;td&gt;1Q26 revenue KRW207.8bn, operating profit KRW38.8bn, 18.7% margin; recovery in global cutting-tool demand and mix&lt;/td&gt;
 &lt;td&gt;Sharp 2Q margin normalization and weaker orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DKT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+223.6%&lt;/td&gt;
 &lt;td&gt;Expansion from OLED components into ESS BMS, battery packs, automotive OLED, and robot charging modules; 2026 estimates revised higher&lt;/td&gt;
 &lt;td&gt;North American ESS expansion fails to convert to revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FiberPro&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+25.9%&lt;/td&gt;
 &lt;td&gt;Defense and space FOG/IMU/INS demand connected to earnings and orders; 1Q revenue about KRW14.3bn, operating profit about KRW4.8bn&lt;/td&gt;
 &lt;td&gt;Backlog conversion delay or one-off margin conclusion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Seegene&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.8%&lt;/td&gt;
 &lt;td&gt;1Q26 revenue KRW129.1bn and operating profit KRW23.6bn; non-respiratory syndromic products plus dividend and buyback cancellation&lt;/td&gt;
 &lt;td&gt;Non-respiratory growth slows and core earnings stall&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;YG-1 has the fastest earnings improvement and the largest price move, but a good company and a good entry price are separate questions. DKT&amp;rsquo;s logic is a 1Q earnings beat combined with ESS BMS, auto, and robotics revenue. &lt;a class="link" href="https://www.newspim.com/news/view/20260604000233" target="_blank" rel="noopener"
 &gt;DKT 1Q and estimate revision&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="6-2-real-business-opportunity-mixed-with-thematic-premium"&gt;6-2. Real business opportunity mixed with thematic premium
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th&gt;Business evidence&lt;/th&gt;
 &lt;th&gt;Remaining test&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Woori Technology&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+208.0%&lt;/td&gt;
 &lt;td&gt;Nuclear MMIS/DCS history, Shin Hanul 3 and 4 revenue, about KRW9.2bn KHNP order&lt;/td&gt;
 &lt;td&gt;Speed of conversion from orders to profit after a 1Q operating loss&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Solid&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+44.9%&lt;/td&gt;
 &lt;td&gt;North American DAS orders, Poland and Saudi expansion, 6G project&lt;/td&gt;
 &lt;td&gt;Whether AI-RAN becomes direct, repeatable revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai Movex&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23.7%&lt;/td&gt;
 &lt;td&gt;Logistics automation, AMR, unmanned warehouses, and KRW55.9bn Kolmar logistics-center order&lt;/td&gt;
 &lt;td&gt;Conversion from 1Q operating loss to profitable backlog execution&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samji Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+145.3%&lt;/td&gt;
 &lt;td&gt;1Q consolidated revenue KRW1.904tn and operating profit KRW334.2bn; electronics distribution generated 97.9% of operating profit&lt;/td&gt;
 &lt;td&gt;Whether the distribution margin spike repeats in 2Q&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samji requires separate treatment. Its &lt;a class="link" href="https://kind.krx.co.kr/external/2026/05/15/000698/20260515001480/11013.htm" target="_blank" rel="noopener"
 &gt;2026 first-quarter filing&lt;/a&gt; shows KRW1.904tn of revenue and KRW334.2bn of operating profit. Electronics distribution contributed KRW327.1bn, or 97.9% of operating profit. The segment margin rose from roughly 2.5% a year earlier to 18.2%.&lt;/p&gt;
&lt;p&gt;Inventory and distribution-margin benefits from higher semiconductor prices are a plausible explanation, but the company did not fully explain the jump. The key question is whether a similar margin repeats in 2Q.&lt;/p&gt;
&lt;p&gt;Solid and KMW also received an AI-RAN premium after Samsung Electronics announced an AI-native network collaboration with NVIDIA. Company-specific direct orders still require separate confirmation. &lt;a class="link" href="https://news.samsung.com/global/samsung-takes-next-stride-toward-ai-native-software-driven-networks-with-nvidia" target="_blank" rel="noopener"
 &gt;Samsung AI-RAN announcement&lt;/a&gt;&lt;/p&gt;
&lt;h3 id="6-3-option-value-led-current-fundamentals"&gt;6-3. Option value led current fundamentals
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th&gt;Option driving price&lt;/th&gt;
 &lt;th&gt;Main risk&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;JS Link&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+118.2%&lt;/td&gt;
 &lt;td&gt;Rare-earth permanent-magnet transition, Malaysia facility, potential US supply-chain entry&lt;/td&gt;
 &lt;td&gt;Loss-making legacy business; certification and mass-production revenue unverified&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sampyo Cement&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65.5%&lt;/td&gt;
 &lt;td&gt;Mixed-use redevelopment of the Sampyo Group&amp;rsquo;s Seongsu ready-mix site&lt;/td&gt;
 &lt;td&gt;No confirmed direct attribution of development profit to Sampyo Cement&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai Bioscience&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+46.2%&lt;/td&gt;
 &lt;td&gt;Xafty dengue phase 2/3 in Vietnam, emergency-use and US trial options&lt;/td&gt;
 &lt;td&gt;Binary clinical and regulatory outcome&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KMW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.9%&lt;/td&gt;
 &lt;td&gt;AI-RAN and Massive MIMO supply-chain expectation&lt;/td&gt;
 &lt;td&gt;Direct AI-RAN orders and 1Q profit recovery unverified&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NatureCell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.6%&lt;/td&gt;
 &lt;td&gt;First-instance court win regarding the JointStem approval rejection and potential FDA path&lt;/td&gt;
 &lt;td&gt;The ruling is not product approval; review and appeal risks remain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;These ideas are not necessarily wrong. Their value, however, changes discretely with a contract, certification, trial, approval, or asset-value attribution. They should not be sized or assessed like recurring-earnings businesses.&lt;/p&gt;
&lt;h2 id="7-only-six-top-30-kosdaq-names-remained-positive-over-20-sessions"&gt;7. Only six top-30 KOSDAQ names remained positive over 20 sessions
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Latest 20D&lt;/th&gt;
 &lt;th&gt;Current status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;YG-1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+324.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+41.2%&lt;/td&gt;
 &lt;td&gt;Earnings and momentum are strong, but expectations are highest&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;DKT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+223.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.7%&lt;/td&gt;
 &lt;td&gt;New-business thesis survives, though momentum slowed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;JS Link&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+118.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.2%&lt;/td&gt;
 &lt;td&gt;Business-transition option still leads price&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;InBody&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+60.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.4%&lt;/td&gt;
 &lt;td&gt;Healthcare quality relative strength remains intact&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;Cosmecca Korea&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+20.6%&lt;/td&gt;
 &lt;td&gt;Export cosmetics and ODM catch-up strength&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;Seegene&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.4%&lt;/td&gt;
 &lt;td&gt;Earnings and capital returns are being priced gradually&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Taihan Fiberoptics, Woori Technology, Seojin System, and RFHIC were major YTD winners but fell more than 30% over the latest 20 sessions. This is why a YTD table cannot be read as a current-leadership table.&lt;/p&gt;
&lt;h2 id="8-three-distinctions-investors-need"&gt;8. Three distinctions investors need
&lt;/h2&gt;&lt;p&gt;First, &lt;strong&gt;sector breadth is not stock breadth&lt;/strong&gt;. KOSPI produced new leadership lanes in grids, construction, retail, financials, and refining, but only a few names remained strong recently.&lt;/p&gt;
&lt;p&gt;Second, &lt;strong&gt;YTD return is not current relative strength&lt;/strong&gt;. Daewoo E&amp;amp;C gained 356.8% YTD but lost 20.1% over 20 sessions. Taihan Fiberoptics gained 491.0% YTD but lost 33.8% recently. A year-to-date winner may be evidence of an earlier flow, not a current entry candidate.&lt;/p&gt;
&lt;p&gt;Third, &lt;strong&gt;recurring earnings and discrete options require different treatment&lt;/strong&gt;. FiberPro, Seegene, and DKT can be tested against quarterly revenue, profit, and orders. Clinical, approval, business-transition, and real-estate options can pay off dramatically but are harder to verify along the way.&lt;/p&gt;
&lt;h2 id="9-research-priority-funnel"&gt;9. Research-priority funnel
&lt;/h2&gt;&lt;p&gt;This is an idea-triage screen, not a final recommendation.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Tier&lt;/th&gt;
 &lt;th&gt;Candidates&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;th&gt;Next evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;A, immediate deeper work&lt;/td&gt;
 &lt;td&gt;Kolmar Korea, S-Oil, KB Financial, FiberPro, Seegene&lt;/td&gt;
 &lt;td&gt;Price action or earnings and order evidence is sufficiently clear to justify earlier follow-up work&lt;/td&gt;
 &lt;td&gt;2Q earnings, estimate revisions, detailed foreign and institutional flows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;B, price and expectations gated&lt;/td&gt;
 &lt;td&gt;Gaon Cable, YG-1, DKT, GS E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Current trend is strong, but price gains and expectations are high&lt;/td&gt;
 &lt;td&gt;Corporate-action adjustment, repeat orders, 2Q margin&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;B, event confirmation&lt;/td&gt;
 &lt;td&gt;Woori Technology, Solid, Hyundai Movex, Samji Electronics&lt;/td&gt;
 &lt;td&gt;Real operating links exist, but profit conversion or repeatability remains open&lt;/td&gt;
 &lt;td&gt;New orders, 2Q margin, exit from losses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C, option watch&lt;/td&gt;
 &lt;td&gt;JS Link, Sampyo Cement, Hyundai Bioscience, KMW, NatureCell&lt;/td&gt;
 &lt;td&gt;Upside can be large, but depends on discrete events&lt;/td&gt;
 &lt;td&gt;Contract, certification, clinical, regulatory, or profit-attribution disclosure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Tier A does not mean “buy now.” It means the company deserves deeper earnings and valuation work before the other candidates. For names such as Gaon Cable and YG-1, company quality and entry-price quality must be separated.&lt;/p&gt;
&lt;h2 id="10-limits-and-falsifiers"&gt;10. Limits and falsifiers
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;This is a cross-sectional study that starts with the top 200 market-cap names in each market, not the full Korean listed universe.&lt;/li&gt;
&lt;li&gt;Semiconductor exclusions combine KRX sectors, ETF inclusion, and manual classification. Conglomerates such as Doosan and Foosung remain despite partial semiconductor exposure.&lt;/li&gt;
&lt;li&gt;Adjusted prices include corporate actions. Gaon Cable requires a separate raw-price interpretation.&lt;/li&gt;
&lt;li&gt;The 20-session return is only a simple trend proxy. Current leadership also requires investor-flow, trading-value, and estimate-revision evidence.&lt;/li&gt;
&lt;li&gt;No new company disclosure was identified that fully explains YG-1&amp;rsquo;s late-June/early-July surge or KMW&amp;rsquo;s July 10 move. Price momentum should not be mistaken for new fundamentals.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The narrow-market diagnosis should be revised if more than half of KOSDAQ non-semiconductor names turn positive YTD and the median 20-session return turns positive. Conversely, if the KOSPI non-semiconductor 20-session positive share falls further and financial, refining, and construction relative strength breaks, the apparent rotation outside semiconductors was only temporary.&lt;/p&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;Removing semiconductors does not leave an empty market. KOSPI clearly produced leadership in grids, construction, retail, financials, holding companies, and refining. Yet few names retained momentum through the latest 20 sessions.&lt;/p&gt;
&lt;p&gt;KOSDAQ is even clearer. Its top ten rose more than KOSPI&amp;rsquo;s top ten, while two-thirds of the non-semiconductor universe remained below year-end levels. This is not a broad KOSDAQ rerating. It is a market where capital concentrated in a small set of stocks, themes, and event groups.&lt;/p&gt;
&lt;p&gt;The next step is not to chase the YTD winner. It is to ask, in order: Is the stock still strong? Can earnings repeat? Do orders convert to revenue? How much expectation is already in the price? Leadership broadened outside semiconductors, but the list of names worth immediate deeper work remains narrow.&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Data as of July 10, 2026. Returns use adjusted closing prices and are rounded to one decimal place. This article is a market-structure study for informational purposes, not a recommendation to buy or sell any security.&lt;/p&gt;

 &lt;/blockquote&gt;</description></item><item><title>The US-China Split in Agentic Inference Infrastructure and the SRAM Opportunity</title><link>https://koreainvestinsights.com/post/us-china-agentic-inference-stack-sram-opportunity-2026-07-09/</link><pubDate>Thu, 09 Jul 2026 17:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/us-china-agentic-inference-stack-sram-opportunity-2026-07-09/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related reading:
&lt;a class="link" href="https://koreainvestinsights.com/en/post/nvidia-vera-rubin-lpx-cmx-inference-stack-samsung-hbm-2026-05-28/" &gt;NVIDIA Vera Rubin inference stack&lt;/a&gt; /
&lt;a class="link" href="https://koreainvestinsights.com/en/post/ai-token-value-memory-value-added-2026-07-09/" &gt;AI token value and memory value capture&lt;/a&gt; /
&lt;a class="link" href="https://koreainvestinsights.com/en/post/ai-datacenter-power-bottleneck-korea-value-chain-screen-2026-07-04/" &gt;AI data center power bottleneck map&lt;/a&gt;&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The United States and China are both scaling AI inference, but they are solving different constraints. The US bottleneck is power, rack density, HBM, advanced packaging, and token-per-watt efficiency. China’s bottleneck is access to leading-edge logic and HBM, so it is more likely to push Ascend-based clusters, optical mesh, parallel scaling, and domestic memory hierarchy workarounds.&lt;/p&gt;
&lt;p&gt;For Korean listed equities, the investable landing zone is not China optical modules by default. The cleaner landing is the US-style AI factory bottleneck: HBM4/HBM4E, power equipment, HBM packaging equipment, and a less appreciated Samsung Electronics option around SRAM/LPU foundry, AI storage, and memory hierarchy.&lt;/p&gt;
&lt;p&gt;The business winner and the best stock setup are not the same thing. SK hynix has the cleanest HBM4 leadership but is already a crowded winner. HD Hyundai Electric has high purity to US data center power bottlenecks, but order momentum is widely recognized. Hanmi Semiconductor is a real HBM tool bottleneck but needs repeat orders and customer diversification. The most asymmetric candidate is Samsung Electronics, if HBM4E, SRAM/LPU foundry, and AI storage start to show up in numbers.&lt;/p&gt;
&lt;h2 id="1-what-is-being-verified"&gt;1. What Is Being Verified
&lt;/h2&gt;&lt;p&gt;This note starts from YS-VC’s article &lt;a href="https://www.ys-vc.com/blog/us-china-ai-inference-stack-divergence"&gt;“Power or Chips: Why US and Chinese AI Inference Stacks Are Diverging”&lt;/a&gt;. The core argument is directionally right: inference scaling is no longer a generic GPU story, and the US and China face different binding constraints.&lt;/p&gt;
&lt;p&gt;The adjustment is in the Korean equity landing. A China optical-mesh story does not automatically translate into Korean optical component revenue. China’s AI infrastructure supply chain is strategically closed and increasingly domestic. Korean equity exposure is more direct through components sold into the US-style AI infrastructure bottleneck.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Claim&lt;/th&gt;
 &lt;th&gt;Assessment&lt;/th&gt;
 &lt;th&gt;Investor read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;US and China AI inference stacks are diverging&lt;/td&gt;
 &lt;td&gt;Mostly true&lt;/td&gt;
 &lt;td&gt;The US optimizes power and rack efficiency; China works around logic and HBM limits&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The US stack is moving toward GPU/HBM/SRAM rack-scale inference&lt;/td&gt;
 &lt;td&gt;True&lt;/td&gt;
 &lt;td&gt;Vera Rubin, LPX, HBM4, and SRAM/LPU become one system&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China is using Ascend, optical mesh, and memory hierarchy workarounds&lt;/td&gt;
 &lt;td&gt;Partly true&lt;/td&gt;
 &lt;td&gt;Directionally plausible, but CloudMatrix performance and reliability need independent evidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;US export controls limit China’s NVIDIA data-center compute access&lt;/td&gt;
 &lt;td&gt;True&lt;/td&gt;
 &lt;td&gt;China cannot count on steady access to top-end GPUs and HBM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM is still a key control point&lt;/td&gt;
 &lt;td&gt;True&lt;/td&gt;
 &lt;td&gt;BIS has explicitly treated HBM as important for AI training and inference at scale&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean opportunity equals China optical modules&lt;/td&gt;
 &lt;td&gt;Too aggressive&lt;/td&gt;
 &lt;td&gt;The cleaner Korean landing is HBM, power, packaging, and Samsung foundry optionality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-why-the-stacks-are-diverging"&gt;2. Why The Stacks Are Diverging
&lt;/h2&gt;&lt;p&gt;Agentic AI increases token volume. Coding agents, research agents, voice agents, and enterprise workflow agents do not just answer once. They read long context, call tools, interpret tool output, and generate new tokens repeatedly. That makes prefill, decode, KV cache, storage, network, and power all matter.&lt;/p&gt;
&lt;p&gt;The bottleneck differs by country.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Area&lt;/th&gt;
 &lt;th&gt;United States&lt;/th&gt;
 &lt;th&gt;China&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Binding constraint&lt;/td&gt;
 &lt;td&gt;Power, rack density, transformers, HBM4, advanced packaging&lt;/td&gt;
 &lt;td&gt;Leading-edge logic, HBM access, export controls, software optimization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Architecture direction&lt;/td&gt;
 &lt;td&gt;Vera Rubin, LPX/LPU, HBM4, high-power AI racks&lt;/td&gt;
 &lt;td&gt;Huawei Ascend, optical mesh, parallel scaling, domestic memory hierarchy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Strength&lt;/td&gt;
 &lt;td&gt;Access to the best GPUs, HBM, and packaging ecosystem&lt;/td&gt;
 &lt;td&gt;Faster power buildout, state-led infrastructure, domestic cloud demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Weakness&lt;/td&gt;
 &lt;td&gt;Time-to-power, grid interconnection, transformer lead times&lt;/td&gt;
 &lt;td&gt;No EUV-based leading logic, HBM limits, closed ecosystem&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean equity landing&lt;/td&gt;
 &lt;td&gt;HBM, power equipment, HBM tools, foundry optionality&lt;/td&gt;
 &lt;td&gt;Limited unless direct suppliers are verified&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The IEA expects data-center electricity demand to reach about 945 TWh by 2030. ([IEA][1]) Power is no longer background infrastructure. It is becoming an upper bound on AI capacity.&lt;/p&gt;
&lt;p&gt;BloombergNEF data cited by Energy Connects says China may add more than 3.4 TW of generation capacity over five years, almost six times the US. The same article says data centers could account for 38% of US power demand growth from 2024 to 2030, versus 6% in China. ([Energy Connects][2]) This is why the US is forced to care about token throughput per MW, while China can lean more heavily on parallel infrastructure buildout.&lt;/p&gt;
&lt;h2 id="3-the-us-stack-hbm-plus-sramlpu"&gt;3. The US Stack: HBM Plus SRAM/LPU
&lt;/h2&gt;&lt;p&gt;The key US signal is NVIDIA’s Vera Rubin and LPX architecture. NVIDIA describes LPX as an inference accelerator for Vera Rubin. It combines Rubin GPUs using HBM with Groq 3 LPX using SRAM-based LPUs. ([NVIDIA LPX][3])&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;NVIDIA LPX disclosure&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Agentic token load&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Up to 15x more tokens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Vera Rubin NVL72 + LPX&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Up to 35x higher throughput per MW&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SRAM per LPU accelerator&lt;/td&gt;
 &lt;td style="text-align: right"&gt;500MB&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SRAM bandwidth per LPU accelerator&lt;/td&gt;
 &lt;td style="text-align: right"&gt;150TB/s&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Scale-up bandwidth per LPU accelerator&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.5TB/s&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LPX rack&lt;/td&gt;
 &lt;td style="text-align: right"&gt;256 LPU chips&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SRAM per LPX rack&lt;/td&gt;
 &lt;td style="text-align: right"&gt;128GB&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DDR5 per LPX rack&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12TB&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SRAM bandwidth per LPX rack&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40PB/s&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is not an HBM replacement story. HBM remains central to the Rubin GPU. SRAM/LPU complements HBM by handling latency-sensitive decode work closer to compute.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Inference stage&lt;/th&gt;
 &lt;th&gt;Simple description&lt;/th&gt;
 &lt;th&gt;Main bottleneck&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Prefill&lt;/td&gt;
 &lt;td&gt;Read the prompt and documents first&lt;/td&gt;
 &lt;td&gt;GPU compute, HBM bandwidth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Decode&lt;/td&gt;
 &lt;td&gt;Generate answers token by token&lt;/td&gt;
 &lt;td&gt;Latency, SRAM, LPU&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KV cache&lt;/td&gt;
 &lt;td&gt;Reuse previous context&lt;/td&gt;
 &lt;td&gt;HBM, DDR, SSD, cache hierarchy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Serving&lt;/td&gt;
 &lt;td&gt;Handle many users reliably&lt;/td&gt;
 &lt;td&gt;Rack network, scheduler, power&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This creates an underappreciated Samsung Electronics angle. Samsung is not only an HBM catch-up candidate. At GTC 2026, Samsung highlighted HBM4 mass production, HBM4E, SOCAMM2 mass production, PCIe 6.0 SSDs, and broader AI computing solutions. ([Samsung Newsroom][4]) If inference moves toward HBM plus SRAM/LPU plus storage hierarchy, Samsung’s combined memory, foundry, packaging, and storage position becomes more relevant.&lt;/p&gt;
&lt;h2 id="4-china-optical-mesh-is-a-competitive-signal-not-a-clean-korean-equity-trade"&gt;4. China: Optical Mesh Is A Competitive Signal, Not A Clean Korean Equity Trade
&lt;/h2&gt;&lt;p&gt;China’s workaround is logical. If access to the latest GPUs and HBM is constrained, the system must rely on more chips, better interconnect, optical mesh, and domestic software optimization. But that is not the same as a direct Korean revenue bridge.&lt;/p&gt;
&lt;p&gt;China’s AI supply chain is strategically local. US export controls also push the system further inward. CloudMatrix-style performance, failure rates, and total cost of ownership still need stronger independent evidence. Therefore, China optical mesh is an important competitive variable, but it is not the best first Korean equity landing.&lt;/p&gt;
&lt;h2 id="5-korean-listed-equity-map"&gt;5. Korean Listed Equity Map
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Bottleneck&lt;/th&gt;
 &lt;th&gt;Korean names&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4/HBM4E&lt;/td&gt;
 &lt;td&gt;Vera Rubin and AI server memory&lt;/td&gt;
 &lt;td&gt;SK hynix, Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Structural beneficiaries. SK hynix leads, Samsung is the catch-up option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SRAM/LPU foundry&lt;/td&gt;
 &lt;td&gt;Low-latency decode accelerators&lt;/td&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Still not fully visible in revenue, but the option is underappreciated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI storage and KV cache&lt;/td&gt;
 &lt;td&gt;Agent memory, eSSD, PCIe 6.0&lt;/td&gt;
 &lt;td&gt;Samsung Electronics, FADU&lt;/td&gt;
 &lt;td&gt;Extends the memory hierarchy below HBM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM packaging tools&lt;/td&gt;
 &lt;td&gt;TC bonders, advanced packaging&lt;/td&gt;
 &lt;td&gt;Hanmi Semiconductor&lt;/td&gt;
 &lt;td&gt;Real bottleneck, but customer concentration and valuation matter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power equipment&lt;/td&gt;
 &lt;td&gt;Transformers, switchgear, grid connection&lt;/td&gt;
 &lt;td&gt;HD Hyundai Electric, LS ELECTRIC, Hyosung Heavy Industries&lt;/td&gt;
 &lt;td&gt;Direct US AI data-center power exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;China optical mesh&lt;/td&gt;
 &lt;td&gt;Optical modules for Chinese clusters&lt;/td&gt;
 &lt;td&gt;Limited direct Korean exposure&lt;/td&gt;
 &lt;td&gt;Avoid unless supplier evidence appears&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="6-power-equipment-is-the-front-end-bottleneck"&gt;6. Power Equipment Is The Front-End Bottleneck
&lt;/h2&gt;&lt;p&gt;In the US, power is a binding constraint. Data centers need grid connection, transformers, switchgear, and high-voltage equipment before GPUs can be monetized.&lt;/p&gt;
&lt;p&gt;HD Hyundai Electric is one of the clearest Korean examples. It raised its 2026 order target by 23%, from $4.222 billion to $5.185 billion, citing North American power infrastructure, 765kV transformers, North American data centers, and onshore generator demand for data centers. ([Seoul Economic Daily][5])&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Factor&lt;/th&gt;
 &lt;th&gt;Investment read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P&lt;/td&gt;
 &lt;td&gt;Power-equipment pricing, high-voltage transformer margins, North America premium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q&lt;/td&gt;
 &lt;td&gt;Data centers, transmission projects, 765kV demand, replacement cycle&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C&lt;/td&gt;
 &lt;td&gt;Raw materials, capacity, delivery timing, quality costs&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;What to watch&lt;/td&gt;
 &lt;td&gt;Book-to-bill, North America orders, margins, 2027 capacity expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Failure mode&lt;/td&gt;
 &lt;td&gt;AI capex delay, grid permitting delay, local sourcing pressure, margin compression&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The company quality and the entry price should still be separated. The theme is visible now. Better setups come when AI capex fears pressure the stock but backlog and pricing remain intact.&lt;/p&gt;
&lt;h2 id="7-hbm4hbm4e-more-tied-to-the-us-stack-than-china"&gt;7. HBM4/HBM4E: More Tied To The US Stack Than China
&lt;/h2&gt;&lt;p&gt;BIS has treated HBM as important for AI training and inference at scale. ([BIS][6]) If China’s access is restricted, allied AI infrastructure becomes more dependent on Korean HBM.&lt;/p&gt;
&lt;p&gt;SK hynix is the clearest leader. Yonhap reported that SK hynix won more than two-thirds of NVIDIA’s HBM orders for a next-generation AI platform. ([Yonhap][7]) But the investment issue is crowding. The market already recognizes SK hynix as the HBM winner.&lt;/p&gt;
&lt;p&gt;Samsung is different. The market has long treated it as an HBM laggard. If HBM4E, SOCAMM2, PCIe 6.0 SSD, foundry, and packaging are repriced as one AI memory-hierarchy platform, the discount can narrow.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Strength&lt;/th&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;th&gt;Current view&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td&gt;HBM4 leadership, NVIDIA allocation, pure HBM exposure&lt;/td&gt;
 &lt;td&gt;Crowding, high expectations, 2027 supply risk&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;HBM4E catch-up, foundry, storage, packaging option&lt;/td&gt;
 &lt;td&gt;HBM qualification delays, foundry losses, customer trust&lt;/td&gt;
 &lt;td&gt;Conditional Buy candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="8-samsungs-sramlpu-foundry-option"&gt;8. Samsung’s SRAM/LPU Foundry Option
&lt;/h2&gt;&lt;p&gt;The non-consensus point is Samsung. If the inference stack moves from GPU-only to HBM plus SRAM/LPU plus cache storage, Samsung is no longer just a memory catch-up story. It is also a foundry and memory-hierarchy option.&lt;/p&gt;
&lt;p&gt;The thesis strengthens if at least two of the following become visible:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th&gt;Evidence to watch&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4/HBM4E customer qualification and shipment&lt;/td&gt;
 &lt;td&gt;Earnings, guidance, HBM mix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LPU or AI ASIC foundry revenue or utilization improvement&lt;/td&gt;
 &lt;td&gt;Foundry loss reduction, HPC customers, AI accelerator production&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI storage and memory-hierarchy revenue&lt;/td&gt;
 &lt;td&gt;SOCAMM2, PCIe 6.0 SSD, eSSD, KV-cache demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;It weakens if HBM4 qualification is delayed by two or more quarters, foundry losses do not narrow, or LPU/AI ASIC revenue does not accrue to Samsung in a material way.&lt;/p&gt;
&lt;h2 id="9-hanmi-semiconductor-real-bottleneck-price-discipline-needed"&gt;9. Hanmi Semiconductor: Real Bottleneck, Price Discipline Needed
&lt;/h2&gt;&lt;p&gt;HBM4 increases packaging complexity. Hanmi Semiconductor is a real TC bonder bottleneck. The Elec reported that Hanmi received its first HBM4 TC bonder order from SK hynix. ([The Elec][8])&lt;/p&gt;
&lt;p&gt;The thesis is sound, but equipment stocks have order-gap risk. Follow-on SK hynix orders, official Samsung or Micron orders, and margin durability matter more than the broad HBM narrative.&lt;/p&gt;
&lt;h2 id="10-practical-view"&gt;10. Practical View
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Priority&lt;/th&gt;
 &lt;th&gt;Exposure&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Conditional Buy candidate&lt;/td&gt;
 &lt;td&gt;HBM4E, SRAM/LPU foundry, and AI storage options may not be fully priced&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;HD Hyundai Electric&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;Direct US data-center power exposure, but order momentum is visible&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;Best HBM4 business position, but crowded winner&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;Hanmi Semiconductor&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;HBM tool bottleneck, but needs repeat orders and customer diversification&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;China optical-mesh beneficiaries&lt;/td&gt;
 &lt;td&gt;Avoid&lt;/td&gt;
 &lt;td&gt;Direct Korean supplier evidence is weak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="conclusion"&gt;Conclusion
&lt;/h2&gt;&lt;p&gt;The Korean opportunity is not “China optical mesh.” It is the US AI factory bottleneck: power, HBM, SRAM/LPU, advanced packaging, and storage hierarchy.&lt;/p&gt;
&lt;p&gt;SK hynix may have the strongest business position. HD Hyundai Electric may have the cleanest power-equipment purity. But the most asymmetric rerating candidate is Samsung Electronics if the market stops viewing it only as an HBM laggard and starts pricing HBM4E, SRAM/LPU foundry, and AI storage together.&lt;/p&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.ys-vc.com/blog/us-china-ai-inference-stack-divergence" target="_blank" rel="noopener"
 &gt;YS-VC, Power or Chips: Why US and Chinese AI inference stacks are diverging&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai" target="_blank" rel="noopener"
 &gt;IEA, Energy demand from AI&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.energyconnects.com/news/renewables/2026/february/china-ramps-up-energy-boom-flagged-by-musk-as-key-to-ai-race/" target="_blank" rel="noopener"
 &gt;Energy Connects, China ramps up energy boom flagged by Musk as key to AI race&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.nvidia.com/en-us/data-center/lpx/" target="_blank" rel="noopener"
 &gt;NVIDIA, NVIDIA LPX&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://news.samsung.com/global/samsung-unveils-hbm4e-showcasing-comprehensive-ai-solutions-nvidia-partnership-and-vision-at-nvidia-gtc-2026" target="_blank" rel="noopener"
 &gt;Samsung Newsroom, Samsung unveils HBM4E and AI solutions at NVIDIA GTC 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.bis.gov/press-release/commerce-strengthens-export-controls-restrict-chinas-capability-produce-advanced-semiconductors-military" target="_blank" rel="noopener"
 &gt;BIS, Commerce strengthens export controls including HBM&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://en.yna.co.kr/view/AEN20260128002800320" target="_blank" rel="noopener"
 &gt;Yonhap, SK hynix reportedly wins over two-thirds of Nvidia HBM orders&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.thelec.net/news/articleView.html?idxno=11132" target="_blank" rel="noopener"
 &gt;The Elec, Hanmi Semiconductor receives first HBM4 TC bonder order&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://en.sedaily.com/finance/2026/07/06/hd-hyundai-electric-raises-2026-order-target-23-percent-on" target="_blank" rel="noopener"
 &gt;Seoul Economic Daily, HD Hyundai Electric raises 2026 order target&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>Samsung and SK Hynix 2028E Profit Valuation: Cheap-Looking Numbers and Cycle Tests</title><link>https://koreainvestinsights.com/post/samsung-hynix-2028e-profit-valuation-cycle-scenarios-2026-07-09/</link><pubDate>Thu, 09 Jul 2026 14:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/samsung-hynix-2028e-profit-valuation-cycle-scenarios-2026-07-09/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context: this is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/ai-token-value-memory-value-added-2026-07-09/" &gt;AI token value and memory value added&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-record-quarter-profit-signals-noise-2026-07-08/" &gt;Samsung 2Q26 signals and noise&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/" &gt;late-July Big Tech earnings and the memory thesis&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 AI memory review&lt;/a&gt;. Relevant hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;Samsung Electronics and SK Hynix look very cheap on 2028E net income. If we use MarketScreener aggregate estimates, 2028E net income is about KRW 439.8tn for Samsung and KRW 346.6tn for SK Hynix. Applying a 9 to 12x P/E gives fair market caps of KRW 3,958tn to 5,277tn for Samsung and KRW 3,119tn to 4,159tn for SK Hynix.&lt;/p&gt;
&lt;p&gt;Against current market caps, that implies roughly +120 to +193% for Samsung and +104 to +172% for SK Hynix. The market is valuing them at only about 4.1x and 4.4x 2028E P/E, respectively.&lt;/p&gt;
&lt;p&gt;But the bottleneck is not the multiple. It is the 2028E profit number. If 2028 earnings are normalized, both stocks are cheap. If 2028 is the cycle peak, a 4 to 5x P/E may not be cheap.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Samsung Electronics&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;MarketScreener 2028E net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 439.8tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 346.6tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;9 to 12x fair market cap&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 3,958tn to 5,277tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 3,119tn to 4,159tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current implied 2028E P/E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 4.1x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 4.4x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Scenario-weighted 2028E net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 272tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 208tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Scenario-weighted 9 to 12x fair market cap&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,444tn to 3,258tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,876tn to 2,502tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Scenario-weighted upside&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About +36 to +81%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About +23 to +64%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The stronger relative setup is Samsung. SK Hynix has higher-quality HBM earnings, but more of that quality is already embedded in the price. Samsung has HBM catch-up, commodity DRAM/NAND leverage and a lower implied 2028E P/E.&lt;/p&gt;
&lt;h2 id="1-current-coordinates"&gt;1. Current Coordinates
&lt;/h2&gt;&lt;p&gt;Samsung&amp;rsquo;s net income belongs economically to both common and preferred shares, so the right denominator is total equity value, not common-share market cap alone. Depending on the public screen, Samsung market cap appears around KRW 1,753tn, while a common-plus-preferred equity value framework is closer to KRW 1,800tn. SK Hynix is around KRW 1,530tn.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Samsung Electronics&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reference share price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 274,000 to 274,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,151,000 to 2,160,000&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Analysis market cap&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 1,800tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 1,530tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026E OP reference&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 371tn to 381tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 280tn to 291tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027E OP reference&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 500tn to 574tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 379tn to 432tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung guided to 2Q26 revenue of KRW 171tn and operating profit of KRW 89.4tn. SK Hynix reported 1Q26 revenue of KRW 52.576tn, operating profit of KRW 37.610tn and net income of KRW 40.346tn, with operating margin of 72%.&lt;/p&gt;
&lt;p&gt;Using net income as roughly 75% of operating profit, current prices imply 2026E P/E near 6 to 7x and 2027E P/E near 4 to 5x. The market is not treating these companies as secular growth stocks. It is treating them as memory companies at or near peak earnings.&lt;/p&gt;
&lt;h2 id="2-the-point-estimate-valuation"&gt;2. The Point-Estimate Valuation
&lt;/h2&gt;&lt;p&gt;MarketScreener aggregate estimates put 2028E net income at KRW 439.781tn for Samsung and KRW 346.562tn for SK Hynix. The same dataset puts 2028E EBIT at KRW 543.635tn and KRW 423.866tn, respectively.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;KRW tn&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2028E net income&lt;/th&gt;
 &lt;th style="text-align: right"&gt;9x fair cap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;10x fair cap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;12x fair cap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current cap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Upside&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;439.8&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,958&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4,398&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5,277&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1,800&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+120% / +144% / +193%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;346.6&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,119&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,466&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4,159&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1,530&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+104% / +127% / +172%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This simple calculation says both stocks are cheap. It also says Samsung is slightly more asymmetric because its implied 2028E P/E is lower.&lt;/p&gt;
&lt;p&gt;The danger is that 2028E is still far away. Memory cycles have repeatedly produced large peak profits that later collapsed. Therefore, the real question is whether 2028E is a normalized profit base or a peak.&lt;/p&gt;
&lt;h2 id="3-how-to-estimate-2028e"&gt;3. How to Estimate 2028E
&lt;/h2&gt;&lt;p&gt;Extending 2027E mechanically is not enough. A top-down “AI capex keeps rising, so memory profit rises” approach is also too thin. Memory profits are often set by supply, not demand. The 2028 supply base is already being shaped by 2026 to 2027 capex.&lt;/p&gt;
&lt;p&gt;The better process is:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Step&lt;/th&gt;
 &lt;th&gt;What to do&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1&lt;/td&gt;
 &lt;td&gt;Split profits into contracted HBM, non-contracted HBM/server DRAM, and spot DRAM/NAND&lt;/td&gt;
 &lt;td&gt;Each bucket has a different earnings floor&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2&lt;/td&gt;
 &lt;td&gt;Track supply instead of forecasting it abstractly&lt;/td&gt;
 &lt;td&gt;2027 capex determines 2028 bit supply&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3&lt;/td&gt;
 &lt;td&gt;Segment demand into AI servers, general servers and consumer devices&lt;/td&gt;
 &lt;td&gt;Price elasticity differs sharply&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;4&lt;/td&gt;
 &lt;td&gt;Force memory-cycle base rates into the model&lt;/td&gt;
 &lt;td&gt;Peak-year earnings historically fall hard&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5&lt;/td&gt;
 &lt;td&gt;Use scenario distributions and reverse implied earnings&lt;/td&gt;
 &lt;td&gt;Current price already embeds a view&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The single most useful future disclosure would be the 2028 HBM long-term agreement coverage ratio. If a large share of 2028 shipment is already contracted, the 2028 profit floor is much higher than in historical cycles.&lt;/p&gt;
&lt;p&gt;The main supply variables are 2027 capex guidance, equipment orders from ASML/AMAT/TEL, Hynix&amp;rsquo;s Yongin and Cheongju investments, Samsung&amp;rsquo;s HBM and packaging catch-up spending, CXMT and YMTC expansion, and whether HBM bottlenecks remain in stacking and packaging rather than wafers alone.&lt;/p&gt;
&lt;p&gt;Demand must also be segmented. AI server demand is relatively inelastic. General server demand is cyclical. PC and smartphone memory demand is price-sensitive and can already show demand destruction when memory prices rise too quickly.&lt;/p&gt;
&lt;p&gt;Historical base rates are harsh.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Cycle&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Operating profit change&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung consolidated 2018 to 2019&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-53%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung DS 2018 to 2019&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-69%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix 2018 to 2019&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-87%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2023 memory downcycle&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Loss-making episodes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This cycle deserves an upward adjustment because of HBM contracts, oligopoly structure, wafer cannibalization by HBM and low AI-server price elasticity. But it also deserves caution because of Chinese supply, consumer demand destruction, policy-driven capacity expansion and the statistical extremity of 70%+ operating margins.&lt;/p&gt;
&lt;h2 id="4-scenario-valuation"&gt;4. Scenario Valuation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Samsung 2028E NI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;9x&lt;/th&gt;
 &lt;th style="text-align: right"&gt;12x&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Upside&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Hynix 2028E NI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;9x&lt;/th&gt;
 &lt;th style="text-align: right"&gt;12x&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Upside&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;S1 shortage persists&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 420tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,780&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5,040&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+110 to +180%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 315tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,835&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,780&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+85 to +147%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;S2 soft landing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 280tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,520&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,360&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+40 to +87%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 215tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,935&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,580&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26 to +69%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;S3 hard landing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 135tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,215&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,620&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-32 to -10%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 110tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;990&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,320&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-35 to -14%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Probability-weighted&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 272tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,444&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3,258&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+36 to +81%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 208tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,876&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2,502&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+23 to +64%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;MarketScreener&amp;rsquo;s estimate sits closer to S1. The probability-weighted model blends S2 and S3 risks. These two approaches do not conflict. One shows how cheap the stocks are if consensus is right. The other shows how much risk discount should be applied to that consensus.&lt;/p&gt;
&lt;h2 id="5-what-the-current-price-already-embeds"&gt;5. What the Current Price Already Embeds
&lt;/h2&gt;&lt;p&gt;Reverse the current market caps by a 9 to 12x P/E and we get the 2028E net income implied by today&amp;rsquo;s prices.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current cap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Implied 2028E NI at 9x&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Implied 2028E NI at 12x&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 1,800tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 200tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 150tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 1,530tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 170tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 127tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Against rough 2027E net income estimates, current prices already imply a meaningful drop.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Rough 2027E NI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Implied 2028E NI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Implied decline&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 400tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 150tn to 200tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-50 to -63%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 310tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 127tn to 170tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-45 to -59%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is the key reframing. The current bet is not necessarily that 2028 earnings will be spectacular. It may simply be that the decline from 2027 will be shallower than the 45 to 63% decline already embedded in prices.&lt;/p&gt;
&lt;h2 id="6-samsung-versus-sk-hynix"&gt;6. Samsung Versus SK Hynix
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Core thesis&lt;/th&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Lower implied 2028E P/E, HBM catch-up, commodity DRAM/NAND leverage&lt;/td&gt;
 &lt;td&gt;HBM4 delay, foundry/system LSI losses, weaker reported-to-core profit conversion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Highest-quality HBM profit pool, customer reference, strong margin&lt;/td&gt;
 &lt;td&gt;More optimism embedded, ADR/event flow, need to prove 2028 LTA coverage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung is not the HBM leader, but that is partly the point. Expectations are lower. If HBM4 qualification, yield and allocation improve, the stock gets both earnings revision and multiple repair. Samsung also has more leverage to conventional DRAM and NAND pricing.&lt;/p&gt;
&lt;p&gt;SK Hynix has the cleaner HBM exposure and better current profit quality. But the market already gives it more credit for structural HBM change. The stock needs clearer evidence that 2028 profit is contracted and not just a peak-cycle extrapolation.&lt;/p&gt;
&lt;h2 id="7-practical-stance"&gt;7. Practical Stance
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;Stance&lt;/th&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;th&gt;What to confirm&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Hold-biased, pullback candidate&lt;/td&gt;
 &lt;td&gt;More asymmetric 2028E valuation and catch-up optionality&lt;/td&gt;
 &lt;td&gt;July 30 segment data, HBM4 progress, NAND/DRAM pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Watchlist, no event-chasing&lt;/td&gt;
 &lt;td&gt;Best HBM quality, but more expectation embedded&lt;/td&gt;
 &lt;td&gt;ADR absorption, 2Q call LTA coverage, HBM4 premium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For Samsung, selling aggressively here requires assigning a very high probability to a 2028 hard landing. The data does not yet justify that. For SK Hynix, the company quality is clear, but ADR, new-share absorption and 2028 contract visibility matter before chasing.&lt;/p&gt;
&lt;h2 id="8-invalidation"&gt;8. Invalidation
&lt;/h2&gt;&lt;p&gt;The thesis weakens if two or more of the following happen.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Invalidation signal&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2027 capex across the three major memory firms rises more than 50% YoY&lt;/td&gt;
 &lt;td&gt;Supply discipline breaks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CXMT qualifies high-end server DDR5 with major customers&lt;/td&gt;
 &lt;td&gt;Commodity DRAM ceiling falls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Chinese suppliers make visible HBM progress&lt;/td&gt;
 &lt;td&gt;Premium durability weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Big Tech capex guidance slows sharply&lt;/td&gt;
 &lt;td&gt;AI server demand downside&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;No disclosure or evidence of 2028 HBM LTA coverage&lt;/td&gt;
 &lt;td&gt;Peak-profit discount remains&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;YMTC NAND ramp accelerates&lt;/td&gt;
 &lt;td&gt;NAND pricing risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;Samsung and SK Hynix both look cheap on 2028E earnings. But cheap-looking is not the same as cheap.&lt;/p&gt;
&lt;p&gt;Using MarketScreener point estimates, both have roughly double-like potential. Using a supply, contract and base-rate scenario model, the upside narrows to +36 to +81% for Samsung and +23 to +64% for SK Hynix. That is still attractive, but the margin of safety is smaller.&lt;/p&gt;
&lt;p&gt;The key is 2028E. It cannot be built from demand stories alone. It depends on HBM LTA coverage, 2027 capex discipline, DRAM/NAND pricing, Chinese supply and Big Tech capex.&lt;/p&gt;
&lt;p&gt;The relative preference is Samsung. SK Hynix has better HBM earnings, but requires more belief. Samsung has catch-up optionality, broader DRAM/NAND leverage and a lower implied 2028E P/E. For both names, the next checks are late-July Big Tech capex, Samsung&amp;rsquo;s July 30 detailed results, SK Hynix ADR absorption and the 2Q call.&lt;/p&gt;
&lt;h2 id="evidence-and-limits"&gt;Evidence and Limits
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Classification&lt;/th&gt;
 &lt;th&gt;Content&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fact&lt;/td&gt;
 &lt;td&gt;Samsung 2Q26 preliminary revenue KRW 171tn and OP KRW 89.4tn. SK Hynix 1Q26 revenue KRW 52.576tn, OP KRW 37.610tn and net income KRW 40.346tn. MarketScreener 2028E net income of KRW 439.781tn for Samsung and KRW 346.562tn for SK Hynix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Inference&lt;/td&gt;
 &lt;td&gt;Net income approximated at 75% of OP. Current prices imply 2028E net income of KRW 150-200tn for Samsung and KRW 127-170tn for SK Hynix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Speculation&lt;/td&gt;
 &lt;td&gt;Scenario probabilities and 2028E net income assumptions: S1 25%, S2 45%, S3 30%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Blocked&lt;/td&gt;
 &lt;td&gt;2028 HBM LTA coverage, customer allocations, actual contract prices and detailed analyst-by-analyst 2028E assumptions are not fully available from public data&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="sources"&gt;Sources
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://news.samsung.com/global/samsung-electronics-announces-earnings-guidance-for-second-quarter-2026" target="_blank" rel="noopener"
 &gt;Samsung Global Newsroom, 2Q26 earnings guidance&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.prnewswire.com/news-releases/sk-hynix-announces-1q26-financial-results-302750959.html" target="_blank" rel="noopener"
 &gt;SK hynix 1Q26 financial results&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.marketscreener.com/quote/stock/SAMSUNG-ELECTRONICS-CO-LT-6494906/finances/" target="_blank" rel="noopener"
 &gt;MarketScreener, Samsung Electronics financial forecasts&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.reuters.com/world/asia-pacific/samsung-estimates-19-fold-rise-q2-operating-profit-beating-expectations-2026-07-06/" target="_blank" rel="noopener"
 &gt;Reuters, Samsung 2Q26 profit reaction&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://en.yna.co.kr/view/AEN20260701005500320" target="_blank" rel="noopener"
 &gt;Yonhap, SK hynix F-1/A and HBM share reference&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>AI Token Value Today and Tomorrow: Value Added for Memory Companies</title><link>https://koreainvestinsights.com/post/ai-token-value-memory-value-added-2026-07-09/</link><pubDate>Thu, 09 Jul 2026 09:00:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/ai-token-value-memory-value-added-2026-07-09/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context: this article follows our work on &lt;a class="link" href="https://koreainvestinsights.com/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/" &gt;late July big tech earnings and the memory thesis&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron&amp;rsquo;s FY3Q26 AI memory earnings&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;the 1H26 AI infrastructure bottleneck review&lt;/a&gt;. Related hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt;. The question is not simply whether AI demand is strong. The question is who keeps the economic value created by AI tokens.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;If we decompose one current dollar of AI token revenue, the largest share of value added still sits at the model and application layer. The equity market, however, is being driven by data center investment that is being pulled forward before final application revenue has fully arrived. That is why today’s AI value chain looks less like a normal pyramid and more like an inverted one.&lt;/p&gt;
&lt;p&gt;The current value-added split of one dollar of AI token revenue is roughly as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current value-added estimate&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;AI model company&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45 to 55 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud and data center&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10 to 16 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU supplier&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 13 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM, server DRAM, SSD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 2.3 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Other infrastructure software and storage&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3 to 5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1 cent&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.3 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The market capitalization and capex cycle cannot be explained by this table alone. Current AI service revenue is estimated around 60 to 100 billion dollars per year, while the leading GPU supplier’s data center revenue run rate has reached around 300 billion dollars, and the big four cloud companies’ 2026 capex is estimated near 725 billion dollars. The industry is building future demand before current revenue catches up.&lt;/p&gt;
&lt;p&gt;Memory is a real bottleneck in this structure. But if memory captures only about 2 to 5 cents of value added per token dollar, the central question for memory stocks is not “is the P/E low?” The real question is whether the industry can sustain mid-cycle net income of roughly 100 to 140 trillion won.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Where does AI token value accrue?&lt;/td&gt;
 &lt;td&gt;Near term: models and cloud. Long term: workflow applications and physical bottlenecks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is memory a structural beneficiary?&lt;/td&gt;
 &lt;td&gt;Yes, but its value-added share is small and the stocks already price in a strong mid-cycle profit base&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;What should investors monitor?&lt;/td&gt;
 &lt;td&gt;HBM4 contract pricing, customer allocation, blended memory ASP, cloud backlog, and application revenue conversion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="1-the-real-question-who-keeps-the-dollar"&gt;1. The Real Question: Who Keeps the Dollar?
&lt;/h2&gt;&lt;p&gt;The most common mistake in AI investing is mixing up revenue attribution with value-added attribution. If a customer pays one dollar for an AI service, that dollar may first appear as revenue at a model company. But inside that dollar are cloud costs, GPU depreciation, memory, foundry capacity, power, networking, servers, cooling, labor, software, taxes, and financing costs.&lt;/p&gt;
&lt;p&gt;So “who books the revenue?” is not the same question as “who keeps the surplus?”&lt;/p&gt;
&lt;p&gt;This article uses five rules.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Rule&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue attribution&lt;/td&gt;
 &lt;td&gt;The first seller that invoices the customer&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Value-added attribution&lt;/td&gt;
 &lt;td&gt;The economic value left after removing double counting&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Token dollar&lt;/td&gt;
 &lt;td&gt;A dollar of API or model usage revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Application dollar&lt;/td&gt;
 &lt;td&gt;A dollar of final workflow or agent software spend&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Depreciation basis&lt;/td&gt;
 &lt;td&gt;Data center capex is annualized over a five-year asset life&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The capex treatment matters. AI data center spending does not disappear as an immediate cost. It is capitalized into GPUs, servers, power gear, buildings, cooling, and networks, then depreciated over several years. Therefore, the right cost in a token-dollar model is annual depreciation, not total capex.&lt;/p&gt;
&lt;p&gt;Time also matters. Inference cost in 2025 is not the same as inference cost in 2028. Cost per unit of performance is falling quickly.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Inference cost as share of revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Model gross margin&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50 to 67%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;33 to 50%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40 to 55%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45 to 60%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027 to 2028&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30 to 40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;60 to 70%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This leads to an important point. Token prices can fall while model gross margins rise, if inference costs fall faster than prices. Much of that margin is then reinvested into training, compute reservations, and customer acquisition. The value does not simply stay on the income statement. It flows back into GPUs, memory, power, and foundry capacity.&lt;/p&gt;
&lt;h2 id="2-the-current-value-waterfall"&gt;2. The Current Value Waterfall
&lt;/h2&gt;&lt;p&gt;The table below decomposes one current AI token dollar. Revenue attribution is the amount a layer effectively invoices or absorbs. Value added is the economic share after removing double counting.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue attribution&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value-added estimate&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;AI model company&lt;/td&gt;
 &lt;td style="text-align: right"&gt;100 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45 to 55 cents&lt;/td&gt;
 &lt;td&gt;Direct customer layer. Falling inference cost is the largest lever&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud and data center&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 50 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10 to 16 cents&lt;/td&gt;
 &lt;td&gt;Operating layer that bundles GPUs, networks, power, and buildings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU supplier&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 18 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 13 cents&lt;/td&gt;
 &lt;td&gt;Main physical bottleneck today&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM, server DRAM, SSD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 3 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 2.3 cents&lt;/td&gt;
 &lt;td&gt;Critical to system performance, but small as a token-dollar share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Infrastructure software, storage, security, observability&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 5 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 3 cents&lt;/td&gt;
 &lt;td&gt;More important as operations become more complex&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.5 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1 cent&lt;/td&gt;
 &lt;td&gt;Common bottleneck across AI chip designs&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server OEM and ODM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 3 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.4 cents&lt;/td&gt;
 &lt;td&gt;High throughput, limited margins&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Third-party networking&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.3 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 0.7 cents&lt;/td&gt;
 &lt;td&gt;Important at cluster scale, but supplier power varies&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 2.5 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.3 cents&lt;/td&gt;
 &lt;td&gt;Small cost line, large site bottleneck&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CPU and IP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 0.5 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 0.3 cents&lt;/td&gt;
 &lt;td&gt;Control and support compute&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Construction, components, materials, residual&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 9 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Distributed&lt;/td&gt;
 &lt;td&gt;Lower layers of the data center stack&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Two observations matter.&lt;/p&gt;
&lt;p&gt;First, power is not the largest cost item inside a token dollar. But it is one of the largest physical bottlenecks. Power matters through grid access, substations, transmission, permits, and long-term power contracts.&lt;/p&gt;
&lt;p&gt;Second, memory’s value-added share per token dollar is still small. This does not mean HBM is unimportant. HBM is indispensable. But the economic share retained by memory companies is far smaller than the share retained by GPUs or model/application layers. If the equity market ignores that difference, valuation errors appear.&lt;/p&gt;
&lt;h2 id="3-the-inverted-pyramid"&gt;3. The Inverted Pyramid
&lt;/h2&gt;&lt;p&gt;The current AI stack looks like this.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Estimated scale&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;AI model and application revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;60 to 100 billion dollars per year&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU data center revenue run rate&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 300 billion dollars&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 big four cloud capex&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 725 billion dollars&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI revenue needed to normalize 700 billion dollars of capex&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 2.8 to 3.0 trillion dollars&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;If AI service revenue is currently 60 to 100 billion dollars, and memory revenue is 3 to 7 cents per token dollar, AI memory flow revenue should be only about 2 to 7 billion dollars per year. Yet the leading memory companies are producing revenue and profit numbers far above what current token revenue alone can explain.&lt;/p&gt;
&lt;p&gt;For example, SK Hynix reported 97.1 trillion won of revenue in 2025 and 52.6 trillion won of revenue in the first quarter of 2026. That scale is hard to explain only from current AI token revenue. Much of today’s memory revenue is therefore tied to capex being pulled forward for future AI demand.&lt;/p&gt;
&lt;p&gt;This is not automatically wrong. Data centers cannot be built instantly after demand appears. Power, land, substations, GPUs, servers, cooling, and networking must be secured years in advance. Pre-investment is necessary.&lt;/p&gt;
&lt;p&gt;The question is speed. If application and cloud revenue catch up, today’s capex is normal growth investment. If final demand lags, the correction begins upstream: memory and server components first, then GPUs, then cloud capex plans.&lt;/p&gt;
&lt;h2 id="4-three-convergence-paths"&gt;4. Three Convergence Paths
&lt;/h2&gt;&lt;p&gt;AI revenue is not determined by token price alone.&lt;/p&gt;
&lt;p&gt;AI revenue = token volume × token price&lt;/p&gt;
&lt;p&gt;Token price per unit of performance is falling quickly. It may decline by roughly 10x per year through the early phase, then slow to 1.5 to 2x per year after 2027. But volume can explode at the same time. A simple question may use a few hundred tokens. A workflow, document process, coding task, or agent job can use tens of thousands, hundreds of thousands, or even more than a million tokens.&lt;/p&gt;
&lt;p&gt;That means price declines do not automatically mean revenue declines. If price falls by 90% but volume rises 100x, revenue still rises 10x.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability sense&lt;/th&gt;
 &lt;th&gt;Description&lt;/th&gt;
 &lt;th&gt;Market implication&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Demand convergence&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45%&lt;/td&gt;
 &lt;td&gt;Application revenue scales quickly and 1.0 to 1.6 trillion dollars of capex normalizes&lt;/td&gt;
 &lt;td&gt;Limited upstream drawdown, bottleneck premium remains&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Digestion then reacceleration&lt;/td&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td&gt;2027 to 2028 capex pauses by 20 to 30%, then workflow AI demand returns&lt;/td&gt;
 &lt;td&gt;Memory ASP could correct 40 to 60%, then recover&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Structural oversupply&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td&gt;Usage elasticity disappoints and data center buildout proves excessive&lt;/td&gt;
 &lt;td&gt;Capex cuts of 50% or more and upstream recession&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;On-device AI changes the distribution, not the whole demand picture. More inference on phones, PCs, cars, robots, and industrial devices can reduce some cloud token revenue. But silicon demand does not disappear. LPDDR, mobile SoCs, edge NPUs, embedded storage, and local servers may all benefit.&lt;/p&gt;
&lt;h2 id="5-long-run-equilibrium-workflows-capture-the-dollar"&gt;5. Long-Run Equilibrium: Workflows Capture the Dollar
&lt;/h2&gt;&lt;p&gt;Over the long run, tokens are unlikely to remain scarce. As token prices fall and model performance gaps narrow, value moves from token production toward workflow ownership.&lt;/p&gt;
&lt;p&gt;On a pure API dollar, the long-run value split may look like this.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Long-run value-added estimate&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Model and API company&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45 to 55 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud and data center&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12 to 15 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU supplier&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8 to 10 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Custom ASIC&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4 to 5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4 to 5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3 to 4 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power and data center infrastructure&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4 to 5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CPU and IP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server OEM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Other infrastructure software&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4 to 6 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Residual&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 5 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;But customers do not really buy tokens. They buy outcomes: customer support, coding productivity, claims processing, contract review, factory quality control, and logistics automation. On a final application dollar, the distribution changes.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Long-run revenue attribution&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Long-run value added&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Workflow applications, agents, distribution&lt;/td&gt;
 &lt;td style="text-align: right"&gt;100 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;45 to 50 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Model and API&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35 to 40 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18 to 24 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud and data center&lt;/td&gt;
 &lt;td style="text-align: right"&gt;13 to 15 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4 to 6 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total silicon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9 to 12 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6 to 8 cents&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power, land, cooling&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.5 to 2 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1 cent&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server OEM and generic infrastructure&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1 cent&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Low&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Other software&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3 to 5 cents&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The long-run conclusion is simple: tokens commoditize, while workflow ownership and physical bottlenecks retain value.&lt;/p&gt;
&lt;h2 id="6-capture-tiers"&gt;6. Capture Tiers
&lt;/h2&gt;&lt;p&gt;The best AI businesses are not simply the ones with the most revenue. The best businesses have three traits.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Customers cannot switch easily.&lt;/li&gt;
&lt;li&gt;Cost declines become margin expansion.&lt;/li&gt;
&lt;li&gt;Supply cannot expand quickly when demand rises.&lt;/li&gt;
&lt;/ol&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Tier&lt;/th&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Tier 1&lt;/td&gt;
 &lt;td&gt;Foundry, workflow applications, integrated cloud platforms&lt;/td&gt;
 &lt;td&gt;Bottleneck supply, customer lock-in, full-stack control&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Tier 2&lt;/td&gt;
 &lt;td&gt;GPU, platform AI suppliers, HBM leaders, custom ASIC&lt;/td&gt;
 &lt;td&gt;Current bottlenecks and high growth, but with price and competition risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Tier 3&lt;/td&gt;
 &lt;td&gt;Server OEM, neocloud, commodity memory, regulated utilities&lt;/td&gt;
 &lt;td&gt;Volume grows, but rent capture is weaker&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Foundry belongs in Tier 1 because AI wafer demand exists whether the winning chip is a GPU or a custom ASIC. Workflow applications belong there because model costs can fall while end-user value remains tied to business outcomes.&lt;/p&gt;
&lt;p&gt;Integrated cloud platforms need case-by-case judgment. Pure GPU rental can become a capital-heavy utility. But a company that owns search, ads, productivity software, cloud, chips, models, and data centers can turn capex into margin.&lt;/p&gt;
&lt;h2 id="7-multiples-cheap-looking-is-not-the-same-as-cheap"&gt;7. Multiples: Cheap Looking Is Not the Same as Cheap
&lt;/h2&gt;&lt;p&gt;AI value-chain multiples should differ by layer.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Appropriate valuation lens&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry&lt;/td&gt;
 &lt;td&gt;EV/S 8 to 11x, P/E 20 to 26x if growth persists&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU supplier&lt;/td&gt;
 &lt;td&gt;Normalized EV/S 9 to 13x, scarcity 15 to 18x, P/E 18 to 24x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Integrated cloud&lt;/td&gt;
 &lt;td&gt;EV/S 5 to 8x, full-stack 8 to 10x, P/E 20 to 28x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Platform AI supplier&lt;/td&gt;
 &lt;td&gt;Hypergrowth EV/ARR 10 to 18x, deceleration 5 to 8x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Workflow AI software&lt;/td&gt;
 &lt;td&gt;Simple tools 2 to 4x, workflow owners 8 to 12x, regulated data owners 12 to 20x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM and memory&lt;/td&gt;
 &lt;td&gt;EV/S 3 to 5x, leaders 5 to 7x, mid-cycle P/E 8 to 14x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Storage&lt;/td&gt;
 &lt;td&gt;EV/S 2 to 3.5x, platform 4 to 7x, normalized P/E 8 to 14x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;CPU and IP&lt;/td&gt;
 &lt;td&gt;CPU 12 to 18x, IP conditionally 30 to 50x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server OEM&lt;/td&gt;
 &lt;td&gt;EV/S 0.5 to 1.5x, P/E 8 to 15x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power and data center infrastructure&lt;/td&gt;
 &lt;td&gt;Contracted assets at EV/EBITDA 10 to 18x or project IRR&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Memory is the layer where investors should be most careful. A 5 to 7x P/E can signal cheapness. At a cycle peak, it can also simply mean that the earnings denominator is temporarily inflated.&lt;/p&gt;
&lt;p&gt;Useful valuation formulas include:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Formula&lt;/th&gt;
 &lt;th&gt;Use&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;EV/NOPAT = (1 - g/ROIC) / (WACC - g)&lt;/td&gt;
 &lt;td&gt;Normalized multiple with growth and capital efficiency&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Justified PBR = (sustainable ROE - g) / (COE - g)&lt;/td&gt;
 &lt;td&gt;Long-run capital-intensive industry PBR&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mid-cycle EPS = (peak + trough + 2×normal) / 4&lt;/td&gt;
 &lt;td&gt;Cycle-aware earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pre-profit value = terminal NOPAT × terminal multiple / (1 + discount rate)^t&lt;/td&gt;
 &lt;td&gt;Early-stage or loss-making firms&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Capital costs also differ.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th style="text-align: right"&gt;WACC or discount rate&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyperscalers&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.5 to 9.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU and foundry&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9 to 10.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10.5 to 12%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Vertical AI software&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12 to 18%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Loss-making model layer&lt;/td&gt;
 &lt;td style="text-align: right"&gt;14 to 20%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="8-the-memory-question-can-mid-cycle-net-income-reach-100-to-140-trillion-won"&gt;8. The Memory Question: Can Mid-Cycle Net Income Reach 100 to 140 Trillion Won?
&lt;/h2&gt;&lt;p&gt;For memory companies, the key question is no longer whether HBM is good. The market already knows that. The key question is whether the current market cap requires a realistic or unrealistic mid-cycle profit base.&lt;/p&gt;
&lt;p&gt;Suppose a leading memory company is valued near 1,400 trillion won. To sustain that value after one or two peak years, mid-cycle net income must be very large.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Assumption&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Peak net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;240 trillion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Trough net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50 trillion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Normalized net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;60 trillion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mid-cycle net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;(240 + 50 + 2×60) / 4 = about 102 trillion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Applying an 8 to 14x mid-cycle multiple gives a fair-value range of about 820 to 1,430 trillion won. If the current market cap is already near the top of that range, the stock is not automatically expensive, but the margin of safety is thin.&lt;/p&gt;
&lt;p&gt;Further upside requires one or more of the following.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Requirement&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Evidence that mid-cycle net income can exceed 140 trillion won&lt;/td&gt;
 &lt;td&gt;HBM premium and high-margin mix remain after the peak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4 custom premium is defended&lt;/td&gt;
 &lt;td&gt;Product customization protects pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ASP holds after Samsung’s supply entry&lt;/td&gt;
 &lt;td&gt;More supply does not crush pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;General DRAM and NAND pricing also improves&lt;/td&gt;
 &lt;td&gt;The cycle broadens beyond HBM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Long-term customer agreements and volume commitments&lt;/td&gt;
 &lt;td&gt;Downcycle earnings risk falls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="9-samsung-sk-hynix-and-micron-are-not-the-same-trade"&gt;9. Samsung, SK Hynix, and Micron Are Not the Same Trade
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Core logic&lt;/th&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;HBM share recovery and memory mix improvement. Expectations are lower than for the leader&lt;/td&gt;
 &lt;td&gt;HBM qualification, yield, foundry and system LSI losses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Highest HBM profit quality and customer reference&lt;/td&gt;
 &lt;td&gt;Much of the good news is already in the price&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td&gt;Leading indicator for AI memory contracts, pricing, FCF, and capex&lt;/td&gt;
 &lt;td&gt;Volatility, short interest, and supply-cycle sensitivity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung has more catch-up optionality. SK Hynix has the highest-quality memory profit pool, but expectations are embedded. Micron is the canary because its pricing, long-term agreements, data center revenue, and capex guide influence Korean memory estimates directly.&lt;/p&gt;
&lt;h2 id="10-cloud-power-and-vertical-ai"&gt;10. Cloud, Power, and Vertical AI
&lt;/h2&gt;&lt;p&gt;Cloud companies split into two groups: integrated platforms and capital-heavy rental platforms. The most important numbers are backlog, revenue conversion, operating margin, internal silicon mix, free cash flow, and bond issuance.&lt;/p&gt;
&lt;p&gt;Power is a small token-dollar cost line but a large site bottleneck. Grid access, substations, transmission, PPAs, gas turbines, nuclear, SMRs, water, land, cooling, and permits matter more than the simple power bill.&lt;/p&gt;
&lt;p&gt;Vertical AI companies can capture the largest long-run share if they own the workflow. The winners will have regulated or physical domains, proprietary data, action authority, measurable ROI, model replaceability, a path to 70%+ gross margin, and compliance moats. Generic chat wrappers, basic retrieval integration, API resale, data-less automation, GPU brokerage, and server assembly should be treated with caution.&lt;/p&gt;
&lt;h2 id="11-what-to-monitor"&gt;11. What to Monitor
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Combined cloud backlog&lt;/td&gt;
 &lt;td&gt;Whether capex is tied to future revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Self-built capex ratio&lt;/td&gt;
 &lt;td&gt;How much demand is moving outside public cloud&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;On-device inference share&lt;/td&gt;
 &lt;td&gt;Cloud token revenue substitution risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4 contract pricing&lt;/td&gt;
 &lt;td&gt;Durability of memory premium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer allocation by HBM supplier&lt;/td&gt;
 &lt;td&gt;Supplier bargaining power&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Blended memory ASP&lt;/td&gt;
 &lt;td&gt;Whether the cycle is broadening&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU lead times&lt;/td&gt;
 &lt;td&gt;Persistence of scarcity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud operating margin&lt;/td&gt;
 &lt;td&gt;Whether capex turns into profit&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI application revenue growth&lt;/td&gt;
 &lt;td&gt;Whether final demand catches infrastructure spending&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The thesis weakens if open-weight systems compress model pricing faster than expected, on-device AI reduces cloud demand more quickly, 2027 cloud capex slows sharply, HBM4 premiums compress, or AI application revenue fails to scale.&lt;/p&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;The AI token economy can be summarized in one sentence:&lt;/p&gt;
&lt;p&gt;Tokens get cheaper, while value moves to workflow ownership and hard-to-expand physical bottlenecks.&lt;/p&gt;
&lt;p&gt;Memory is one of those bottlenecks. HBM, high-performance DRAM, and SSDs are essential to AI servers. But memory’s value-added share per token dollar remains limited. That is why memory stocks require more than the statement that HBM is strong. Current prices already demand a strong mid-cycle profit base.&lt;/p&gt;
&lt;p&gt;The four key questions are:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Does HBM4 preserve custom pricing?&lt;/li&gt;
&lt;li&gt;Do leading suppliers hold ASP and margins after Samsung’s entry?&lt;/li&gt;
&lt;li&gt;Do cloud backlog and AI application revenue catch up to capex?&lt;/li&gt;
&lt;li&gt;Can memory companies sustain mid-cycle net income of 100 to 140 trillion won?&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;If those four conditions are met, today’s large memory market caps can be justified. If not, low P/E ratios may prove to be a cycle-peak illusion rather than a margin of safety.&lt;/p&gt;</description></item><item><title>Samsung Electronics recorded the world's largest quarterly operating profit. Signals and noise poured in today.</title><link>https://koreainvestinsights.com/post/samsung-record-quarter-profit-signals-noise-2026-07-08/</link><pubDate>Wed, 08 Jul 2026 22:35:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/samsung-record-quarter-profit-signals-noise-2026-07-08/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to the Samsung 2Q26 preview, Micron FY3Q26 review, NVIDIA earnings analogue work, late-July Big Tech earnings-call scenario map, and Big Tech financing relay note. Related hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;Samsung Electronics presented provisional sales of 171 trillion won and operating profit of 89.4 trillion won for the second quarter of 2026. Based on Samsung Electronics&amp;rsquo; official newsroom, this is a very large number compared to the operating profit of 57.23 trillion won in the previous quarter and 4.68 trillion won in the same period last year. Looking at memory prices, AI server demand, hyperscaler investment, and HBM4 ramp-up together, the quality of Samsung Electronics&amp;rsquo; investment argument has actually become stronger.&lt;/p&gt;
&lt;p&gt;However, the stock price moved in the opposite direction. Based on local DB, Samsung Electronics closed at 296,000 won on July 7 and 277,500 won on July 8. On July 7, foreigners net sold about 1.82 trillion won, and on July 8, they sold about 874.1 billion won. Individuals made net purchases of approximately 2.32 trillion won and 584.7 billion won, respectively. Although good performance was achieved, foreign and programmatic supply and demand saw peak-out, interest rates, exchange rates, and overcrowding of positions before profit levels.&lt;/p&gt;
&lt;p&gt;The conclusion of this article is as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Judgment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics Investment Argument&lt;/td&gt;
 &lt;td&gt;enforce. DRAM/NAND prices, AI servers, and hyperscaler CapEx are all favorable.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Short term timing&lt;/td&gt;
 &lt;td&gt;damage. Despite good performance, there was selling pressure for two days in a row.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;supply and demand&lt;/td&gt;
 &lt;td&gt;danger. Foreign selling and private absorption are conflicting, and the Korean market is weak.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sell ​​decision&lt;/td&gt;
 &lt;td&gt;no. The current data is closer to digesting expectations and supply and demand pressures than to damaging the thesis.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Additional purchase decision&lt;/td&gt;
 &lt;td&gt;Not yet. You must check ASML, TSMC, Samsung Electronics IR, foreign supply and demand, and exchange rates and interest rates.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;action&lt;/td&gt;
 &lt;td&gt;Strong holding, freeze on additional purchases, make a decision after confirming the next gateway.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;To sum it up in one line, the information released today is not “Samsung Electronics has worsened,” but is closer to “the market has already had too many good numbers, and now demands the next number and the quality of supply and demand.”&lt;/p&gt;
&lt;h2 id="1-why-should-we-separate-signal-and-noise"&gt;1. Why should we separate signal and noise?
&lt;/h2&gt;&lt;p&gt;Today&amp;rsquo;s data is not read in only one direction. Numbers are good. However, stock prices and supply and demand are bad. Memory price comments are strong. However, spot trading takes a breather on some items. Hyperscaler CapEx continues to grow. However, long-term interest rates in the United States are also high.&lt;/p&gt;
&lt;p&gt;On days like this, looking at only one side is wrong. “It’s the world’s largest quarterly operating profit, so you must buy it” is also hasty, and “the cycle is over because it fell behind despite good performance” is also hasty.&lt;/p&gt;
&lt;p&gt;Therefore, this article divides the data into four layers.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;floor&lt;/th&gt;
 &lt;th&gt;Content&lt;/th&gt;
 &lt;th&gt;Role in Investment Decisions&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Official Performance&lt;/td&gt;
 &lt;td&gt;Samsung Electronics 2Q26 provisional sales and operating profit&lt;/td&gt;
 &lt;td&gt;Check profit level&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;industrial signals&lt;/td&gt;
 &lt;td&gt;DRAM·NAND price, AI server, hyperscaler CapEx&lt;/td&gt;
 &lt;td&gt;Check the quality of the thesis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;market noise&lt;/td&gt;
 &lt;td&gt;Sluggish spot trading, Morgan Stanley peak logic, ASML·TSMC high bar&lt;/td&gt;
 &lt;td&gt;Check short-term timing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Supply/Demand/Macro&lt;/td&gt;
 &lt;td&gt;Foreign selling, exchange rate, interest rate, Korean market breadth&lt;/td&gt;
 &lt;td&gt;Position Sizing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The key is to separate the profit direction and the stock price direction. Samsung Electronics’ main business profit direction has improved. However, the drivers of short-term stock prices were not profits, but supply and demand and the discount rate.&lt;/p&gt;
&lt;h2 id="2-officially-confirmed-values-sales-171-trillion-won-operating-profit-894-trillion-won"&gt;2. Officially confirmed values: Sales 171 trillion won, operating profit 89.4 trillion won
&lt;/h2&gt;&lt;p&gt;[Fact] Samsung Electronics announced its second quarter provisional results through its official newsroom on July 7, 2026. Consolidated sales are approximately 171 trillion won and operating profit is approximately 89.4 trillion won. According to Korean disclosure regulations, the company presented a median value rather than a range, and explained that the actual estimated range was KRW 170 trillion to KRW 172 trillion in sales and KRW 89.3 trillion to KRW 89.5 trillion in operating profit. Source: &lt;a class="link" href="https://news.samsung.com/global/samsung-electronics-announces-earnings-guidance-for-second-quarter-2026" target="_blank" rel="noopener"
 &gt;Samsung Electronics Official Newsroom&lt;/a&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q26 provisional figures&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Compared to previous quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Year-on-year comparison&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Sales&lt;/td&gt;
 &lt;td style="text-align: right"&gt;171.0 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;133.87 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;74.57 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+129.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;89.4 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;57.23 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.68 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+56.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,810.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This number is extremely unusual for a single quarter&amp;rsquo;s operating profit. The expression “world&amp;rsquo;s largest” may be subject to consideration depending on the comparison target and accounting standards, but it is clear that it is an overwhelmingly large number, at least compared to the normal quarterly profits of global mega-caps.&lt;/p&gt;
&lt;p&gt;What is important is not the operating profit itself of 89.4 trillion won, but the quality within it. The interpretation presented in the attached research is as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;reported operating profit&lt;/td&gt;
 &lt;td&gt;89.4 trillion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjustment of one-time provisions such as DS performance bonus&lt;/td&gt;
 &lt;td&gt;The logic that profits from main business should be considered higher&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core OP estimation&lt;/td&gt;
 &lt;td&gt;Possibility of over 100 trillion won excluding provisions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM ASP&lt;/td&gt;
 &lt;td&gt;+47% compared to previous quarter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND ASP&lt;/td&gt;
 &lt;td&gt;+66% compared to previous quarter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4&lt;/td&gt;
 &lt;td&gt;Maintain ramp-up expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;[Inference] Just looking at the reported OP of KRW 89.4 trillion is strong, but if DRAM and NAND prices rose significantly at the same time and there was an effect of provisioning, the strength of the main business may be stronger than the reported number. This point is the most important signal of Samsung Electronics&amp;rsquo; thesis.&lt;/p&gt;
&lt;h2 id="3-stock-price-and-supply-and-demand-foreigners-sold-after-good-numbers"&gt;3. Stock price and supply and demand: Foreigners sold after good numbers
&lt;/h2&gt;&lt;p&gt;Despite good performance, the stock price fell. The prices and supply and demand based on local DB are as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;date&lt;/th&gt;
 &lt;th style="text-align: right"&gt;closing price&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Daily fluctuations&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign net purchase estimates&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Estimated institutional net purchase&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Individual net purchase estimate&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-03&lt;/td&gt;
 &lt;td style="text-align: right"&gt;309,500 won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 387.2 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.3063 trillion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 918.4 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-06&lt;/td&gt;
 &lt;td style="text-align: right"&gt;318,000 won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.75%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-640.9 billion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.8 billion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+596.4 billion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-07&lt;/td&gt;
 &lt;td style="text-align: right"&gt;296,000 won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.92%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.8207 trillion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 547.5 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.3232 trillion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-08&lt;/td&gt;
 &lt;td style="text-align: right"&gt;277,500 won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 874.1 billion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+282.7 billion won&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+584.7 billion won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The source is the local Kiwoom-based &lt;code&gt;prices_daily&lt;/code&gt; and &lt;code&gt;investor_flow_daily&lt;/code&gt; tables. On July 8, the foreign ownership rate was recorded at 46.58%.&lt;/p&gt;
&lt;p&gt;The message of this table is clear. The market did not believe that Samsung Electronics&amp;rsquo; profits had weakened, but that it was a structure in which foreigners reduced their risk and individuals accepted it in stocks that had already risen. On July 7, foreigners and institutions sold together, and individuals bought significantly. Institutions received some on July 8, but foreign selling continued.&lt;/p&gt;
&lt;p&gt;Therefore, today’s judgment is “performance is good, but supply and demand are bad.” It is not a basis for selling, but a basis for freezing additional purchases.&lt;/p&gt;
&lt;h2 id="4-four-axes-of-samsung-electronics-thesis"&gt;4. Four axes of Samsung Electronics thesis
&lt;/h2&gt;&lt;p&gt;The argument for investing in Samsung Electronics can be organized into four axes.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;axis&lt;/th&gt;
 &lt;th&gt;Content&lt;/th&gt;
 &lt;th&gt;Current judgment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;AI server DRAM·HBM&lt;/td&gt;
 &lt;td&gt;Demand for high-bandwidth and high-capacity memory for AI servers and GPU/ASIC systems&lt;/td&gt;
 &lt;td&gt;Strengthening&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;eSSD·NAND mix&lt;/td&gt;
 &lt;td&gt;AI server storage and enterprise SSD demand&lt;/td&gt;
 &lt;td&gt;Strengthening&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyperscaler CapEx&lt;/td&gt;
 &lt;td&gt;Long-term investment in cloud and AI data centers&lt;/td&gt;
 &lt;td&gt;Strengthening&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry turnaround&lt;/td&gt;
 &lt;td&gt;Edge logic·packaging·customer recovery options&lt;/td&gt;
 &lt;td&gt;Not yet an option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Among these, the key ones are numbers 1 to 3. Foundry is a valuation upside option. For the stock price to hold up now, the foundry does not need to improve immediately. Investments in memory, eSSD, and hyperscaler alone explain the strength of profits. However, additional rerating requires reducing the foundry deficit and confirming HBM4 and HBM4E customers.&lt;/p&gt;
&lt;h2 id="5-signal-1-legacy-dram-and-nand-prices-are-stronger-than-expected"&gt;5. Signal 1: Legacy DRAM and NAND prices are stronger than expected
&lt;/h2&gt;&lt;p&gt;The most important bullish signal in the attached research is the memory price trend summarized in the comments of Meritz Securities Commissioner Kim Seon-woo. Since the full text of the original data was not directly compared, this article classifies it as “market comment-based interpretation.” Still, numbers and direction are important in interpreting Samsung Electronics&amp;rsquo; investment argument.&lt;/p&gt;
&lt;p&gt;The key point is that seasonal demand for the second half of the year began to arrive in July, and while B2B server demand was absorbing supply, B2C smartphone and PC customers also began to feel pressure to secure supplies.&lt;/p&gt;
&lt;h3 id="legacy-dram"&gt;Legacy DRAM
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Figures and Content&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Nanya 3Q26 DRAM price proposal&lt;/td&gt;
 &lt;td&gt;DDR4·LPDDR4 +50~90% QoQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Consumer DRAM price&lt;/td&gt;
 &lt;td&gt;Above $3/Gb, upward acceleration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;supply expansion&lt;/td&gt;
 &lt;td&gt;Judgment that meaningful supply expansion will be difficult before 2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cause&lt;/td&gt;
 &lt;td&gt;Decrease in investment between 2023 and 2025 after loss in 2023, lack of cleanroom&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="legacy-nand"&gt;Legacy NAND
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Figures and Content&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Windbond·MXIC SLC NAND proposal&lt;/td&gt;
 &lt;td&gt;3Q26 +70~100% QoQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;price range&lt;/td&gt;
 &lt;td&gt;$1.9~2.0/Gb&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Demand changes&lt;/td&gt;
 &lt;td&gt;SLC NAND orders increase due to MLC NAND shortage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Interpretation&lt;/td&gt;
 &lt;td&gt;Price pressure spreads not only to high-performance AI NAND but also to legacy NAND&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="supply-structure"&gt;Supply Structure
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Content&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM total supply growth rate in 2027&lt;/td&gt;
 &lt;td&gt;Expected 20~25%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server demand growth rate next year&lt;/td&gt;
 &lt;td&gt;50~70% expected&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bottleneck causes&lt;/td&gt;
 &lt;td&gt;Low-end/legacy shortage occurs due to concentration on high-end memory&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;[Inference] If this data is correct, Samsung Electronics&amp;rsquo; bull case is wider than SK Hynix&amp;rsquo;s HBM pure exposure. Samsung Electronics has a discount factor for being behind in HBM, but it also receives price increases for general DRAM, server DRAM, eSSD, NAND, and legacy products. In other words, “the frame of disparaging Samsung Electronics based solely on HBM” is weakened.&lt;/p&gt;
&lt;h2 id="6-signal-2-it-is-difficult-to-say-that-ai-capex-has-decreased-yet"&gt;6. Signal 2: It is difficult to say that AI CapEx has decreased yet
&lt;/h2&gt;&lt;p&gt;Hyperscaler investment also supports Samsung Electronics&amp;rsquo; thesis. According to numbers shared by Goldman Sachs charts, hyperscaler CapEx grows as follows:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Year&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Hyperscaler CapEx&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2024&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$227 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$406 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$740 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$996 billion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The forecast for 2027 is about 4.4 times that of 2024. This number is also in line with the bond and capital increase flows covered in &lt;a class="link" href="https://koreainvestinsights.com/ko/post/hyperscaler-financing-race-ai-capex-memory-bottleneck-2026-07-07/" &gt;Big Tech Financing Relay&lt;/a&gt;. Amazon, Alphabet, and Meta are even using corporate bonds and stock issuances to invest in AI infrastructure at a level that is difficult to cover with internal cash flow alone.&lt;/p&gt;
&lt;p&gt;What is important for Samsung Electronics is that “AI CapEx does not end at GPU.” Investing in an AI data center is not just about purchasing GPUs. Server DRAM, HBM, eSSD, NAND, substrate, power, network, and cooling will also increase.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;AI CapEx Items&lt;/th&gt;
 &lt;th&gt;Connection with Samsung Electronics&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;GPU·ASIC server&lt;/td&gt;
 &lt;td&gt;HBM, server DRAM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;data center storage device&lt;/td&gt;
 &lt;td&gt;eSSD, NAND&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High-density server proliferation&lt;/td&gt;
 &lt;td&gt;High capacity DDR5, LPDDR series, storage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Rack-level AI factory&lt;/td&gt;
 &lt;td&gt;Increased memory payload&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Long-term investment cycle&lt;/td&gt;
 &lt;td&gt;Strengthening visibility in 2027 rather than short-term performance in 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;[Inference] If hyperscaler CapEx increases by 2027, it is difficult to conclude that Samsung Electronics&amp;rsquo; memory profits will immediately decline from the peak in one quarter. The problem is not whether there is no demand, but how much the market has already reflected that demand in the price.&lt;/p&gt;
&lt;h2 id="7-signal-3-ai-server-supply-chain-is-not-yet-broken"&gt;7. Signal 3: AI server supply chain is not yet broken
&lt;/h2&gt;&lt;p&gt;Taiwan AI server supply chain data included in the attached research is also important.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Enterprise&lt;/th&gt;
 &lt;th&gt;Latest Signals&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Wistron&lt;/td&gt;
 &lt;td&gt;June sales +10.9% MoM, +53.8% YoY&lt;/td&gt;
 &lt;td&gt;GB300 NVL72, B300 shipment increases&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Wiwynn&lt;/td&gt;
 &lt;td&gt;June sales +32.5% MoM, +29.8% YoY&lt;/td&gt;
 &lt;td&gt;AMD Helios enters, prepares for mass shipment of 3Q ASIC AI servers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Unimicron&lt;/td&gt;
 &lt;td&gt;ABF·HDI substrate revenue record high&lt;/td&gt;
 &lt;td&gt;ABF expansion CapEx increase mentioned&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This data does not fit the claim that “demand for AI servers has declined.” If server OEMs, ODMs, and substrate companies still show strong shipments and sales, memory demand is unlikely to disappear easily.&lt;/p&gt;
&lt;p&gt;물론 이 데이터만으로 삼성전자 실적을 바로 산출할 수는 없다. But the direction is clear. AI 서버 공급망은 아직 멈추지 않았다. 오히려 GPU, ASIC, 서버, 기판, 메모리, storage가 함께 움직이는 구조다.&lt;/p&gt;
&lt;h2 id="8-소음-1-trendforce-spot-데이터는-약하지만-붕괴는-아니다"&gt;8. 소음 1: TrendForce spot 데이터는 약하지만 붕괴는 아니다
&lt;/h2&gt;&lt;p&gt;There are also opposing signs. TrendForce spot market comments were summarized as an atmosphere where trading was quiet and buyers were waiting.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;spot change&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DDR5 16Gb&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.50%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DDR5 eTT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.00%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DDR4 16Gb 3200&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.42%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DDR4 eTT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.51%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DDR4 8Gb 3200&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.88%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Looking at this table, it is excessive to say, “Memory prices have fallen.” 대부분은 상승 또는 보합이고, 일부 eTT 품목만 하락했다. 더 정확한 표현은 “거래는 한산하고, spot 시장은 가격 발견 중”이다.&lt;/p&gt;
&lt;p&gt;The important distinction is spot and contract.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Category&lt;/th&gt;
 &lt;th&gt;meaning&lt;/th&gt;
 &lt;th&gt;Relevance to Samsung Electronics thesis&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Spot&lt;/td&gt;
 &lt;td&gt;Short-term transaction prices in distribution and module markets&lt;/td&gt;
 &lt;td&gt;Quickly reflects sentiment and short-term inventory&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Contract&lt;/td&gt;
 &lt;td&gt;Large customer long-term contract price&lt;/td&gt;
 &lt;td&gt;Server DRAM, HBM, eSSD more important to profits&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;삼성전자 thesis의 핵심은 spot이 아니다. 핵심은 contract 가격, server DRAM, HBM, eSSD 주문, hyperscaler 수요다. 진짜 위험 신호는 spot 거래 부진 자체가 아니라 다음 네 가지가 함께 나오는 경우다.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Contract price cut&lt;/li&gt;
&lt;li&gt;Slowdown in server DRAM orders&lt;/li&gt;
&lt;li&gt;HBM·eSSD order slowdown&lt;/li&gt;
&lt;li&gt;주요 고객의 inventory correction 코멘트&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Currently, all four have not been confirmed. Therefore, it is correct to view TrendForce data not as bear confirmation but as a signal of anticipation digestion and price discovery.&lt;/p&gt;
&lt;h2 id="9-noise-2-morgan-stanley-peak-logic-is-only-partially-correct"&gt;9. Noise 2: Morgan Stanley peak logic is only partially correct
&lt;/h2&gt;&lt;p&gt;Morgan Stanley-type peak thesis should not be ignored either. The key point is that the speed of improvement in memory indicators is approaching its peak, and inventory normalization and price increases may slow down. AI trade 안에서는 반도체보다 Alphabet, Amazon 같은 AI cloud 쪽을 선호한다는 관점도 제시됐다.&lt;/p&gt;
&lt;p&gt;This claim is partly correct. This is especially important for short-term stock prices.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Morgan Stanley-style warning&lt;/th&gt;
 &lt;th&gt;Areas to acknowledge&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Improvement speed peak possibility&lt;/td&gt;
 &lt;td&gt;Stock prices are more sensitive to slope than level.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price increases slowing&lt;/td&gt;
 &lt;td&gt;3Q 이후 상승률이 낮아지면 멀티플 압박 가능&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Semiconductor overcrowding position&lt;/td&gt;
 &lt;td&gt;Sell-the-news is possible even with good performance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI cloud preferred&lt;/td&gt;
 &lt;td&gt;Possibility of rotation towards consumers rather than semiconductors&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;However, extending this to the logic of structural selling is weak. There is too much opposing data.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Opposite logic&lt;/th&gt;
 &lt;th&gt;Evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Continued expansion of AI investment&lt;/td&gt;
 &lt;td&gt;Hyperscaler CapEx expected to increase by 2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Prolonged supply shortage&lt;/td&gt;
 &lt;td&gt;Legacy shortage intensifies due to transition to cleanroom and high value-added products&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server demand growth&lt;/td&gt;
 &lt;td&gt;Server demand expected to increase by 50~70% next year&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Domestic securities firms continue to buy&lt;/td&gt;
 &lt;td&gt;Target prices maintained or raised, including KB 600,000 won, Daol 585,000 won, NH 530,000 won, and Hana/iM 480,000 won&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;[Inference] The Morgan Stanley peak thesis is valid as a warning, “Don’t chase now.” However, the basis for “discarding Samsung Electronics’ thesis” is insufficient.&lt;/p&gt;
&lt;h2 id="10-noise-3-asml-and-tsmc-are-high-bar-events"&gt;10. Noise 3: ASML and TSMC are high bar events
&lt;/h2&gt;&lt;p&gt;The next gateway is ASML and TSMC. ASML&amp;rsquo;s official financial calendar displays Q2 2026 financial results on July 15, 2026. TSMC&amp;rsquo;s official IR page announces the 2Q26 earnings conference on July 16, 2026. Source: &lt;a class="link" href="https://www.asml.com/investors/financial-calendar" target="_blank" rel="noopener"
 &gt;ASML financial calendar&lt;/a&gt;, &lt;a class="link" href="https://investor.tsmc.com/english/quarterly-results/2026/q2" target="_blank" rel="noopener"
 &gt;TSMC 2Q26 quarterly results&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The problem is that even if good performance is achieved, it is impossible to know whether the stock price will rise. Samsung Electronics&amp;rsquo; 2Q26 provisional results are already a preview. If the high bar is high, even good numbers seem lacking.&lt;/p&gt;
&lt;p&gt;Here&amp;rsquo;s what to look for in ASML and TSMC:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Event&lt;/th&gt;
 &lt;th&gt;checkpoint&lt;/th&gt;
 &lt;th&gt;Connection with Samsung Electronics&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;ASML 2Q26&lt;/td&gt;
 &lt;td&gt;bookings, EUV·High-NA demand, Chinese regulatory comments&lt;/td&gt;
 &lt;td&gt;Confirmation of semiconductor capex expectations and equipment demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;TSMC 2Q26&lt;/td&gt;
 &lt;td&gt;AI/HPC revenue, CoWoS·advanced packaging, CapEx guidance&lt;/td&gt;
 &lt;td&gt;Identifying AI accelerator and HBM packaging bottlenecks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Market reaction&lt;/td&gt;
 &lt;td&gt;SOXX, NVDA, MU, TSM, ASML stock price reaction&lt;/td&gt;
 &lt;td&gt;Check if it is sold or relieved with good performance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;As Mizuho commented, the fact that Samsung Electronics&amp;rsquo; stock price fell despite its good preliminary results shows the possibility of a global IT momentum shift. So ASML and TSMC are events that see “how the market digests good numbers” rather than “confirming the numbers.”&lt;/p&gt;
&lt;h2 id="11-noise-4-press-multiple-for-interest-rates-and-exchange-rates"&gt;11. Noise 4: Press multiple for interest rates and exchange rates
&lt;/h2&gt;&lt;p&gt;Even if Samsung Electronics&amp;rsquo; thesis is good, the stock price may be depressed if the discount rate increases. The interest rate environment based on the attached research is as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;level&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;US 10-year note&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 4.58%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;US 30-year note&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 5.07%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Japanese 10-year note&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 2.88%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Japanese 40-year note&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 4.01%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;These numbers do not undermine the demand thesis. However, PER can be lowered. A typical section where stock prices fall despite strong profits is “the section where EPS rises but the multiple falls faster.”&lt;/p&gt;
&lt;p&gt;The same goes for exchange rates. A weak won may be beneficial to Samsung Electronics&amp;rsquo; performance. However, from a foreigner&amp;rsquo;s perspective, it increases foreign exchange losses, foreign exchange hedging costs, and the risk of Korea&amp;rsquo;s share. The foreign ownership rate of 46.58% and continuous selling on July 8 are signals that this part needs to be checked.&lt;/p&gt;
&lt;h2 id="12-noise-5-foreign-outflow-and-leverage-piping"&gt;12. Noise 5: Foreign outflow and leverage piping
&lt;/h2&gt;&lt;p&gt;The attached research includes comments on foreign outflow and leverage piping based on Bloomberg and Telegram. The figures presented here, such as foreign outflow in the Korean market since the beginning of the year -$100.16 billion, foreign securities outflow in March -$36.55 billion, and USD/KRW 1,554 won, need to be compared with official statistics by collection range. Therefore, in this article, only directional signals are used.&lt;/p&gt;
&lt;p&gt;Still, the message is important.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Observation&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreigners reduced Korea&lt;/td&gt;
 &lt;td&gt;Rather than avoiding Korea, it may be a memory peak-out bet.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Domestic institutions and individuals absorbed the rebound&lt;/td&gt;
 &lt;td&gt;It is a domestic liquidity absorption market rather than a foreign-led rally.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Single stock leverage/short gamma structure&lt;/td&gt;
 &lt;td&gt;Forced selling and hedging flows during declines can exaggerate prices.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Exchange rate rise&lt;/td&gt;
 &lt;td&gt;It is a barrier to additional purchases by foreigners.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;[Inference] The short-term plunge is more likely to be the result of memory peak-out betting, liquidation of leverage positions, and reduction of foreign risks, rather than the collapse of Samsung Electronics&amp;rsquo; main business. However, we cannot conclude that this is not a structural risk. We need to check whether foreign spot sales stop.&lt;/p&gt;
&lt;h2 id="13-korean-market-performance-is-also-not-good"&gt;13. Korean market performance is also not good
&lt;/h2&gt;&lt;p&gt;This is not just a problem for Samsung Electronics. As of 06:10 on July 8, the local KR integrated screener classified the Korean market as NEUTRAL.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;figures&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Proportion of stocks above the 50-day line&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Proportion of stocks above the 200-day line&lt;/td&gt;
 &lt;td style="text-align: right"&gt;21.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Passed items&lt;/td&gt;
 &lt;td style="text-align: right"&gt;17&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;New Passed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Breakaway&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Breakout&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Newly passed stocks were APR, Handsome, GS Retail, Hana Financial Group, and KT&amp;amp;G. Stocks that left included TSE, YG-1, and Samsung Electronics.&lt;/p&gt;
&lt;p&gt;13 passed the quality compounder screen, and at the top were Gigabis, SK Hynix, VM, AP, and PSK. Only three companies, APR, Seegene, and Kolmar Korea, passed the smart money + quality condition.&lt;/p&gt;
&lt;p&gt;The meaning of this data is simple. Market breadth is weak. Even if Samsung Electronics&amp;rsquo; thesis is good, it is not widely accepted by the entire Korean market. In this environment, additional purchases should not be made by reporting only profit data. We need to see whether supply and demand and market width return.&lt;/p&gt;
&lt;h2 id="14-then-what-should-i-do"&gt;14. Then what should I do?
&lt;/h2&gt;&lt;p&gt;The current conclusion is “strong holding, freeze on additional purchases.”&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;action&lt;/th&gt;
 &lt;th&gt;Judgment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;holding&lt;/td&gt;
 &lt;td&gt;possible. The thesis is intact.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Additional Purchase&lt;/td&gt;
 &lt;td&gt;Still waiting. Need to confirm supply/demand, ASML, TSMC, IR&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sell ​​&lt;/td&gt;
 &lt;td&gt;It&amp;rsquo;s not just data now. The profit direction actually improved.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Short term trading&lt;/td&gt;
 &lt;td&gt;After confirming the slowdown in foreign selling and the global semiconductor response from July 15th to 16th&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;There are four additional purchase conditions:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Foreign spot sales of Samsung Electronics are slowing down, and programmatic selling pressure is decreasing.&lt;/li&gt;
&lt;li&gt;ASML bookings and TSMC AI/HPC·CoWoS·CapEx comments do not indicate a slowdown in AI investment.&lt;/li&gt;
&lt;li&gt;DS core profit, DRAM·NAND contract price, HBM4·HBM4E customer progress, and eSSD orders are confirmed in Samsung Electronics IR.&lt;/li&gt;
&lt;li&gt;The sluggish spot market does not spread to contract price cuts or a slowdown in server DRAM orders.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The conditions for invalidation are as follows:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Invalidation Conditions&lt;/th&gt;
 &lt;th&gt;meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Contract DRAM/NAND price decline&lt;/td&gt;
 &lt;td&gt;price thesis undermined&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Server DRAM·HBM·eSSD order slowdown&lt;/td&gt;
 &lt;td&gt;AI memory demand thesis undermined&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ASML·TSMC strongly talks about CapEx slowdown&lt;/td&gt;
 &lt;td&gt;Hyperscaler investment thesis weakened&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign selling and won weakening intensify simultaneously&lt;/td&gt;
 &lt;td&gt;Pressure to reduce Korea&amp;rsquo;s proportion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Market breadth further worsens&lt;/td&gt;
 &lt;td&gt;Even if Samsung Electronics is good, supply and demand absorption is weakened&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="15-conclusion"&gt;15. Conclusion
&lt;/h2&gt;&lt;p&gt;Samsung Electronics&amp;rsquo; 2Q26 provisional performance is not a number that weakens the thesis. Rather, it is the opposite. Operating profit of KRW 89.4 trillion, DRAM ASP +47%, NAND ASP +66%, soaring legacy memory prices, hyperscaler CapEx continuing until 2027, and AI server supply chain sales all strengthen the case for investing in Samsung Electronics.&lt;/p&gt;
&lt;p&gt;However, today&amp;rsquo;s stock prices and supply and demand have damaged the timing. Foreigners sold, and individuals received. Interest rates are high, exchange rates are a burden, and high bar events such as ASML and TSMC remain. Market width is also weak.&lt;/p&gt;
&lt;p&gt;So the conclusion is clear.&lt;/p&gt;
&lt;p&gt;Samsung Electronics has not become worse. However, today’s market called for “the next good number and supply and demand to believe in that number” rather than “a good number.” Now is not the time to abandon the thesis, but rather to protect the position, stop making additional purchases, and check for the next gateway.&lt;/p&gt;
&lt;h2 id="basis-classification"&gt;Basis classification
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Category&lt;/th&gt;
 &lt;th&gt;Content&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fact&lt;/td&gt;
 &lt;td&gt;Samsung Electronics&amp;rsquo; official 2Q26 provisional sales of 171 trillion won and operating profit of 89.4 trillion won. Comparative figures for 1Q26 and 2Q25. ASML July 15th, TSMC July 16th earnings schedule. Samsung Electronics price, supply and demand, and foreign ownership rate in local DB.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Inference&lt;/td&gt;
 &lt;td&gt;Samsung Electronics&amp;rsquo; thesis was strengthened, but the short-term timing was judged to be damaged. TrendForce spot data is interpreted as price discovery, not bear confirmation. Morgan Stanley&amp;rsquo;s peak thesis is a catch-up warning and not a basis for structural selling.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Speculation&lt;/td&gt;
 &lt;td&gt;A scenario in which hyperscaler CapEx increases to approximately $1 trillion by 2027 and can support Samsung Electronics&amp;rsquo; memory demand for a longer period of time.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Blocked&lt;/td&gt;
 &lt;td&gt;Comparison of Bloomberg foreign outflow figures with official statistics, confirmation of the full text report by Meritz·TrendForce, 2Q26 actual profit by Samsung Electronics DS division, and volume and margin by HBM4·HBM4E customer.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="evidence-ledger"&gt;Evidence Ledger
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Source&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics 2Q26 provisional performance&lt;/td&gt;
 &lt;td&gt;&lt;a class="link" href="https://news.samsung.com/global/samsung-electronics-announces-earnings-guidance-for-second-quarter-2026" target="_blank" rel="noopener"
 &gt;Samsung Electronics Official Newsroom&lt;/a&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ASML 2Q26 Schedule&lt;/td&gt;
 &lt;td&gt;&lt;a class="link" href="https://www.asml.com/investors/financial-calendar" target="_blank" rel="noopener"
 &gt;ASML financial calendar&lt;/a&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;TSMC 2Q26 Schedule&lt;/td&gt;
 &lt;td&gt;&lt;a class="link" href="https://investor.tsmc.com/english/quarterly-results/2026/q2" target="_blank" rel="noopener"
 &gt;TSMC 2Q26 quarterly results&lt;/a&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics price/supply&lt;/td&gt;
 &lt;td&gt;Local Research DB: &lt;code&gt;prices_daily&lt;/code&gt;, &lt;code&gt;investor_flow_daily&lt;/code&gt;, &lt;code&gt;foreign_ownership_daily&lt;/code&gt;, as of 2026-07-08&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory price·AI server·foreign outflow comments&lt;/td&gt;
 &lt;td&gt;Attached research packet. Uncollated items in the full text are separated from the main text as market comments or directional signals&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Related prior analysis&lt;/td&gt;
 &lt;td&gt;Korea Invest Insights&amp;rsquo; Samsung Electronics 2Q26 preview, Micron FY3Q26, Big Tech earnings call, Big Tech financing, NVIDIA similar quarter analysis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;</description></item><item><title>Which NVIDIA Quarter Does Samsung's Earnings Selloff Resemble?</title><link>https://koreainvestinsights.com/post/samsung-earnings-selloff-nvidia-q4fy26-rebound-trigger-2026-07-08/</link><pubDate>Wed, 08 Jul 2026 14:30:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/samsung-earnings-selloff-nvidia-q4fy26-rebound-trigger-2026-07-08/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-2q26-preview-micron-surprise-erased-core-op-hbm-2026-06-29/" &gt;Samsung 2Q26 preview&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 earnings&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/nvidia-earnings-elasticity-hbm-cycle-samsung-hynix-2026-06-28/" &gt;NVIDIA elasticity and the Korea HBM cycle&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/" &gt;late-July Big Tech calls and the memory thesis&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/hyperscaler-financing-race-ai-capex-memory-bottleneck-2026-07-07/" &gt;hyperscaler financing and memory bottlenecks&lt;/a&gt;. Related hubs: &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;Samsung Electronics&amp;rsquo; 2Q26 preliminary earnings were strong on the surface. Revenue was about KRW 171 trillion and operating profit was about KRW 89.4 trillion. Sales rose 27.74% quarter over quarter and operating profit rose 56.21%. Operating profit was up 1,810.26% year over year. These figures were disclosed by Samsung Electronics.&lt;/p&gt;
&lt;p&gt;Yet the stock fell 6.92% on July 7 and another 6.25% on July 8. The two-day decline was 12.7%.&lt;/p&gt;
&lt;p&gt;That is not the same as the famous NVIDIA August 2024 sell-on reaction. NVIDIA fell 6.4% on the first reaction day in August 2024, but rebounded 1.5% the next day. Samsung did not bounce. It fell again.&lt;/p&gt;
&lt;p&gt;Across 17 NVIDIA post-earnings reactions, the closest analogue is NVIDIA&amp;rsquo;s Q4 FY26 report on February 25, 2026. NVIDIA fell 5.5% on the first reaction day and 4.2% the next day, for a two-day decline of 9.4%. The reaction-vector distance from Samsung&amp;rsquo;s path is 2.55, the smallest in the sample.&lt;/p&gt;
&lt;p&gt;The key point is not that a strong earnings print is useless. The point is that a strong print alone was not enough once the market started asking whether AI infrastructure growth was slowing. NVIDIA&amp;rsquo;s large post-print drawdowns eventually recovered, but the recovery came only after a new catalyst directly killed the fear that had caused the drawdown. For Samsung, the equivalent catalyst is late-July Big Tech CapEx guidance.&lt;/p&gt;
&lt;p&gt;The practical conclusion is Wait. This is not a mechanical dip-buy setup yet. The next gates are late-July hyperscaler CapEx commentary, DRAM/NAND price momentum, and Samsung&amp;rsquo;s clean 3Q operating-profit path.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="1-what-was-strange-about-samsungs-earnings-reaction"&gt;1. What was strange about Samsung&amp;rsquo;s earnings reaction
&lt;/h2&gt;&lt;p&gt;The numbers were strong.&lt;/p&gt;
&lt;p&gt;Samsung Electronics announced 2Q26 preliminary earnings on July 7, 2026. The company disclosed the following consolidated figures.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q26 preliminary&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q25&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 171T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 133.87T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 74.57T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 89.4T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 57.23T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.68T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27.74% QoQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+129.31% YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+56.21% QoQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1,810.26% YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung also disclosed estimate ranges of KRW 170T to KRW 172T for sales and KRW 89.3T to KRW 89.5T for operating profit. Korean disclosure rules required the company to publish a single median figure.&lt;/p&gt;
&lt;p&gt;The stock reaction was the problem. Samsung closed at KRW 318,000 on July 6, KRW 296,000 on July 7, and KRW 277,500 on July 8. Based on yfinance daily close data:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Date&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Close&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Daily return&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-06&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 318,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.75%&lt;/td&gt;
 &lt;td&gt;pre-announcement base&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-07&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 296,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.92%&lt;/td&gt;
 &lt;td&gt;announcement-day selloff&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-08&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 277,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.25%&lt;/td&gt;
 &lt;td&gt;no bounce&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Two-day cumulative&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.74%&lt;/td&gt;
 &lt;td&gt;post-print path&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Compared with the June 18 high of KRW 362,500, the July 8 close was already 23.45% lower. Samsung was not falling from a calm setup. It was already volatile before the earnings print, and the good number failed to stop the selling.&lt;/p&gt;
&lt;p&gt;The key is that the market was no longer asking only how much Samsung earned. It was asking whether the pace of earnings improvement could keep accelerating.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="2-why-compare-samsung-with-nvidia"&gt;2. Why compare Samsung with NVIDIA
&lt;/h2&gt;&lt;p&gt;Samsung is a memory company. NVIDIA is a GPU platform company. Their businesses are different. But both have been core price signals for the AI infrastructure cycle.&lt;/p&gt;
&lt;p&gt;NVIDIA moved first when the AI infrastructure boom began. As bottlenecks moved downstream into HBM, DRAM, eSSD, substrates, and power infrastructure, Samsung and SK hynix began facing the same market question: is the next leg still accelerating?&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Did earnings improve?&lt;/td&gt;
 &lt;td&gt;confirmed result&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Did earnings beat the bar?&lt;/td&gt;
 &lt;td&gt;surprise versus expectations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Will the next quarter be better?&lt;/td&gt;
 &lt;td&gt;guidance and pricing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is the growth rate still rising?&lt;/td&gt;
 &lt;td&gt;the part stocks care about most&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Will hyperscalers keep spending?&lt;/td&gt;
 &lt;td&gt;the final payer for AI memory demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Samsung&amp;rsquo;s two-day selloff came from that last layer. The print was strong, but expectations were already high. Some investors had expected operating profit above KRW 90T. At the same time, the market was worried about possible hyperscaler CapEx moderation. Korean single-stock leverage products, futures, and ETF flows amplified the move.&lt;/p&gt;
&lt;p&gt;This analysis therefore does not compare Samsung&amp;rsquo;s business model with NVIDIA&amp;rsquo;s business model. It compares post-earnings price paths and the fear behind those paths.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="3-timing-adjustment-nvidia-reports-after-the-close"&gt;3. Timing adjustment: NVIDIA reports after the close
&lt;/h2&gt;&lt;p&gt;NVIDIA usually reports after the US market close. The earnings information is not reflected in the announcement-day close. The real reaction appears in the next regular trading session.&lt;/p&gt;
&lt;p&gt;Samsung&amp;rsquo;s guidance was reflected in the Korean session on the announcement day. So the timing has to be aligned as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Reporting style&lt;/th&gt;
 &lt;th&gt;First trading session with full reaction&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;reflected in the Korean session&lt;/td&gt;
 &lt;td&gt;announcement day&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA&lt;/td&gt;
 &lt;td&gt;after-market release&lt;/td&gt;
 &lt;td&gt;next trading day&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;In this note, NVIDIA&amp;rsquo;s reaction D0 means the first trading day after the after-market release. NVIDIA&amp;rsquo;s D+1 means the following trading day. This aligns NVIDIA&amp;rsquo;s D0 with Samsung&amp;rsquo;s July 7 and NVIDIA&amp;rsquo;s D+1 with Samsung&amp;rsquo;s July 8.&lt;/p&gt;
&lt;p&gt;Without this adjustment, the comparison would be wrong.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="4-nvidias-17-quarter-post-earnings-reaction-table"&gt;4. NVIDIA&amp;rsquo;s 17-quarter post-earnings reaction table
&lt;/h2&gt;&lt;p&gt;The table below combines NVIDIA quarterly revenue, growth rates, and post-earnings stock reactions. Revenue and report timing are sourced from NVIDIA IR and SEC 8-K releases. Stock returns are calculated from yfinance daily close data.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Quarter&lt;/th&gt;
 &lt;th&gt;Report date&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;QoQ&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YoY&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Reaction D0&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D+1&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Two-day&lt;/th&gt;
 &lt;th style="text-align: right"&gt;T+20&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 FY23&lt;/td&gt;
 &lt;td&gt;2022-05-25&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$8.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;n/a&lt;/td&gt;
 &lt;td style="text-align: right"&gt;n/a&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 FY23&lt;/td&gt;
 &lt;td&gt;2022-08-24&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$6.7B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-19.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;n/a&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-27.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 FY23&lt;/td&gt;
 &lt;td&gt;2022-11-16&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5.9B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;n/a&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY23&lt;/td&gt;
 &lt;td&gt;2023-02-22&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$6.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;n/a&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 FY24&lt;/td&gt;
 &lt;td&gt;2023-05-24&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$7.2B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+18.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+27.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+38.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 FY24&lt;/td&gt;
 &lt;td&gt;2023-08-23&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$13.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+87.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+101.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 FY24&lt;/td&gt;
 &lt;td&gt;2023-11-21&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$18.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+34.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+206.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY24&lt;/td&gt;
 &lt;td&gt;2024-02-21&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$22.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+22.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+262.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+33.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 FY25&lt;/td&gt;
 &lt;td&gt;2024-05-22&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$26.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+261.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+33.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 FY25&lt;/td&gt;
 &lt;td&gt;2024-08-28&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$30.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+122.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 FY25&lt;/td&gt;
 &lt;td&gt;2024-11-20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$35.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+93.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-10.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY25&lt;/td&gt;
 &lt;td&gt;2025-02-26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$39.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+77.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 FY26&lt;/td&gt;
 &lt;td&gt;2025-05-28&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$44.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+69.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 FY26&lt;/td&gt;
 &lt;td&gt;2025-08-27&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$46.7B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q3 FY26&lt;/td&gt;
 &lt;td&gt;2025-11-19&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$57.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+22.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+62.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY26&lt;/td&gt;
 &lt;td&gt;2026-02-25&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$68.1B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+19.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+73.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 FY27&lt;/td&gt;
 &lt;td&gt;2026-05-20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$81.6B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+19.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+85.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The key message is clear. Strong results did not always lift the stock. As the AI cycle matured, the market became more sensitive to the change in growth rate than to the level of revenue.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="5-the-closest-analogue-is-nvidia-q4-fy26"&gt;5. The closest analogue is NVIDIA Q4 FY26
&lt;/h2&gt;&lt;p&gt;Samsung&amp;rsquo;s path was:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company/event&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D0&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D+1&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Two-day&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26 preliminary earnings&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.92%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.74%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;We compare this two-day vector with all 17 NVIDIA quarters.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;NVIDIA quarter&lt;/th&gt;
 &lt;th&gt;Report date&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D0&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D+1&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Two-day&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Distance from Samsung&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Q4 FY26&lt;/td&gt;
 &lt;td&gt;2026-02-25&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.46%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.16%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-9.39%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.55&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;Q3 FY24&lt;/td&gt;
 &lt;td&gt;2023-11-21&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.46%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.93%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.34%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.21&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Q3 FY26&lt;/td&gt;
 &lt;td&gt;2025-11-19&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.97%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.10%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.48&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;Q1 FY27&lt;/td&gt;
 &lt;td&gt;2026-05-20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.77%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.90%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.64%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.74&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;Q2 FY26&lt;/td&gt;
 &lt;td&gt;2025-08-27&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.79%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.32%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.09%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.79&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Q4 FY26 is the clear match. It is not just the numbers. The drivers were similar.&lt;/p&gt;
&lt;p&gt;NVIDIA&amp;rsquo;s Q4 FY26 revenue was $68.1B, up 73% year over year and 20% quarter over quarter. The next-quarter revenue guide was $78B. The print was strong. But the stock fell 5.5% on the first reaction day and 4.2% on the next day.&lt;/p&gt;
&lt;p&gt;The market was asking whether the next step could still get better. Customer concentration, hyperscaler spending durability, AI CapEx peak risk, and competition were all in focus.&lt;/p&gt;
&lt;p&gt;Samsung faced the same structure. KRW 89.4T of operating profit was strong, but investors had already moved the bar higher. The market wanted proof that the growth rate was still accelerating, not only proof that the current profit level was high.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="6-why-the-august-2024-nvidia-sell-on-is-the-wrong-template"&gt;6. Why the August 2024 NVIDIA sell-on is the wrong template
&lt;/h2&gt;&lt;p&gt;The intuitive comparison is NVIDIA Q2 FY25 in August 2024. NVIDIA fell 6.4% on the first reaction day. That resembles Samsung&amp;rsquo;s July 7 decline.&lt;/p&gt;
&lt;p&gt;But the next day changes the conclusion.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Case&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D0&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D+1&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA Q2 FY25, August 2024&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.5%&lt;/td&gt;
 &lt;td&gt;one-day selloff, then bounce&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA Q4 FY25, February 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.0%&lt;/td&gt;
 &lt;td&gt;one-day selloff, then strong bounce&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA Q4 FY26, February 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.2%&lt;/td&gt;
 &lt;td&gt;two-day selloff&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.92%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.25%&lt;/td&gt;
 &lt;td&gt;two-day selloff&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The August 2024 and February 2025 NVIDIA reactions were dip-and-bounce. Samsung was dip-and-continue. That distinction matters. A one-day sell-on can be a dip-buy setup. A two-day selloff can mean the market has not finished digesting the fear.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="7-did-nvidia-recover-immediately-after-large-post-print-selloffs"&gt;7. Did NVIDIA recover immediately after large post-print selloffs?
&lt;/h2&gt;&lt;p&gt;No.&lt;/p&gt;
&lt;p&gt;There were three NVIDIA quarters with a first-reaction-day decline worse than 5%: Q2 FY25, Q4 FY25, and Q4 FY26. All three were still below the pre-announcement price after 20 trading days.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;NVIDIA quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;D0&lt;/th&gt;
 &lt;th style="text-align: right"&gt;T+5&lt;/th&gt;
 &lt;th style="text-align: right"&gt;T+10&lt;/th&gt;
 &lt;th style="text-align: right"&gt;T+20&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 FY25, August 2024&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-14.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY25, February 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-10.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY26, February 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;All three fell further by T+5. All three failed to recover the pre-announcement price by T+20.&lt;/p&gt;
&lt;p&gt;The sample is small, so this is not a deterministic rule. But it is enough to reject the simple idea that strong earnings plus a large drop automatically equals an immediate buying opportunity.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="8-the-real-low-and-recovery-took-weeks"&gt;8. The real low and recovery took weeks
&lt;/h2&gt;&lt;p&gt;The second attached analysis goes further. It asks when NVIDIA actually bottomed and when it recovered.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Anchor&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Pre-print base&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Low&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Max drawdown&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Time to low&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Breakeven&lt;/th&gt;
 &lt;th style="text-align: right"&gt;New high&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Low-to-later-high&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 FY25, August 2024&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$125.55&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$102.78 on 2024-09-06&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-18.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6 sessions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2024-10-07, 5.5 weeks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2024-10-14, 6.5 weeks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+45.3% to 2025-01-06&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY25, February 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$131.24&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$94.29 around 2025-04-04 to 2025-04-08&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-28.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;27 sessions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-05-14, 11 weeks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-06-25, 16 weeks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+119.6% to 2025-10-29&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY26, February 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$195.55&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$165.17 on 2026-03-30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-15.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23 sessions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026-04-14, 6.5 weeks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026-04-24, 8 weeks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+42.7% to 2026-05-14&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The pattern is important.&lt;/p&gt;
&lt;p&gt;First, the low did not arrive immediately. Even the fastest case took six trading sessions. The slower cases took more than a month.&lt;/p&gt;
&lt;p&gt;Second, breakeven took 5.5 to 11 weeks.&lt;/p&gt;
&lt;p&gt;Third, the rebound was eventually large. But it did not come from cheapness alone. It came after a catalyst removed the fear behind the selloff.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="9-the-rebound-trigger-had-to-kill-the-exact-fear"&gt;9. The rebound trigger had to kill the exact fear
&lt;/h2&gt;&lt;h3 id="9-1-august-2024-the-fed-addressed-macro-fear"&gt;9-1. August 2024: the Fed addressed macro fear
&lt;/h3&gt;&lt;p&gt;The August 2024 drawdown was about macro fear. Weak labor-market data raised recession concerns. Regulatory noise also hurt sentiment.&lt;/p&gt;
&lt;p&gt;The rebound trigger was the Fed&amp;rsquo;s 50bp cut on September 18, 2024. The fear was rates and recession. The Fed directly addressed that fear.&lt;/p&gt;
&lt;h3 id="9-2-february-2025-tariff-fear-required-a-policy-reversal"&gt;9-2. February 2025: tariff fear required a policy reversal
&lt;/h3&gt;&lt;p&gt;The February 2025 post-print decline later deepened because of tariff shock. NVIDIA moved toward a 12-month low in early April.&lt;/p&gt;
&lt;p&gt;The rebound trigger was the 90-day tariff pause. A policy shock created the fear; a policy reversal removed it.&lt;/p&gt;
&lt;h3 id="9-3-february-2026-gtc-addressed-ai-capex-peak-fear"&gt;9-3. February 2026: GTC addressed AI CapEx peak fear
&lt;/h3&gt;&lt;p&gt;This is the most relevant anchor for Samsung.&lt;/p&gt;
&lt;p&gt;NVIDIA&amp;rsquo;s Q4 FY26 print was strong, but the market worried that AI investment might peak in 2026. Hyperscaler concentration, AI return-on-investment questions, and customer spending durability all mattered.&lt;/p&gt;
&lt;p&gt;The rebound catalyst was GTC 2026. Jensen Huang said NVIDIA had visibility into at least $1T of demand or orders through 2027, compared with the prior year&amp;rsquo;s $500B figure through 2026. Data Center Dynamics and Constellation Research reported the comments.&lt;/p&gt;
&lt;p&gt;This directly addressed the fear. The market worried that AI CapEx was peaking; NVIDIA answered that demand through 2027 was much larger.&lt;/p&gt;
&lt;p&gt;Even then, the bottom came after GTC, not on GTC day. The stock bottomed on March 30, about two weeks later.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="10-applying-the-lesson-to-samsung"&gt;10. Applying the lesson to Samsung
&lt;/h2&gt;&lt;p&gt;Samsung&amp;rsquo;s selloff fear has two parts.&lt;/p&gt;
&lt;p&gt;First, memory pricing and HBM demand may have been priced too quickly.&lt;/p&gt;
&lt;p&gt;Second, if hyperscalers slow AI infrastructure spending, memory pricing and volume could both weaken.&lt;/p&gt;
&lt;p&gt;Therefore, the rebound does not need another reminder that 2Q earnings were good. The market already knows that. The rebound needs evidence that the fear is wrong.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Samsung selloff fear&lt;/th&gt;
 &lt;th&gt;Required catalyst&lt;/th&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Big Tech AI CapEx slowdown&lt;/td&gt;
 &lt;td&gt;Alphabet, Microsoft, Meta, and Amazon maintaining or raising 2026 to 2027 CapEx&lt;/td&gt;
 &lt;td&gt;late-July 2026 earnings calls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory price-growth slowdown&lt;/td&gt;
 &lt;td&gt;DRAM/NAND spot and contract-price momentum staying firm&lt;/td&gt;
 &lt;td&gt;weekly checks and 3Q negotiations&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM catch-up doubt&lt;/td&gt;
 &lt;td&gt;HBM4/HBM4E qualification, production shipments, NVIDIA mix&lt;/td&gt;
 &lt;td&gt;2H26 IR and earnings calls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q one-off cost confusion&lt;/td&gt;
 &lt;td&gt;clean 3Q operating-profit path after bonus/provision adjustments&lt;/td&gt;
 &lt;td&gt;3Q preview and 3Q earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;flow-driven liquidation&lt;/td&gt;
 &lt;td&gt;foreign and program-selling slowdown, lower leverage-product volatility&lt;/td&gt;
 &lt;td&gt;daily flow checks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The most important gate is late-July Big Tech earnings. If Alphabet, Microsoft, Meta, and Amazon keep or raise AI infrastructure spending plans, Samsung&amp;rsquo;s selloff can be reclassified as a flow correction rather than demand damage.&lt;/p&gt;
&lt;p&gt;If two or more hyperscalers talk about slowing the pace of AI infrastructure investment, the NVIDIA Q4 FY26 template becomes more relevant and the correction can last longer.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="11-why-the-fed-put-is-weaker-this-time"&gt;11. Why the Fed put is weaker this time
&lt;/h2&gt;&lt;p&gt;Two of NVIDIA&amp;rsquo;s prior recoveries had macro or policy support.&lt;/p&gt;
&lt;p&gt;In August 2024, the Fed cut rates by 50bp.&lt;/p&gt;
&lt;p&gt;In early 2025, a tariff pause reversed a policy shock.&lt;/p&gt;
&lt;p&gt;In February 2026, GTC provided a company-driven demand reset.&lt;/p&gt;
&lt;p&gt;Samsung now has less help from the first two channels. Since June 2026, strong US labor data and inflation concern have made rates a constraint again. The market is not confidently pricing a near-term Fed rescue.&lt;/p&gt;
&lt;p&gt;That makes the third channel more important: direct demand confirmation. Late-July hyperscaler CapEx commentary can play the same role for Samsung that GTC played for NVIDIA in March 2026.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="12-practical-view-wait-for-proof-then-scale-in"&gt;12. Practical view: wait for proof, then scale in
&lt;/h2&gt;&lt;p&gt;The conclusion is not a strong sell call. It is also not an automatic buy call.&lt;/p&gt;
&lt;p&gt;The stance is Wait.&lt;/p&gt;
&lt;p&gt;A strong print followed by a two-day selloff can be a signal that the market has moved from current profit level to next growth rate. That is a different regime.&lt;/p&gt;
&lt;p&gt;The entry gates are:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Gate&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;th&gt;If confirmed&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;late-July Big Tech CapEx maintained or raised&lt;/td&gt;
 &lt;td&gt;final demand for AI memory remains intact&lt;/td&gt;
 &lt;td&gt;selloff fear weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM/NAND price momentum holds&lt;/td&gt;
 &lt;td&gt;2Q pricing logic continues into 3Q&lt;/td&gt;
 &lt;td&gt;3Q earnings support improves&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung clean 3Q QoQ remains healthy&lt;/td&gt;
 &lt;td&gt;underlying earnings path survives provision noise&lt;/td&gt;
 &lt;td&gt;quality of earnings improves&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;If two of these three gates pass, the case for staged entry improves. If all three weaken, the correction can extend.&lt;/p&gt;
&lt;p&gt;The main mistake would be to assume that record earnings automatically lead to a bounce. NVIDIA&amp;rsquo;s history says otherwise. The rebound required a catalyst that killed the specific fear behind the selloff.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="13-samsung-versus-sk-hynix"&gt;13. Samsung versus SK hynix
&lt;/h2&gt;&lt;p&gt;This note focuses on Samsung, but the read-through also matters for SK hynix.&lt;/p&gt;
&lt;p&gt;Samsung has a huge earnings base, a lower multiple, and HBM4/HBM4E catch-up optionality. But investors also have to deal with HBM purity, foundry and S.LSI losses, bonus provisions, and foreign-flow pressure.&lt;/p&gt;
&lt;p&gt;SK hynix has cleaner HBM exposure and a stronger position in NVIDIA&amp;rsquo;s supply chain. But it is already priced as the winner. If Big Tech CapEx remains strong, SK hynix is the trend-reacceleration candidate, while Samsung is the re-rating candidate where the Micron/NVIDIA read-through has been partly erased.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Lens&lt;/th&gt;
 &lt;th&gt;Samsung Electronics&lt;/th&gt;
 &lt;th&gt;SK hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Strength&lt;/td&gt;
 &lt;td&gt;earnings level, low multiple, HBM catch-up option&lt;/td&gt;
 &lt;td&gt;HBM purity, customer trust, leadership position&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Burden&lt;/td&gt;
 &lt;td&gt;HBM certification, foundry/S.LSI, provision interpretation&lt;/td&gt;
 &lt;td&gt;high expectations, crowded-winner risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Rebound condition&lt;/td&gt;
 &lt;td&gt;clean 3Q earnings and HBM customer proof&lt;/td&gt;
 &lt;td&gt;Big Tech CapEx durability and HBM allocation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Style&lt;/td&gt;
 &lt;td&gt;under-reflected alpha candidate&lt;/td&gt;
 &lt;td&gt;trend re-acceleration candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Both depend on the same final question: will hyperscalers keep spending?&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="14-conclusion-samsung-needs-better-disproof-not-just-better-earnings"&gt;14. Conclusion: Samsung needs better disproof, not just better earnings
&lt;/h2&gt;&lt;p&gt;Samsung&amp;rsquo;s 2Q26 preliminary earnings were strong. But the stock fell sharply for two straight days. Compared with 17 NVIDIA post-earnings reactions, Samsung&amp;rsquo;s path looks much closer to NVIDIA Q4 FY26 in February 2026 than to the August 2024 one-day sell-on.&lt;/p&gt;
&lt;p&gt;In that NVIDIA case, the company posted strong numbers but failed to recover the pre-announcement price for about a month. The recovery began only after GTC 2026 directly challenged the AI CapEx peak fear with a $1T through-2027 demand statement.&lt;/p&gt;
&lt;p&gt;Samsung&amp;rsquo;s equivalent proof point is not another reminder that 2Q earnings were strong. It is hyperscaler confirmation that AI infrastructure spending remains firm into 2027.&lt;/p&gt;
&lt;p&gt;That is why the late-July Big Tech earnings calls matter. Until then, the right discipline is to wait for evidence rather than buying only because the stock has fallen after a record print.&lt;/p&gt;
&lt;hr&gt;
&lt;h2 id="evidence-classification"&gt;Evidence classification
&lt;/h2&gt;&lt;h3 id="fact"&gt;Fact
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26 preliminary earnings&lt;/td&gt;
 &lt;td&gt;Revenue about KRW 171T and operating profit about KRW 89.4T, Samsung official newsroom, 2026-07-07&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung comparison periods&lt;/td&gt;
 &lt;td&gt;1Q26 revenue KRW 133.87T and operating profit KRW 57.23T; 2Q25 revenue KRW 74.57T and operating profit KRW 4.68T, Samsung official newsroom&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung price path&lt;/td&gt;
 &lt;td&gt;July 6 KRW 318,000, July 7 KRW 296,000, July 8 KRW 277,500, yfinance daily close calculation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung two-day reaction&lt;/td&gt;
 &lt;td&gt;July 7 -6.92%, July 8 -6.25%, two-day -12.74%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA Q4 FY26 event&lt;/td&gt;
 &lt;td&gt;February 25, 2026 event confirmed on NVIDIA IR&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA Q1 FY27&lt;/td&gt;
 &lt;td&gt;Revenue $81.6B, up 85% YoY and 20% QoQ, NVIDIA IR financial reports&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GTC 2026 $1T comment&lt;/td&gt;
 &lt;td&gt;Demand/order visibility through 2027 reported by Data Center Dynamics and Constellation Research&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA price-path calculations&lt;/td&gt;
 &lt;td&gt;yfinance daily closes, 17 quarters aligned with next-day reaction for after-market releases&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="inference"&gt;Inference
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung/NVIDIA comparison&lt;/td&gt;
 &lt;td&gt;This is a price-path and fear comparison, not a business-model comparison.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q4 FY26 similarity&lt;/td&gt;
 &lt;td&gt;Two-day path is the closest, and the fear set also overlaps: expectations, CapEx durability, and customer concentration.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Late-July Big Tech calls&lt;/td&gt;
 &lt;td&gt;If AI CapEx concern drove the Samsung selloff, hyperscaler guidance is the most direct disproof.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Wait stance&lt;/td&gt;
 &lt;td&gt;The print was strong, but the stock is asking about growth-rate durability. Confirmation should come before aggressive buying.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="speculation"&gt;Speculation
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Possible path&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Positive path&lt;/td&gt;
 &lt;td&gt;Big Tech CapEx holds or rises, DRAM/NAND price momentum holds, and Samsung&amp;rsquo;s clean 3Q earnings path improves.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Negative path&lt;/td&gt;
 &lt;td&gt;Two or more hyperscalers slow AI infrastructure spending or memory price momentum weakens.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Recovery timing&lt;/td&gt;
 &lt;td&gt;NVIDIA analogues suggest breakeven can take weeks, not days. For Samsung, mid-August to September is more realistic if catalysts turn positive.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="blocked"&gt;Blocked
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Why it is blocked&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26 segment detail&lt;/td&gt;
 &lt;td&gt;Preliminary earnings do not yet disclose DS, foundry, or S.LSI detail.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Late-July hyperscaler CapEx guidance&lt;/td&gt;
 &lt;td&gt;Not yet reported.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung clean 3Q QoQ&lt;/td&gt;
 &lt;td&gt;Requires later earnings and IR detail after provision adjustments.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Exact consensus beat for all 17 NVIDIA quarters&lt;/td&gt;
 &lt;td&gt;Some quarters are verified, but a uniform consensus data set for every quarter was not fully obtained.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="sources-and-calculation-notes"&gt;Sources and calculation notes
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;&lt;a class="link" href="https://news.samsung.com/global/samsung-electronics-announces-earnings-guidance-for-second-quarter-2026" target="_blank" rel="noopener"
 &gt;Samsung Electronics, Earnings Guidance for Second Quarter 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://investor.nvidia.com/financial-info/financial-reports/default.aspx" target="_blank" rel="noopener"
 &gt;NVIDIA Investor Relations, Financial Reports&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://investor.nvidia.com/events-and-presentations/events-and-presentations/event-details/2026/NVIDIA-4th-Quarter-FY26-Financial-Results-2026-sO6kGS3C2P/default.aspx" target="_blank" rel="noopener"
 &gt;NVIDIA Investor Relations, 4th Quarter FY26 Financial Results&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.datacenterdynamics.com/en/news/nvidia-ceo-jensen-huang-says-company-has-one-trillion-dollars-in-orders-through-2027/" target="_blank" rel="noopener"
 &gt;Data Center Dynamics, Jensen Huang says NVIDIA has one trillion dollars in orders through 2027&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.constellationr.com/insights/news/nvidia-gtc-2026-ai-inference-fueling-demand-boom-1-trillion-order-flow" target="_blank" rel="noopener"
 &gt;Constellation Research, Nvidia GTC 2026 and $1 trillion order flow&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Stock calculations: yfinance daily closes for &lt;code&gt;005930.KS&lt;/code&gt; and &lt;code&gt;NVDA&lt;/code&gt;, from 2022-05-20 to 2026-07-08. For NVIDIA, after-market releases are aligned so the next regular trading day is reaction D0.&lt;/li&gt;
&lt;/ol&gt;</description></item><item><title>The Big Tech Financing Relay: AI Capex Has Not Been Cut, And The Memory Bottleneck Looks Stronger</title><link>https://koreainvestinsights.com/post/hyperscaler-financing-race-ai-capex-memory-bottleneck-2026-07-07/</link><pubDate>Tue, 07 Jul 2026 21:45:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/hyperscaler-financing-race-ai-capex-memory-bottleneck-2026-07-07/</guid><description>&lt;p&gt;This is a follow-up to the late-July Big Tech earnings-call memory-thesis preview, the Micron FY3Q26 review, the 2027 semiconductor consensus-payability work, and the AI data-center capex bottleneck series.&lt;/p&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The Big Tech financing relay says one thing clearly: AI infrastructure capex has not been cut. If anything, it has outgrown the old assumption that elite hyperscalers could fund everything internally.&lt;/p&gt;
&lt;p&gt;Top hyperscaler 2026 capex guidance now sits around $695bn to $750bn. That compares with about $410bn in 2025 and about $245bn in 2024, implying roughly three consecutive years of 60% plus growth. The more important change is funding mix. The path has moved from internal cash to bonds, multi-currency bonds, off-balance-sheet structures, and now equity issuance.&lt;/p&gt;
&lt;p&gt;For memory, the implication is cleaner than for hyperscaler equities. Microsoft has tied roughly $25bn of its 2026 capex to component-price inflation. Meta has cited component pricing, especially memory, as one reason for higher 2026 capex. Amazon has also warned about component and memory-cost pressure. The buyers are absorbing the price burden. Memory suppliers receive it.&lt;/p&gt;
&lt;p&gt;The public-equity read-through is therefore not “own hyperscalers simply because they are spending.” It is more specific: watch the memory and HBM bottleneck nodes that can turn that spending into ASP, margin, and long-term contract visibility.&lt;/p&gt;
&lt;h2 id="the-financing-timeline"&gt;The Financing Timeline
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;th&gt;Issuer&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Type&lt;/th&gt;
 &lt;th&gt;Size&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2025-10-30&lt;/td&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonds&lt;/td&gt;
 &lt;td&gt;$30bn&lt;/td&gt;
 &lt;td&gt;Long-term AI and data-center funding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2025-11&lt;/td&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonds&lt;/td&gt;
 &lt;td&gt;about $15bn&lt;/td&gt;
 &lt;td&gt;First major return to US dollar bonds in years&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2025-11&lt;/td&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonds&lt;/td&gt;
 &lt;td&gt;about $17.5bn&lt;/td&gt;
 &lt;td&gt;Pre-funding AI infrastructure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-02&lt;/td&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonds&lt;/td&gt;
 &lt;td&gt;about $32bn&lt;/td&gt;
 &lt;td&gt;Multi-currency deal including a 100-year tech bond&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-02&lt;/td&gt;
 &lt;td&gt;Oracle&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Debt-led funding&lt;/td&gt;
 &lt;td&gt;$25bn&lt;/td&gt;
 &lt;td&gt;Part of a larger cloud-infrastructure funding plan&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-03-10/11&lt;/td&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonds&lt;/td&gt;
 &lt;td&gt;about $54bn&lt;/td&gt;
 &lt;td&gt;Dollar plus euro record-scale funding&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-04-30&lt;/td&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonds&lt;/td&gt;
 &lt;td&gt;$25bn&lt;/td&gt;
 &lt;td&gt;Issued just after Q1 results and higher capex guide&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-05-12&lt;/td&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;CHF bonds&lt;/td&gt;
 &lt;td&gt;CHF 2.82bn&lt;/td&gt;
 &lt;td&gt;Currency diversification&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-06-01/03&lt;/td&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Equity&lt;/td&gt;
 &lt;td&gt;$84.75bn&lt;/td&gt;
 &lt;td&gt;The decisive signal: dilution accepted for AI compute&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-07&lt;/td&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Reported bond launch&lt;/td&gt;
 &lt;td&gt;$25bn plus&lt;/td&gt;
 &lt;td&gt;Reported eight-tranche US IG deal. Final terms not yet confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The July 7 Amazon deal should not be mixed with the March record financing. As of publication, the final July 7 size, tranche spreads, new-issue concession and order book were not confirmed in public sources. It is therefore treated as a reported latest financing step, not a fully settled fact.&lt;/p&gt;
&lt;h2 id="capex-is-still-rising"&gt;Capex Is Still Rising
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026E capex&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YoY&lt;/th&gt;
 &lt;th&gt;Key point&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $200bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about +50%&lt;/td&gt;
 &lt;td&gt;AWS, custom chips and data-center buildout&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about $190bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+61%&lt;/td&gt;
 &lt;td&gt;Roughly $25bn attributed to component-price inflation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$180bn to $190bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 2x&lt;/td&gt;
 &lt;td&gt;2027 expected to rise substantially again&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$125bn to $145bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 2x&lt;/td&gt;
 &lt;td&gt;Memory and component pricing cited as pressure points&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle&lt;/td&gt;
 &lt;td style="text-align: right"&gt;volatile&lt;/td&gt;
 &lt;td style="text-align: right"&gt;high&lt;/td&gt;
 &lt;td&gt;Cloud buildout is highly capital-intensive&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Top 5&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$695bn to $750bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about +67%&lt;/td&gt;
 &lt;td&gt;Versus about $410bn in 2025 and $245bn in 2024&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The absolute dollar amount is still rising. The risk is the second derivative: even if 2027 capex reaches $1tn, the growth rate will likely slow from the 2025-2026 pace. That matters more for hyperscaler equity multiples than for memory suppliers. If memory keeps taking a larger share of the server bill of materials, memory revenue can still grow faster than total capex.&lt;/p&gt;
&lt;h2 id="why-funding-mix-matters"&gt;Why Funding Mix Matters
&lt;/h2&gt;&lt;p&gt;The escalation path is visible:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Internal cash.&lt;/li&gt;
&lt;li&gt;Straight bonds.&lt;/li&gt;
&lt;li&gt;Multi-currency bonds.&lt;/li&gt;
&lt;li&gt;SPVs, private credit and off-balance-sheet structures.&lt;/li&gt;
&lt;li&gt;Equity issuance.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Alphabet&amp;rsquo;s June 2026 equity raise is the key tell. This is not a weak company. It is one of the strongest cash-flow machines in the world. If it chooses dilution to fund AI compute, the capital intensity of the AI buildout has moved beyond the old internal-cash model.&lt;/p&gt;
&lt;p&gt;Microsoft looks different. Its visible public benchmark funding has been less aggressive than Alphabet&amp;rsquo;s or Amazon&amp;rsquo;s, which may point to stronger internal funding capacity rather than lower AI capex ambition.&lt;/p&gt;
&lt;h2 id="the-memory-read-through"&gt;The Memory Read-Through
&lt;/h2&gt;&lt;p&gt;The key change is not simply that memory prices are up. It is that hyperscalers are not cutting demand in response. They are increasing budgets, signing long-term contracts and raising capital.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Signal&lt;/th&gt;
 &lt;th&gt;Memory implication&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft component-inflation comment&lt;/td&gt;
 &lt;td&gt;More dollars are needed for similar capacity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta memory-price comment&lt;/td&gt;
 &lt;td&gt;Memory is moving the whole capex budget&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon component-cost warning&lt;/td&gt;
 &lt;td&gt;Costs are higher, but capex continues&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron SCAs&lt;/td&gt;
 &lt;td&gt;Customers are locking in long-term supply&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4/HBM4E transition&lt;/td&gt;
 &lt;td&gt;Higher value per accelerator and more wafer intensity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is why hyperscaler equities and memory suppliers should be separated. Hyperscalers bear the cost, FCF drag and payback risk. Memory suppliers receive the pricing and allocation benefit.&lt;/p&gt;
&lt;h2 id="public-equity-checklist"&gt;Public Equity Checklist
&lt;/h2&gt;&lt;p&gt;For Samsung Electronics, the questions are DS core profit, HBM4/HBM4E customer qualification, 3Q DRAM and NAND pricing, and the scale of foundry/System LSI losses.&lt;/p&gt;
&lt;p&gt;For SK Hynix, the questions are HBM4 long-term contract structure, 2027 allocation, and whether US-listed access through the SKHY ADR broadens the investor base without distorting common-share pricing.&lt;/p&gt;
&lt;p&gt;For Micron, the question is whether the SCA structure becomes the benchmark for how investors value memory-cycle durability.&lt;/p&gt;
&lt;p&gt;For hyperscalers, the clean stance is more cautious: demand is real, but FCF, debt, dilution and payback period risk are also real.&lt;/p&gt;
&lt;h2 id="scenarios"&gt;Scenarios
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th&gt;Memory read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bull&lt;/td&gt;
 &lt;td&gt;Amazon order book strong, late-July calls maintain or lift 2027 capex, memory costs remain absorbed&lt;/td&gt;
 &lt;td&gt;HBM and server DRAM pricing durability strengthens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base&lt;/td&gt;
 &lt;td&gt;2026 capex maintained, 2027 growth continues but slows, memory-price increases moderate&lt;/td&gt;
 &lt;td&gt;Memory thesis remains intact but stock reactions stay selective&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bear&lt;/td&gt;
 &lt;td&gt;Capex guides cut, payback concerns rise, component inflation causes deployment delays&lt;/td&gt;
 &lt;td&gt;ASP thesis weakens until contract pricing is confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Structural damage&lt;/td&gt;
 &lt;td&gt;Customers reduce memory content per server, HBM contract prices fall, bond demand weakens&lt;/td&gt;
 &lt;td&gt;Memory thesis needs full re-underwriting&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;The Big Tech financing relay confirms that AI infrastructure demand remains real. But it is not a simple hyperscaler-equity bullish signal. The more precise read-through is that memory is one of the bottlenecks taking a larger share of AI infrastructure dollars.&lt;/p&gt;
&lt;p&gt;The next tests are the final terms of Amazon&amp;rsquo;s July 7 deal, the late-July Big Tech calls, Samsung&amp;rsquo;s July 30 earnings call, SK Hynix&amp;rsquo;s 2Q call, and 3Q/4Q memory contract-price data.&lt;/p&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1652044/000119312526251733/d160205dfwp.htm" target="_blank" rel="noopener"
 &gt;Alphabet proposed equity raise, SEC FWP&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://abc.xyz/investor/news/news-details/2026/Alphabet-Announces-Upsize-and-Pricing-of-84-75-Billion-Equity-Capital-Raise-to-Expand-AI-Infrastructure--and-Compute-2026-QzN3D9yMAj/default.aspx" target="_blank" rel="noopener"
 &gt;Alphabet upsized $84.75bn equity raise&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://blog.google/alphabet/investor-presentation-june-2026/" target="_blank" rel="noopener"
 &gt;Alphabet investor presentation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-First-Quarter-Results/default.aspx" target="_blank" rel="noopener"
 &gt;Amazon Q1 2026 results&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1018724/000110465926026393/tm266670d7_fwp.htm" target="_blank" rel="noopener"
 &gt;Amazon March 2026 SEC FWP&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.investing.com/news/sec-filings/amazon-closes-1447-billion-multitranche-eurodenominated-bond-sale-93CH-4564195" target="_blank" rel="noopener"
 &gt;Amazon euro notes filing report&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.tomshardware.com/tech-industry/big-tech/big-techs-ai-spending-plans-reach-725-billion" target="_blank" rel="noopener"
 &gt;Big Tech AI capex spending plans&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://investors.micron.com/quarterly-results" target="_blank" rel="noopener"
 &gt;Micron quarterly results&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>Late-July Big Tech Earnings Calls and the Memory Thesis Scenario Map</title><link>https://koreainvestinsights.com/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/</link><pubDate>Tue, 07 Jul 2026 20:59:50 +0900</pubDate><guid>https://koreainvestinsights.com/post/big-tech-july-earnings-call-memory-thesis-scenarios-2026-07-07/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 earnings&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-2q26-preview-micron-surprise-erased-core-op-hbm-2026-06-29/" &gt;Samsung Electronics 2Q26 preview&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/semiconductor-2027-earnings-hyperscaler-payability-memory-nvidia-2026-06-21/" &gt;Who pays for 2027 semiconductor consensus?&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/nvidia-earnings-elasticity-hbm-cycle-samsung-hynix-2026-06-28/" &gt;NVIDIA&amp;rsquo;s elasticity as a memory-cycle guide&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;the 1H26 AI infrastructure bottleneck review&lt;/a&gt;. Related hubs: &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;AI HBM Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The late-July earnings calls from Alphabet, Microsoft, Meta and Amazon are the next stress test for the memory thesis. My probability map is: strengthening 50%, unchanged 35%, weakening 15%. Within the weakening bucket, true structural damage is only around 4%.&lt;/p&gt;
&lt;p&gt;The key question is not whether Big Tech cuts capex. The four companies are still constrained by compute, power and component availability. The real questions are narrower: whether 2027 capex falls meaningfully short of the market&amp;rsquo;s roughly $1 trillion expectation, and whether memory pricing plus memory&amp;rsquo;s share of server BOM stay intact.&lt;/p&gt;
&lt;p&gt;The execution rule is simple: do not chase before the calls. Wait for the checklist. Korean memory stocks have already risen sharply over the last year, so the upside from good news can be smaller than the downside from a bad surprise.&lt;/p&gt;
&lt;h2 id="samsungs-july-7-update"&gt;Samsung&amp;rsquo;s July 7 Update
&lt;/h2&gt;&lt;p&gt;Samsung Electronics released its 2Q26 preliminary result on July 7. It guided to revenue of about KRW 171 trillion and operating profit of about KRW 89.4 trillion. The official ranges were KRW 170-172 trillion for revenue and KRW 89.3-89.5 trillion for operating profit. Source: &lt;a class="link" href="https://news.samsung.com/global/samsung-electronics-announces-earnings-guidance-for-second-quarter-2026" target="_blank" rel="noopener"
 &gt;Samsung Electronics 2Q26 earnings guidance&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;That is broadly in line with the earlier reported OP range of KRW 87-92 trillion. It does not settle the debate. The important event is the July 30 Samsung earnings call, where investors need to separate reported OP from core semiconductor profit, check 3Q DRAM and NAND pricing tone, and listen for HBM4/HBM4E customer qualification and allocation comments.&lt;/p&gt;
&lt;h2 id="what-the-event-judges"&gt;What the Event Judges
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Local timing&lt;/th&gt;
 &lt;th&gt;Event&lt;/th&gt;
 &lt;th&gt;Status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;July 7&lt;/td&gt;
 &lt;td&gt;Samsung Electronics 2Q26 preliminary result&lt;/td&gt;
 &lt;td&gt;Officially released&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Around July 22-23 US&lt;/td&gt;
 &lt;td&gt;Alphabet 2Q26 call&lt;/td&gt;
 &lt;td&gt;Aggregator estimate, official IR date still needs confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Around July 28 US&lt;/td&gt;
 &lt;td&gt;Microsoft FY4Q26 call&lt;/td&gt;
 &lt;td&gt;Aggregator estimate, official IR page still needs confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Late July&lt;/td&gt;
 &lt;td&gt;SK Hynix 2Q26 result&lt;/td&gt;
 &lt;td&gt;Aggregators conflict, official page needs confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Around July 29 US&lt;/td&gt;
 &lt;td&gt;Meta 2Q26 call&lt;/td&gt;
 &lt;td&gt;Aggregator estimate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Around July 30 US&lt;/td&gt;
 &lt;td&gt;Amazon 2Q26 call&lt;/td&gt;
 &lt;td&gt;Aggregator estimate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;August&lt;/td&gt;
 &lt;td&gt;SK Hynix Nasdaq ADR listing&lt;/td&gt;
 &lt;td&gt;Pending SEC process&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The headline question is not &amp;ldquo;will Big Tech cut spending?&amp;rdquo; The real question is whether 2027 spending growth is still steep enough to support the memory earnings path.&lt;/p&gt;
&lt;h2 id="big-tech-capex-baseline"&gt;Big Tech Capex Baseline
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2025 actual&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026 plan&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Growth&lt;/th&gt;
 &lt;th&gt;2027 commentary&lt;/th&gt;
 &lt;th&gt;Demand evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$91.4B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$180-190B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 2x&lt;/td&gt;
 &lt;td&gt;CFO said 2027 spending should rise materially above 2026&lt;/td&gt;
 &lt;td&gt;Cloud backlog of $462B, roughly doubled sequentially&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $118B, inferred&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $190B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+61%&lt;/td&gt;
 &lt;td&gt;Double-digit revenue and operating income growth, but no explicit capex figure&lt;/td&gt;
 &lt;td&gt;RPO $627B, +99% YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$72B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$125-145B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+87.5% at midpoint&lt;/td&gt;
 &lt;td&gt;No number, flexible language&lt;/td&gt;
 &lt;td&gt;Contractual commitments rose $107B in 1Q&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$125B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $200B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+60%&lt;/td&gt;
 &lt;td&gt;No number&lt;/td&gt;
 &lt;td&gt;AWS backlog $364B, Anthropic contract separate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $410B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $710B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+73%&lt;/td&gt;
 &lt;td&gt;Market expectation above $1T, +41%&lt;/td&gt;
 &lt;td&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Alphabet is the strongest setup. It raised 2026 capex guidance, said 2027 would be meaningfully higher, and later announced an $80B capital raise to expand AI infrastructure and compute. Microsoft is the key swing factor: RPO is huge, but sequential RPO growth needs to reaccelerate. Meta is two-sided: it raised guidance because of component and memory costs, but it also kept flexibility language for 2027. Amazon has the largest absolute capex plan, but its free cash flow pressure makes its tone worth watching.&lt;/p&gt;
&lt;h2 id="memory-price-facts"&gt;Memory Price Facts
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;DRAM&lt;/th&gt;
 &lt;th style="text-align: right"&gt;NAND&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55-60%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+33-38%&lt;/td&gt;
 &lt;td&gt;Server DRAM was even stronger&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+58-63%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+70-75%&lt;/td&gt;
 &lt;td&gt;Peak rate of increase&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3Q26 forecast&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+13-18%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+10-15%&lt;/td&gt;
 &lt;td&gt;Still rising, but at a slower pace&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;4Q26 reference&lt;/td&gt;
 &lt;td style="text-align: right"&gt;PC DRAM +3-8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Not specified&lt;/td&gt;
 &lt;td&gt;A consumer-side reference point&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Source: &lt;a class="link" href="https://www.trendforce.com/presscenter/news/20260703-13134.html" target="_blank" rel="noopener"
 &gt;TrendForce 3Q26 price outlook&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;The 3Q deceleration is not a demand collapse. It reflects base effects, consumer affordability limits and a rising share of long-term contracts. The more important point is that server memory remains tight. If Big Tech capex grows more slowly while memory takes a larger share of server BOM, memory revenue can still grow faster than aggregate capex.&lt;/p&gt;
&lt;h2 id="micron-sca-matters"&gt;Micron SCA Matters
&lt;/h2&gt;&lt;p&gt;Micron disclosed 16 strategic customer agreements in its FY3Q26 materials. The contracts run from 2026 through 2030, include take-or-pay structures, and cover about 20% of DRAM volume and roughly one-third of NAND volume. Minimum contracted revenue is around $100B, and customer deposits plus commitments are around $22B. Source: &lt;a class="link" href="https://investors.micron.com/static-files/631b1a32-5537-46ae-8f40-82e42fc79dfe" target="_blank" rel="noopener"
 &gt;Micron prepared remarks&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;This is the most important underappreciated change in the cycle. Big Tech is not merely accepting higher memory prices. It is signing multi-year, non-cancelable supply assurance contracts. That lowers the historical cyclicality discount for memory earnings. It does not eliminate risk, because price ceilings may cap upside on contracted volume, but it changes the downside profile.&lt;/p&gt;
&lt;p&gt;For SK Hynix, the key question is whether its long-term contracts have similar floors without the same upside caps. That detail is not officially confirmed. It should be the first question on the SK Hynix 2Q call.&lt;/p&gt;
&lt;h2 id="scenario-map"&gt;Scenario Map
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Probability&lt;/th&gt;
 &lt;th&gt;Meaning for memory stocks&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Strengthening&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50%&lt;/td&gt;
 &lt;td&gt;Earnings revision cycle extends. Upside depends on how explicit 2027 capex becomes.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Neutral&lt;/td&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td&gt;2026 plans remain intact, but 2027 language stays cautious. Short-term headline noise likely.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Weakening&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td&gt;Split into narrative weakening and structural damage.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Narrative weakening&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 11%&lt;/td&gt;
 &lt;td&gt;Stocks correct, but price and contract data stay healthy. Could become an entry window.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Structural damage&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 4%&lt;/td&gt;
 &lt;td&gt;Thesis reset. Examples: 2026 capex cut, memory-content redesign, digestion language from multiple companies.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="call-by-call-update-rules"&gt;Call-by-Call Update Rules
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Sequence&lt;/th&gt;
 &lt;th&gt;Result&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Strengthening probability update&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td&gt;Plan maintained, 2027 reaffirmed, backlog up&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50 → 55&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alphabet&lt;/td&gt;
 &lt;td&gt;2027 language retreats or backlog stalls&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50 → below 40&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td&gt;$190B plan maintained, supply tightness, 2027 growth tone, RPO growth resumes&lt;/td&gt;
 &lt;td style="text-align: right"&gt;55 → 62&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td&gt;Normalization or single-digit growth tone&lt;/td&gt;
 &lt;td style="text-align: right"&gt;55 → 45&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td&gt;Plan maintained, commitments rise, cloud monetization becomes official&lt;/td&gt;
 &lt;td style="text-align: right"&gt;62 → 66&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td&gt;April-style flexibility language returns with numbers&lt;/td&gt;
 &lt;td style="text-align: right"&gt;62 → 50&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td&gt;$200B maintained, memory price absorbed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;66 → 68-70&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td&gt;Amazon-only pacing comments&lt;/td&gt;
 &lt;td style="text-align: right"&gt;66 → 60, not enough to kill the thesis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;If two or more companies cut plans or use digestion language, the table no longer applies. That becomes a thesis reset exercise.&lt;/p&gt;
&lt;h2 id="korean-memory-checklist"&gt;Korean Memory Checklist
&lt;/h2&gt;&lt;p&gt;This section is a public research checklist, not a trading instruction. The late-July Big Tech calls matter here because they help test the Korean memory thesis.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Role&lt;/th&gt;
 &lt;th&gt;What to verify&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Korean memory core with DRAM and NAND volume leverage&lt;/td&gt;
 &lt;td&gt;DS core profit, 3Q memory price assumptions and HBM4 customer qualification.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;HBM purity and long-term contract structure&lt;/td&gt;
 &lt;td&gt;HBM shipment, ASP durability, contract pricing structure and HBM4 allocation.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="phrases-to-listen-for"&gt;Phrases To Listen For
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Direction&lt;/th&gt;
 &lt;th&gt;Phrases&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Positive&lt;/td&gt;
 &lt;td&gt;supply cannot keep up with demand / customer commitments rising / backlog rising / component or memory price pressure absorbed by budgets / 2027 capex expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Negative&lt;/td&gt;
 &lt;td&gt;digestion phase / lower utilization / investment pacing / optimization / delayed customer commitments / easing component price pressure / memory supply improving earlier than expected / memory content redesign&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="kill-conditions"&gt;Kill Conditions
&lt;/h2&gt;&lt;p&gt;One item triggers reassessment. Two or more start the thesis-reset process.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Any of the four companies lowers 2026 capex plans.&lt;/li&gt;
&lt;li&gt;Microsoft RPO or Alphabet Cloud backlog turns negative sequentially.&lt;/li&gt;
&lt;li&gt;Companies use &amp;ldquo;digestion&amp;rdquo; or &amp;ldquo;pacing purchases to power availability&amp;rdquo; language.&lt;/li&gt;
&lt;li&gt;A concrete server redesign reduces memory content per server because of price.&lt;/li&gt;
&lt;li&gt;TrendForce cuts 3Q memory price growth back to single digits while Samsung DS core profit disappoints.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The fourth item is the real kill condition. Supply-driven forced reductions are different from price-driven voluntary memory-content cuts.&lt;/p&gt;
&lt;h2 id="evidence-classification"&gt;Evidence Classification
&lt;/h2&gt;&lt;p&gt;Facts: Samsung&amp;rsquo;s 2Q26 preliminary result, Samsung&amp;rsquo;s July 30 call schedule, Alphabet&amp;rsquo;s 2026 capex range and AI infrastructure capital raise, Microsoft&amp;rsquo;s $190B capex run-rate and $627B RPO, Meta&amp;rsquo;s $125-145B plan and $107B commitment increase, Amazon&amp;rsquo;s roughly $200B plan and AWS backlog, TrendForce&amp;rsquo;s 3Q price outlook, Micron&amp;rsquo;s FY3Q26 SCA disclosures.&lt;/p&gt;
&lt;p&gt;Inferences: the 50/35/15 scenario map, the idea that +73% to +41% capex deceleration is already embedded, the view that Meta&amp;rsquo;s cloud report is more monetization than capex-cut signal, and the view that long-term contracts reduce memory cyclicality.&lt;/p&gt;
&lt;p&gt;Speculation: SK Hynix contract upside without price ceilings, possible Microsoft financing, the exact timing of a 2027 digestion phase, and Meta&amp;rsquo;s cloud economics.&lt;/p&gt;
&lt;p&gt;Blocked: verified 2027 consensus multiples for Samsung and SK Hynix, Marvell&amp;rsquo;s exact share in Trainium 3 and 4, SK Hynix contract details, company-level 2027 capex consensus distributions, memory supplier allocation by hyperscaler and some official late-July earnings dates.&lt;/p&gt;
&lt;p&gt;The conclusion is practical: the odds favor a stronger memory thesis, but price already reflects part of that. The right way to trade the event is to update probabilities as evidence arrives, not to make a blind prediction before the calls.&lt;/p&gt;</description></item><item><title>Dukovany And The Westinghouse Settlement: Contract Risk Has Fallen, Margin Risk Remains</title><link>https://koreainvestinsights.com/post/dukovany-westinghouse-ip-settlement-korea-nuclear-margin-risk-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 22:25:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/dukovany-westinghouse-ip-settlement-korea-nuclear-margin-risk-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This memo follows &lt;a class="link" href="https://koreainvestinsights.com/post/us-investment-special-act-team-korea-nuclear-opportunity-2026-06-13/" &gt;Team Korea&amp;rsquo;s nuclear opportunity under the US investment package&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/hyundai-ec-woojin-us-nuclear-expansion-bargaining-power-2026-06-13/" &gt;Hyundai E&amp;amp;C and Woojin in US nuclear expansion&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/construction-epc-rerating-ai-power-nuclear-datacenter-reconstruction-2026-06-14/" &gt;construction EPC rerating&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The Westinghouse IP risk that could have blocked the Czech Dukovany 5 and 6 project has fallen meaningfully. Westinghouse, KEPCO and KHNP announced a global settlement on January 16, 2025, and the Dukovany contract was signed in June 2025.&lt;/p&gt;
&lt;p&gt;That does not mean Korean nuclear exports are now free to expand across Europe without constraints. The remaining issue is economic: how much Westinghouse can take as a toll, how broad the reported market-access restrictions are, how much local content is required, and who carries fixed-price or schedule risk.&lt;/p&gt;
&lt;p&gt;The confirmed project numbers are large. World Nuclear News reported the Dukovany 5 and 6 EPC contract for two APR1000 units at about CZK 407 billion, or USD 18.6 billion. KEPCO E&amp;amp;C announced an A/E services contract worth EUR 725 million, about KRW 1.2 trillion. Doosan Enerbility announced a KRW 320 billion steam turbine and turbine-control contract through Doosan Skoda Power.&lt;/p&gt;
&lt;p&gt;The investment conclusion is simple: &lt;strong&gt;contract survival risk has fallen, but margin risk remains.&lt;/strong&gt; The question has moved from &amp;ldquo;can Korea win and sign the project?&amp;rdquo; to &amp;ldquo;how much profit remains for Korea&amp;rsquo;s listed value chain after Westinghouse, localization and project risk?&amp;rdquo;&lt;/p&gt;
&lt;h2 id="1-what-has-been-resolved"&gt;1. What Has Been Resolved
&lt;/h2&gt;&lt;p&gt;In August 2024, Westinghouse protested CEZ&amp;rsquo;s selection of KHNP as preferred bidder for the Dukovany project. Westinghouse argued that KHNP&amp;rsquo;s APR1000 and APR1400 designs used Westinghouse-licensed System 80 technology and that KHNP lacked the right to sublicense or retransfer that technology to CEZ and local suppliers. Westinghouse also raised US export-control issues under DOE Part 810.&lt;/p&gt;
&lt;p&gt;The landscape changed on January 16, 2025. Westinghouse announced a global settlement with KEPCO and KHNP to resolve the IP dispute. The company said the agreement would allow both parties to move forward with new nuclear reactor deployment, while the terms remained confidential.&lt;/p&gt;
&lt;p&gt;The US Department of Energy welcomed the settlement on the same date. CEZ&amp;rsquo;s Dukovany project page then confirmed that the contract for two new Dukovany units was signed in June 2025 after the legal injunction was lifted.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Issue&lt;/th&gt;
 &lt;th&gt;Current state&lt;/th&gt;
 &lt;th&gt;Investment reading&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Westinghouse IP dispute&lt;/td&gt;
 &lt;td&gt;Global settlement announced in January 2025&lt;/td&gt;
 &lt;td&gt;The main legal obstacle to signing has fallen&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Settlement terms&lt;/td&gt;
 &lt;td&gt;Confidential&lt;/td&gt;
 &lt;td&gt;Cost, market access and technology-independence risks remain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Dukovany 5 and 6&lt;/td&gt;
 &lt;td&gt;Contract signed in June 2025&lt;/td&gt;
 &lt;td&gt;The project is alive&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EU foreign subsidy review&lt;/td&gt;
 &lt;td&gt;EC did not open an in-depth FSR investigation, according to KHNP/WNN&lt;/td&gt;
 &lt;td&gt;One regulatory risk has eased&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EDF Czech litigation&lt;/td&gt;
 &lt;td&gt;EDF lost and said it would not pursue further Czech legal action&lt;/td&gt;
 &lt;td&gt;Tender-procedure risk has eased&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-why-westinghouse-did-not-simply-lose"&gt;2. Why Westinghouse Did Not Simply Lose
&lt;/h2&gt;&lt;p&gt;The US court dismissal of Westinghouse&amp;rsquo;s Part 810 declaratory-judgment pathway should not be read as a full defeat on the economics. Finnegan&amp;rsquo;s legal analysis notes that the court found Part 810 did not give Westinghouse an independent judicially remediable right for that claim. It did not decide that Westinghouse&amp;rsquo;s broader contractual or IP economics were worthless.&lt;/p&gt;
&lt;p&gt;This is why the settlement matters. Westinghouse may have moved from a legal blocker to an economic toll collector.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Simplistic view&lt;/th&gt;
 &lt;th&gt;More accurate view&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Westinghouse lost&lt;/td&gt;
 &lt;td&gt;The Part 810 court route was limited, but settlement economics may remain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea can export freely&lt;/td&gt;
 &lt;td&gt;Dukovany is open, but regional market access may still be constrained&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;IP risk disappeared&lt;/td&gt;
 &lt;td&gt;signing risk fell, but cost and margin risk remain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="3-the-reported-toll"&gt;3. The Reported Toll
&lt;/h2&gt;&lt;p&gt;Yonhap reported, citing industry sources, that KHNP may be restricted from new nuclear bids in North America, the EU, Britain, Japan and Ukraine, with the Czech Republic carved out as an exception. The same report said the agreement may require about USD 650 million of Westinghouse goods and services per exported reactor and about USD 175 million of technology license fees per reactor.&lt;/p&gt;
&lt;p&gt;Korea Times also reported that the presidential office ordered a review of whether the KHNP, KEPCO and Westinghouse agreement followed proper legal basis and procedure.&lt;/p&gt;
&lt;p&gt;These reported terms are not public settlement terms. They should be treated as reported but unconfirmed. Still, the direction is important.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Reported condition&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;th&gt;Investment implication&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;USD 650 million of Westinghouse goods and services per reactor&lt;/td&gt;
 &lt;td&gt;Part of the headline order may flow to the US supply chain&lt;/td&gt;
 &lt;td&gt;Korean revenue capture is lower than headline EPC value&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;USD 175 million technology fee per reactor&lt;/td&gt;
 &lt;td&gt;IP toll&lt;/td&gt;
 &lt;td&gt;Project margin is reduced&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reported limits in North America, EU, UK, Japan and Ukraine&lt;/td&gt;
 &lt;td&gt;Czech project may be an exception, not a broad European beachhead&lt;/td&gt;
 &lt;td&gt;Do not extrapolate Dukovany into all-Europe TAM&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Technology independence verification&lt;/td&gt;
 &lt;td&gt;New reactor or SMR exports may also face constraints&lt;/td&gt;
 &lt;td&gt;The premium for independent Korean exports should be moderated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Using only the reported figures, two reactors would imply USD 1.65 billion of Westinghouse goods, services and license fees. Against a USD 18.6 billion project, that is roughly 8.9%. This is not a confirmed contract calculation, but it is a useful way to think about margin leakage.&lt;/p&gt;
&lt;h2 id="4-project-size-is-not-profit"&gt;4. Project Size Is Not Profit
&lt;/h2&gt;&lt;p&gt;World Nuclear News reported that the Dukovany 5 and 6 EPC contract covers two APR1000 units at a projected cost of CZK 407 billion, or USD 18.6 billion, with construction targeted to start in 2029.&lt;/p&gt;
&lt;p&gt;That is a large project. But nuclear EPC value is not the same as shareholder value. The key variables are price escalation, localization, financing, schedule guarantees, delay penalties, equipment scope and Westinghouse-related costs.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Driver&lt;/th&gt;
 &lt;th&gt;What to watch&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Price&lt;/td&gt;
 &lt;td&gt;EPC value, escalation clauses, FX and inflation treatment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Quantity&lt;/td&gt;
 &lt;td&gt;Dukovany 5 and 6 plus possible Temelin options&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cost&lt;/td&gt;
 &lt;td&gt;Westinghouse toll, local content, fixed-price risk, financing and schedule guarantees&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The most common mistake is to buy the headline order while ignoring the cost stack.&lt;/p&gt;
&lt;h2 id="5-value-chain-implications"&gt;5. Value Chain Implications
&lt;/h2&gt;&lt;h3 id="kepco-ec"&gt;KEPCO E&amp;amp;C
&lt;/h3&gt;&lt;p&gt;KEPCO E&amp;amp;C is the cleanest direct exposure. It announced a EUR 725 million A/E services contract with KHNP for Dukovany 5 and 6, equivalent to about KRW 1.2 trillion. The scope includes A/E services, procurement support for auxiliary equipment and licensing activities for two 1,000-MW-class APR1000 units.&lt;/p&gt;
&lt;p&gt;This gives KEPCO E&amp;amp;C direct design and licensing exposure. The next question is revenue recognition, staffing intensity, change-order scope and operating margin.&lt;/p&gt;
&lt;h3 id="doosan-enerbility"&gt;Doosan Enerbility
&lt;/h3&gt;&lt;p&gt;Doosan Enerbility announced a KRW 320 billion contract with Doosan Skoda Power for steam turbines and turbine-control systems for Dukovany 5 and 6. This is important because it combines Korean nuclear expertise with Czech local manufacturing through Doosan Skoda Power.&lt;/p&gt;
&lt;p&gt;Doosan is the strongest beta to Czech nuclear equipment and potential Temelin follow-on work. But the stock already carries nuclear, SMR, gas turbine and AI-power narratives. Margin and follow-on contracts matter more than the headline.&lt;/p&gt;
&lt;h3 id="daewoo-ec"&gt;Daewoo E&amp;amp;C
&lt;/h3&gt;&lt;p&gt;Daewoo E&amp;amp;C can be a construction exposure, but it needs more caution. Construction and EPC have large revenue potential, but also carry schedule, cost and delay-penalty risk. The key question is contract structure, not just participation.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th&gt;Key check&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;A/E revenue recognition, margin, change orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Doosan Enerbility&lt;/td&gt;
 &lt;td&gt;Watchlist / Wait&lt;/td&gt;
 &lt;td&gt;Equipment margin, Temelin option, additional nuclear scope&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;Fixed-price exposure, schedule risk and local partner split&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KEPCO KPS, Woori Technology, Woojin&lt;/td&gt;
 &lt;td&gt;Follow-up candidates&lt;/td&gt;
 &lt;td&gt;Actual O&amp;amp;M, I&amp;amp;C or instrumentation contracts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="6-catalysts-and-invalidation"&gt;6. Catalysts And Invalidation
&lt;/h2&gt;&lt;p&gt;Upside catalysts:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Catalyst&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Additional subcontract announcements&lt;/td&gt;
 &lt;td&gt;Korean listed value-chain capture expands&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Clarity on Czech state-aid and financing structure&lt;/td&gt;
 &lt;td&gt;Financing risk falls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Temelin 3 and 4 options&lt;/td&gt;
 &lt;td&gt;Dukovany becomes a repeat project, not a one-off&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KHNP-Westinghouse cooperation model&lt;/td&gt;
 &lt;td&gt;Market-access uncertainty may fall&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pre-2029 permitting, site and supply-chain milestones&lt;/td&gt;
 &lt;td&gt;Project execution risk falls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Invalidation risks:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;th&gt;Impact&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Settlement terms are worse than reported&lt;/td&gt;
 &lt;td&gt;Korean nuclear export economics deteriorate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Fixed-price and schedule risk fall heavily on Korean parties&lt;/td&gt;
 &lt;td&gt;EPC and construction stocks de-rate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Temelin and broader Europe remain closed&lt;/td&gt;
 &lt;td&gt;Multiple expansion based on Europe TAM fades&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Key subcontracting shifts to Czech or Westinghouse supply chains&lt;/td&gt;
 &lt;td&gt;Korean listed-company capture shrinks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reported margins disappoint in quarterly earnings&lt;/td&gt;
 &lt;td&gt;Headline-order premium reverses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="bottom-line"&gt;Bottom Line
&lt;/h2&gt;&lt;p&gt;Dukovany is alive. The legal and procedural risks that could have blocked the contract have fallen. But the equity question has changed. Investors should no longer ask only whether Korea won the project. They should ask how much of the project becomes high-quality profit after Westinghouse tolls, localization, fixed-price risk and financing terms.&lt;/p&gt;
&lt;p&gt;The best framing is:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Dukovany confirms Korea&amp;rsquo;s nuclear export survival. The next rerating depends on margin and repeatability.&lt;/strong&gt;&lt;/p&gt;</description></item><item><title>SK Square Investment Memo: When Does SK Hynix Cash Flow Become SK Square Cash?</title><link>https://koreainvestinsights.com/post/sk-square-hynix-cash-flow-nav-discount-listed-ai-semi-allocator-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 21:40:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/sk-square-hynix-cash-flow-nav-discount-listed-ai-semi-allocator-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This memo follows &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-f1a-f6-skhy-adr-hbm-capex-2026-07-04/" &gt;SK Hynix F-1/A and F-6&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-adr-000660-price-path-leverage-etf-plumbing-2026-07-04/" &gt;SK Hynix ADR price path&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-hbm-market-share-ai-memory-demand-2026/" &gt;SK Hynix HBM market share 2026&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 earnings&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;SK Square should not be framed as the investment vehicle for the entire SK Group. SK Inc. is the official group holding company and portfolio coordinator. For public-market investors, SK Square is better defined as:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;A listed AI and semiconductor capital allocator built on SK Hynix equity value, NAV discount compression and a call option on future SK Hynix special cash dividends.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;The key variable is not SK Hynix net income by itself. Higher SK Hynix earnings lift SK Square&amp;rsquo;s equity-method income and NAV, but they do not automatically become SK Square cash. Cash reaches SK Square through SK Hynix dividends, stake sales, collateralization or SK Square&amp;rsquo;s own portfolio exits.&lt;/p&gt;
&lt;p&gt;Current view: &lt;strong&gt;Wait / Conditional Buy&lt;/strong&gt;. The buy case strengthens when SK Hynix free cash flow converts into cash dividends, SK Square receives that cash, and the company allocates it to buyback cancellation or Hynix-linked AI and semiconductor investments.&lt;/p&gt;
&lt;h2 id="1-correct-framing"&gt;1. Correct Framing
&lt;/h2&gt;&lt;p&gt;SK Inc. is the official holding company for SK Group. It manages the broader group portfolio across energy, AI, semiconductors and other businesses. SK Square is different. It is an ICT and semiconductor investment company whose NAV is overwhelmingly driven by SK Hynix.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;SK Inc.&lt;/th&gt;
 &lt;th&gt;SK Square&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Role&lt;/td&gt;
 &lt;td&gt;Group holding company and portfolio manager&lt;/td&gt;
 &lt;td&gt;ICT and semiconductor investment company&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Main exposure&lt;/td&gt;
 &lt;td&gt;Broad SK Group portfolio&lt;/td&gt;
 &lt;td&gt;SK Hynix stake and AI/semi portfolio&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investor thesis&lt;/td&gt;
 &lt;td&gt;Group restructuring and holding-company discount&lt;/td&gt;
 &lt;td&gt;SK Hynix NAV, cash conversion, NAV discount compression&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Main risk&lt;/td&gt;
 &lt;td&gt;Conglomerate complexity&lt;/td&gt;
 &lt;td&gt;Hynix dividends may not convert into SK Square cash&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This distinction matters. Calling SK Square the group&amp;rsquo;s treasury overstates the case. The cleaner thesis is a listed AI and semiconductor allocator tied to SK Hynix cash-flow conversion.&lt;/p&gt;
&lt;h2 id="2-nav-is-mostly-sk-hynix"&gt;2. NAV Is Mostly SK Hynix
&lt;/h2&gt;&lt;p&gt;SK Square&amp;rsquo;s official NAV page showed the following as of July 2, 2026.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Total NAV&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 324.7T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix stake value&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 319.52T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net cash&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 0.79T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAV per share&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,460,288&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square share price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,525,000&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAV discount&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 38.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The discount is calculated as &lt;code&gt;1 - 1,525,000 / 2,460,288&lt;/code&gt;. SK Square holds 146,124,531 SK Hynix shares, or 20.5%.&lt;/p&gt;
&lt;h2 id="3-earnings-are-not-cash"&gt;3. Earnings Are Not Cash
&lt;/h2&gt;&lt;p&gt;SK Hynix earnings affect SK Square through accounting income and NAV. But SK Square needs actual cash for buybacks, cancellations, dividends or new investments.&lt;/p&gt;
&lt;p&gt;SK Hynix&amp;rsquo;s fixed dividend for 2025-2027 is KRW 1,500 per share. Applied to SK Square&amp;rsquo;s 146,124,531 shares, this equals KRW 219.2B, or about KRW 0.219T per year. That is about 28% of SK Square&amp;rsquo;s KRW 0.79T net cash. It is meaningful, but not enough to re-rate a holding company with KRW 319.52T of SK Hynix stake value.&lt;/p&gt;
&lt;p&gt;The real option is special or additional cash dividends.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Total SK Hynix cash dividend&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Square pre-tax receipt&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Estimated after-tax receipt&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 0.205T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 0.194T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 5T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.025T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 0.969T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 10T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.050T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 1.937T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 20T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.100T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 3.875T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The after-tax estimates use a simplified Korean dividend received deduction framework. A domestic parent owning at least 20% and less than 50% of a domestic subsidiary may exclude 80% of dividends from taxable income. Applying a 27.5% tax rate to the remaining 20% implies about 5.5% leakage.&lt;/p&gt;
&lt;h2 id="4-nav-discount-math"&gt;4. NAV Discount Math
&lt;/h2&gt;&lt;p&gt;Holding NAV per share constant at KRW 2,460,288, the implied share price changes materially as the discount narrows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;NAV discount&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Implied price&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;50%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,230,144&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;45%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,353,158&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;40%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,476,173&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;38%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,525,379&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;35%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,599,187&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;30%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,722,202&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1,845,216&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;SK Square has stated an objective of reducing its NAV discount to 30% or less by 2028. That is a target, not a guarantee. The market needs evidence: Hynix cash dividends, SK Square buyback cancellations, non-core exits and disciplined AI/semi investments.&lt;/p&gt;
&lt;h2 id="5-shareholder-return-and-capital-allocation"&gt;5. Shareholder Return And Capital Allocation
&lt;/h2&gt;&lt;p&gt;SK Square&amp;rsquo;s 2023-2025 shareholder-return policy uses at least 30% of recurring dividend income and part of meaningful investment gains for shareholder returns. Methods include buybacks, cancellations and cash dividends.&lt;/p&gt;
&lt;p&gt;At a high NAV discount, buyback cancellation is powerful. If a company trades at a 38% NAV discount, buying back stock is economically similar to buying KRW 1 of NAV for KRW 0.62. That means low-return investments are hard to justify unless they have clear Hynix-linked strategic value.&lt;/p&gt;
&lt;h2 id="6-us-cash-is-not-automatically-trapped-cash"&gt;6. U.S. Cash Is Not Automatically Trapped Cash
&lt;/h2&gt;&lt;p&gt;SK Hynix&amp;rsquo;s U.S. customer revenue is large. Reported 2025 U.S. customer revenue was KRW 66.885T, about 69% of total revenue. SK hynix America reportedly generated KRW 58.693T of sales and KRW 235.9B of net income, implying a net margin of about 0.40%.&lt;/p&gt;
&lt;p&gt;That low margin suggests the U.S. sales entity may be a limited-margin distributor rather than the place where most economic profit accumulates. Cash can return to the Korean parent through product payment, transfer-pricing settlement, intercompany payments, dividends, royalties or other structures.&lt;/p&gt;
&lt;p&gt;The important question for SK Square is not whether U.S. revenue is large. It is whether SK Hynix parent-level free cash flow grows and becomes cash dividends.&lt;/p&gt;
&lt;h2 id="7-skhy-adr-read-through"&gt;7. SKHY ADR Read-Through
&lt;/h2&gt;&lt;p&gt;SK Hynix is pursuing a U.S. ADR listing. Reuters reported that the company may raise up to $29.4B, with proceeds for Korean semiconductor factories and EUV equipment.&lt;/p&gt;
&lt;p&gt;For SK Square, the ADR is positive for investor access and potentially for SK Hynix&amp;rsquo;s valuation. But it is not a direct cash-in event for SK Square. The proceeds go to SK Hynix. SK Square benefits through NAV, valuation and future dividends, while also facing potential ownership dilution from new share issuance.&lt;/p&gt;
&lt;h2 id="8-investment-ideas"&gt;8. Investment Ideas
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Idea&lt;/th&gt;
 &lt;th&gt;Thesis&lt;/th&gt;
 &lt;th&gt;Required proof&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Hynix FCF-to-cash option&lt;/td&gt;
 &lt;td&gt;Hynix earnings matter only when they become SK Square cash&lt;/td&gt;
 &lt;td&gt;FCF after capex and meaningful cash dividends&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Listed AI/semi allocator&lt;/td&gt;
 &lt;td&gt;SK Square becomes a disciplined Hynix-linked investment platform&lt;/td&gt;
 &lt;td&gt;Investments tied to Hynix ecosystem, not generic AI hype&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAV discount compression&lt;/td&gt;
 &lt;td&gt;Discount narrows from about 38% toward 30%&lt;/td&gt;
 &lt;td&gt;Buyback cancellation, non-core exits, capital discipline&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. trapped-cash misunderstanding&lt;/td&gt;
 &lt;td&gt;Large U.S. revenue does not mean large trapped U.S. profit&lt;/td&gt;
 &lt;td&gt;Parent-level FCF and tax-efficient cash flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="9-practical-view"&gt;9. Practical View
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Security&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square&lt;/td&gt;
 &lt;td&gt;Wait / Conditional Buy&lt;/td&gt;
 &lt;td&gt;NAV discount is meaningful, but cash conversion and capital allocation need confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;Source of cash flow and NAV&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Inc.&lt;/td&gt;
 &lt;td&gt;Separate thesis&lt;/td&gt;
 &lt;td&gt;Official holding company, but less direct Hynix cash-flow exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Entry conditions:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th&gt;What to see&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Wider discount&lt;/td&gt;
 &lt;td&gt;Updated NAV discount at or above 45% while the Hynix HBM thesis remains intact&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cash conversion&lt;/td&gt;
 &lt;td&gt;Hynix special cash dividend or early additional shareholder return, followed by SK Square buyback cancellation or Hynix-synergy investment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Allocator trust&lt;/td&gt;
 &lt;td&gt;Non-core exits, AI/semi discipline, no group-support capital allocation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Invalidation:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Hynix earnings are strong but FCF is absorbed by capex and working capital&lt;/td&gt;
 &lt;td&gt;No SK Square cash conversion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hynix returns capital mainly through buybacks rather than cash dividends&lt;/td&gt;
 &lt;td&gt;Limited SK Square cash inflow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square uses Hynix dividends for group support&lt;/td&gt;
 &lt;td&gt;Holding-company discount widens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Low-return investments take priority over buyback cancellation&lt;/td&gt;
 &lt;td&gt;Capital allocation trust declines&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADR dilution outweighs lower cost of capital&lt;/td&gt;
 &lt;td&gt;SK Square&amp;rsquo;s Hynix stake economics weaken&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;SK Square is not cash-rich today. Official net cash is KRW 0.79T. But it owns KRW 319.52T of SK Hynix stake value. If SK Hynix AI-memory earnings convert into free cash flow and cash dividends, SK Square can receive trillions of won over time.&lt;/p&gt;
&lt;p&gt;The strongest thesis is not “SK Hynix is good, therefore SK Square is good.” It is more specific: SK Hynix FCF must become cash dividends, SK Square must allocate that cash to buyback cancellation and Hynix-linked investments, and the NAV discount must compress structurally.&lt;/p&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sksquare.com/eng/ir/nav.do" target="_blank" rel="noopener"
 &gt;SK Square Shareholding Structure &amp;amp; NAV&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sksquare.com/eng/ir/return.do" target="_blank" rel="noopener"
 &gt;SK Square Shareholder Return&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.skhynix.com/ir/UI-FR-IR04/" target="_blank" rel="noopener"
 &gt;SK hynix Shareholder Status&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.skhynix.com/ir/UI-FR-IR05/" target="_blank" rel="noopener"
 &gt;SK hynix Shareholder Return&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sk-inc.com/en/main/mainpage.aspx" target="_blank" rel="noopener"
 &gt;SK Inc.&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://taxsummaries.pwc.com/republic-of-korea/corporate/income-determination" target="_blank" rel="noopener"
 &gt;PwC Korea corporate income determination&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.irs.gov/pub/irs-trty/korea.pdf" target="_blank" rel="noopener"
 &gt;U.S.-Korea tax treaty PDF&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.reuters.com/world/asia-pacific/south-koreas-sk-hynix-says-raise-29-bln-us-adr-listing-2026-06-24/" target="_blank" rel="noopener"
 &gt;Reuters: SK Hynix targets up to $29.4bn U.S. ADR listing&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://news.skhynix.com/sk-hynix-announces-fy25-financial-results/" target="_blank" rel="noopener"
 &gt;SK hynix FY25 financial results&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.asiae.co.kr/en/article/2026031717471236785" target="_blank" rel="noopener"
 &gt;AsiaE: SK hynix U.S. revenue coverage&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>SK Gas LNG Cold Energy: How Much Of The AI Data-Center Cooling Option Can Become Earnings?</title><link>https://koreainvestinsights.com/post/sk-gas-lng-cold-energy-ai-datacenter-cooling-option-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 14:48:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/sk-gas-lng-cold-energy-ai-datacenter-cooling-option-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context
This follows &lt;a class="link" href="https://koreainvestinsights.com/post/sk-gas-sk-oceanplant-ai-datacenter-power-second-line-2026-07-04/" &gt;SK Gas vs SK Oceanplant&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-power-bottleneck-korea-value-chain-screen-2026-07-04/" &gt;AI data-center power bottleneck map&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-capex-5p3t-korea-power-substrate-storage-bottleneck-2026-06-05/" &gt;AI data-center CapEx bottlenecks&lt;/a&gt;. The related hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-equipment-ip-hub/" &gt;Korean Semiconductor Value Chain Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The right wording matters. The confirmed fact is not that SK Gas has disclosed a commercial AI data-center cooling project. The confirmed fact is that an SK Gas engineer published the technical logic for connecting LNG cold energy with AI data-center cooling through SK Group&amp;rsquo;s DevOcean channel.&lt;/p&gt;
&lt;p&gt;The January 2, 2025 DevOcean post, authored by Byeon Jae-jun of SK Gas, says he works on a task force linking LNG cold-energy utilization technology with AI data centers. The post explains that LNG produces about 0.2 MWh of cold energy per ton when warmed from about -162°C liquid to about 5°C gas, and that most of this cold energy is currently discarded through seawater heat exchange.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The investment conclusion is precise:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Confirmed&lt;/td&gt;
 &lt;td&gt;SK Gas-linked technical logic exists. Ulsan LNG infrastructure and the SK-AWS AI data-center direction are publicly visible.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investment logic&lt;/td&gt;
 &lt;td&gt;LNG cold energy is not new cooling energy. It is the reallocation of otherwise wasted low-temperature energy from LNG vaporization into data-center cooling load.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Unconfirmed&lt;/td&gt;
 &lt;td&gt;Actual commercial contract, target PUE, realized PUE improvement, tariff, capex, payback, AWS/SK economics and SK Gas margin attribution remain undisclosed.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;So SK Gas is not a pure &amp;ldquo;cold-energy theme stock.&amp;rdquo; The cleaner thesis is &lt;strong&gt;undervalued core energy-infrastructure earnings plus an under-modeled AI data-center cooling option&lt;/strong&gt;.&lt;/p&gt;
&lt;h2 id="1-why-ulsan-matters"&gt;1. Why Ulsan Matters
&lt;/h2&gt;&lt;p&gt;LNG cold energy is hard to move over long distances. The key condition is physical proximity between the LNG terminal and the data center. Ulsan may satisfy that condition.&lt;/p&gt;
&lt;p&gt;KET, a joint venture between Korea National Oil Corporation and SK Gas, operates an energy storage terminal in Ulsan North Port and began commercial operation of its oil and LNG terminal in 2024.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt; SK Telecom has also announced that it is working with AWS on an AI data center in Ulsan, using SK Group land and energy supply capabilities.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;SK Gas also has Ulsan GPS, an LNG/LPG dual-fuel combined-cycle power plant. SK Gas sold a 49% stake in Ulsan GPS for KRW 1.2242 trillion while retaining 51% control.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt; This makes Ulsan a combined power, LNG terminal, cooling and AI data-center campus candidate rather than a generic LNG story.&lt;/p&gt;
&lt;h2 id="2-the-cooling-share-should-not-be-overstated"&gt;2. The Cooling Share Should Not Be Overstated
&lt;/h2&gt;&lt;p&gt;Cooling is important, but the common claim that cooling is 30-40% of all data-center power should not be applied to all modern hyperscale facilities. IEA data show that cooling and environmental control may be around 7% in efficient hyperscale data centers, while it can exceed 30% in less efficient enterprise data centers.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;That distinction matters. LNG cold energy may lower electric chiller load, but the actual PUE benefit depends on facility design, rack density, cooling loop efficiency, pumps, heat exchangers, backup chillers and utilization matching.&lt;/p&gt;
&lt;h2 id="3-how-the-system-would-work"&gt;3. How The System Would Work
&lt;/h2&gt;&lt;p&gt;The simplified chain is:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Step&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;LNG storage&lt;/td&gt;
 &lt;td&gt;LNG is stored near -162°C.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Vaporization&lt;/td&gt;
 &lt;td&gt;LNG must be warmed and vaporized before use. Cold energy is released in the process.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Recovery&lt;/td&gt;
 &lt;td&gt;Instead of discarding that cold energy into seawater, a heat-exchange system captures it.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Coolant loop&lt;/td&gt;
 &lt;td&gt;Refrigerant or brine loops move the cooling effect safely toward the data-center cooling plant.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Data-center cooling&lt;/td&gt;
 &lt;td&gt;Electric chiller load can be reduced.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Redundancy&lt;/td&gt;
 &lt;td&gt;Mission-critical data centers still need backup chillers, pumps and control systems.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The value is real only if cold recovery, reliability, capex and contractual economics work together.&lt;/p&gt;
&lt;h2 id="4-investment-idea-one-a-cold-energy-toll-road"&gt;4. Investment Idea One: A Cold-Energy Toll Road
&lt;/h2&gt;&lt;p&gt;If commercialized, SK Gas and KET could turn otherwise wasted cold energy into a cooling service.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Variable&lt;/th&gt;
 &lt;th&gt;Driver&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Price&lt;/td&gt;
 &lt;td&gt;Avoided chiller electricity cost, water cost and carbon cost can partly become a cold-energy tariff.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Quantity&lt;/td&gt;
 &lt;td&gt;Ulsan AI data-center MW scale, rack density, LNG vaporization volume and cold-energy utilization rate.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cost&lt;/td&gt;
 &lt;td&gt;Heat exchangers, coolant or brine loops, pipes, pumps, control systems, backup chillers and maintenance.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The upside is a potential infrastructure-style toll road without directly buying GPUs. The missing proof is a long-term cold-energy supply contract, minimum volume, tariff, capex payback, target PUE and SK Gas margin attribution.&lt;/p&gt;
&lt;h2 id="5-investment-idea-two-firm-power-call-option"&gt;5. Investment Idea Two: Firm Power Call Option
&lt;/h2&gt;&lt;p&gt;The cold-energy idea sits on top of a more basic thesis. AI data centers need firm power. They do not only need cheap power. They need power that does not stop.&lt;/p&gt;
&lt;p&gt;Ulsan GPS gives SK Gas a firm-power option. The key variables are power-market price, capacity payments, LNG/LPG fuel spreads, plant utilization and SK Gas&amp;rsquo;s retained 51% exposure after the stake sale.&lt;/p&gt;
&lt;p&gt;The KRW 1.2242 trillion price for 49% of Ulsan GPS implies a simple 100% value of roughly KRW 2.5 trillion and a retained 51% value near KRW 1.27 trillion. This is only a simple cross-check, not a full valuation, but it supports the idea that the market may still be applying a large discount to SK Gas&amp;rsquo;s energy infrastructure assets.&lt;/p&gt;
&lt;h2 id="6-what-is-still-missing"&gt;6. What Is Still Missing
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Target PUE for the Ulsan AI data center&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LNG cold-energy share of total cooling load&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cold-energy supply tariff&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Long-term contract structure&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cold-recovery capex and payback&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AWS/SK economic allocation&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Gas EBITDA or net-income contribution&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Proprietary technology or patent scope&lt;/td&gt;
 &lt;td&gt;Not disclosed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This table is the core of the analysis. The technical logic is credible, but the numbers required for an EPS model are not public yet.&lt;/p&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;SK Gas&amp;rsquo;s LNG cold-energy option is not empty. It has technical logic, a relevant Ulsan location and a broader SK-AWS AI data-center context. But it is not yet a confirmed earnings driver.&lt;/p&gt;
&lt;p&gt;The right investment stance is Watchlist / conditional buy. SK Gas should be read as an undervalued energy infrastructure company with an AI data-center cooling option. The cold-energy thesis becomes investable only when the market sees a supply contract, PUE improvement, capex payback and SK Gas margin capture.&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://devocean.sk.com/blog/techBoardDetail.do?ID=167883&amp;amp;boardType=techBlog" target="_blank" rel="noopener"
 &gt;SK DevOcean, AI Data Center LNG Cold Energy Solution&lt;/a&gt;&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.ket.co.kr/intro/summary/index.jsp" target="_blank" rel="noopener"
 &gt;KET company overview&lt;/a&gt;&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://news.sktelecom.com/213527" target="_blank" rel="noopener"
 &gt;SK Telecom Newsroom, SKT and AWS Ulsan AI data-center plan&lt;/a&gt;&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://v.daum.net/v/20260630160302257" target="_blank" rel="noopener"
 &gt;The Fact, SK Gas sells 49% of Ulsan GPS for KRW 1.2242 trillion&lt;/a&gt;&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.iea.org/reports/energy-and-ai" target="_blank" rel="noopener"
 &gt;IEA, Energy and AI&lt;/a&gt;&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>AI Data Center Power Bottleneck Map: After GnCenergy, VinaTech and Seojin System, Who Is Next?</title><link>https://koreainvestinsights.com/post/ai-datacenter-power-bottleneck-korea-value-chain-screen-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 13:40:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/ai-datacenter-power-bottleneck-korea-value-chain-screen-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;July 23, 2026 update
Our follow-up &lt;a class="link" href="https://koreainvestinsights.com/post/korea-ai-datacenter-beneficiaries-power-cooling-operator-flow-ranking-2026-07-23/" &gt;re-ranks 19 Korean AI data-center stocks&lt;/a&gt; using direct exposure, current flow, valuation and entry price. GnCenergy and LS ELECTRIC lead on directness, but the present risk-adjusted order changes to LG Electronics, Iljin Electric, Samsung SDS, GST and SK Gas.&lt;/p&gt;

 &lt;/blockquote&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Related context
This post follows &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-capex-5p3t-korea-power-substrate-storage-bottleneck-2026-06-05/" &gt;AI data-center CapEx and Korean bottlenecks&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/gnc-energy-ai-datacenter-backup-generator-sustainable-rally-2026-07-04/" &gt;GnCenergy and AI data-center backup power&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/vinatech-bloom-ai-datacenter-supercapacitor-power-buffer-2026-07-04/" &gt;VinaTech and Bloom Energy power buffering&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-gas-sk-oceanplant-ai-datacenter-power-second-line-2026-07-04/" &gt;SK Gas vs SK Oceanplant&lt;/a&gt;, and the &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;H1 2026 AI infrastructure postmortem&lt;/a&gt;. The relevant hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-equipment-ip-hub/" &gt;Korean Semiconductor Value Chain Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;AI data-center power is not one trade. It breaks into several layers: &lt;strong&gt;generation and backup power, fuel cells and distributed power, ESS and UPS, power conversion and transformers, cables and bus ducts, cooling and EPC&lt;/strong&gt;. Each layer maps to a different listed Korean supplier.&lt;/p&gt;
&lt;p&gt;The first screen is straightforward.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Conclusion&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Strong direct evidence&lt;/td&gt;
 &lt;td&gt;GnCenergy, VinaTech and Seojin System have the clearest supply evidence. GnCenergy maps to Korean data-center backup generators. VinaTech maps to Bloom Energy-linked supercapacitors. Seojin System maps to ESS enclosures and racks.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Evidence still weak&lt;/td&gt;
 &lt;td&gt;HD Hyundai Marine Engine and Hanwha Engine have optionality in data-center engine power, but listed-entity-level supply evidence remains weak. Separate group-level orders from the listed company&amp;rsquo;s earnings capture.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Underpriced sub-layers&lt;/td&gt;
 &lt;td&gt;Backup generators, supercapacitors, ESS enclosures, internal power conversion, cable bus and bus ducts may still be less systematically priced than large transformer winners.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The first diligence priority is &lt;strong&gt;GnCenergy, VinaTech, Seojin System, Sanil Electric, Iljin Electric, Gaon Cable and SK Gas&lt;/strong&gt;. Large electrical-equipment names are high quality, but a lot of the rerating is already known. The question now is which sub-bottlenecks turn into repeat revenue.&lt;/p&gt;
&lt;h2 id="1-why-power-is-the-bottleneck"&gt;1. Why Power Is The Bottleneck
&lt;/h2&gt;&lt;p&gt;The IEA&amp;rsquo;s &lt;i&gt;Energy and AI&lt;/i&gt; report sees global data-center electricity consumption rising from about 415 TWh in 2024 to around 945 TWh by 2030. In the United States, data centers may account for nearly half of electricity-demand growth through 2030.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The issue is not only generation. Grid equipment has become a bottleneck. The IEA notes that procurement can now take two to three years for cables and up to four years for large power transformers, about twice as long as in 2021.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;For investors, this means the opportunity set is wider than transformer makers. Backup power, ESS, UPS, internal distribution, cable systems and cooling can all become bottlenecks when AI data centers scale.&lt;/p&gt;
&lt;h2 id="2-value-chain-map"&gt;2. Value Chain Map
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Products and services&lt;/th&gt;
 &lt;th&gt;Bottleneck&lt;/th&gt;
 &lt;th&gt;Listed Korean candidates&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Backup and onsite generation&lt;/td&gt;
 &lt;td&gt;Diesel and gas generators, medium-speed engines, LNG/LPG power&lt;/td&gt;
 &lt;td&gt;Reliability, tier certification, delivery, fuel supply&lt;/td&gt;
 &lt;td&gt;GnCenergy, SK Gas, Doosan Enerbility, Hanwha Engine, HD Hyundai Marine Engine, SGC Energy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Fuel cells and distributed power&lt;/td&gt;
 &lt;td&gt;SOFC, PEMFC, fuel-cell BOP, supercapacitors&lt;/td&gt;
 &lt;td&gt;Load response, heat use, PPA structure, auxiliary power&lt;/td&gt;
 &lt;td&gt;VinaTech, Doosan Fuel Cell, SK Eternix, S-Fuelcell, Bumhan Fuel Cell&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ESS, UPS and BBU&lt;/td&gt;
 &lt;td&gt;LFP/NCA batteries, racks, enclosures, UPS, supercapacitors&lt;/td&gt;
 &lt;td&gt;Fire safety, customer qualification, U.S. manufacturing, container design&lt;/td&gt;
 &lt;td&gt;Samsung SDI, LG Energy Solution, SK Innovation/SK On, Seojin System, VinaTech&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power conversion and distribution&lt;/td&gt;
 &lt;td&gt;PCS, DC solutions, switchgear, breakers, special transformers&lt;/td&gt;
 &lt;td&gt;Hyperscaler approval, density, efficiency&lt;/td&gt;
 &lt;td&gt;LS ELECTRIC, Sanil Electric, HD Hyundai Electric, Hyosung Heavy Industries&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Grid equipment&lt;/td&gt;
 &lt;td&gt;Power transformers, GIS, reactors, breakers&lt;/td&gt;
 &lt;td&gt;Production slots, North American certification, lead times&lt;/td&gt;
 &lt;td&gt;HD Hyundai Electric, Hyosung Heavy Industries, LS ELECTRIC, Iljin Electric, Cheryong Electric&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cables and bus ducts&lt;/td&gt;
 &lt;td&gt;High-voltage cables, cable bus, bus duct, internal wiring&lt;/td&gt;
 &lt;td&gt;Copper, certification, installation&lt;/td&gt;
 &lt;td&gt;Gaon Cable, LS Eco Energy, Taihan Cable, Iljin Electric&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cooling and mechanical&lt;/td&gt;
 &lt;td&gt;HVAC, immersion cooling, CDU, cleanroom&lt;/td&gt;
 &lt;td&gt;PUE, heat density, coolant and refrigerant&lt;/td&gt;
 &lt;td&gt;Shinsung E&amp;amp;G, GST, LG Electronics, S-Oil&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EPC and operations&lt;/td&gt;
 &lt;td&gt;Data-center construction, electrical work, power operations&lt;/td&gt;
 &lt;td&gt;Permitting, schedule, integrated design&lt;/td&gt;
 &lt;td&gt;Samsung C&amp;amp;T, HanmiGlobal, Hyundai E&amp;amp;C, DL E&amp;amp;C&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="3-evidence-based-ranking"&gt;3. Evidence-Based Ranking
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Ticker&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;GnCenergy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;119850&lt;/td&gt;
 &lt;td&gt;Watchlist / conditional Buy candidate&lt;/td&gt;
 &lt;td&gt;The most direct Korean data-center backup-generator supplier. Evidence includes Naver, kt cloud, LG CNS and other IDC projects.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;VinaTech&lt;/td&gt;
 &lt;td style="text-align: right"&gt;126340&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;A direct Bloom Energy AI data-center supercapacitor supply contract, with overseas revenue around 90% according to the company.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Seojin System&lt;/td&gt;
 &lt;td style="text-align: right"&gt;178320&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;ESS enclosures, racks and containers. Samsung SDI America and U.S. ESS supply-chain exposure are the key evidence points.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;Sanil Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;062040&lt;/td&gt;
 &lt;td&gt;Wait / Watchlist&lt;/td&gt;
 &lt;td&gt;Strong Bloom Energy special-transformer evidence, but valuation may already reflect the theme.&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;LS ELECTRIC&lt;/td&gt;
 &lt;td style="text-align: right"&gt;010120&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;A complete power-distribution, DC, ESS and transformer platform. The issue is price.&lt;sup id="fnref:7"&gt;&lt;a href="#fn:7" class="footnote-ref" role="doc-noteref"&gt;7&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;HD Hyundai Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;267260&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;Very strong North American data-center equipment evidence, but the large-cap rerating is already advanced.&lt;sup id="fnref:8"&gt;&lt;a href="#fn:8" class="footnote-ref" role="doc-noteref"&gt;8&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;Iljin Electric&lt;/td&gt;
 &lt;td style="text-align: right"&gt;103590&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;Both heavy electrical equipment and cables. The Malaysia data-center cable order is direct evidence.&lt;sup id="fnref:9"&gt;&lt;a href="#fn:9" class="footnote-ref" role="doc-noteref"&gt;9&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;8&lt;/td&gt;
 &lt;td&gt;Gaon Cable&lt;/td&gt;
 &lt;td style="text-align: right"&gt;000500&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;U.S. AI data-center cable supply and U.S. capacity expansion are direct signals.&lt;sup id="fnref:10"&gt;&lt;a href="#fn:10" class="footnote-ref" role="doc-noteref"&gt;10&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;9&lt;/td&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td style="text-align: right"&gt;018670&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;Ulsan GPS and SK&amp;rsquo;s Ulsan AI data-center power architecture make it a low-valuation infrastructure candidate.&lt;sup id="fnref:11"&gt;&lt;a href="#fn:11" class="footnote-ref" role="doc-noteref"&gt;11&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td&gt;Samsung SDI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;006400&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;Data-center ESS, UPS and BBU exposure is real, but EV batteries still dominate the equity story.&lt;sup id="fnref:12"&gt;&lt;a href="#fn:12" class="footnote-ref" role="doc-noteref"&gt;12&lt;/a&gt;&lt;/sup&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="4-main-ideas"&gt;4. Main Ideas
&lt;/h2&gt;&lt;h3 id="gncenergy-backup-power-pure-play"&gt;GnCenergy: Backup power pure-play
&lt;/h3&gt;&lt;p&gt;GnCenergy is the clearest direct supplier in this screen. AI data centers cannot tolerate downtime, so backup generation is not an old-economy afterthought. It is the last reliability layer behind compute capacity. The risk is project lumpiness, margin volatility and potential environmental regulation against diesel backup generators.&lt;/p&gt;
&lt;h3 id="vinatech-and-sanil-electric-bloom-energy-supply-chain"&gt;VinaTech and Sanil Electric: Bloom Energy supply chain
&lt;/h3&gt;&lt;p&gt;The Bloom Energy supply chain is not only about fuel cells. If onsite SOFC power becomes more common for AI data centers, power smoothing and conversion hardware also matter. VinaTech supplies supercapacitors for AI data centers through Bloom Energy, while Sanil Electric has Bloom-linked special-transformer evidence. The key test is whether these become repeat orders rather than one-off projects.&lt;/p&gt;
&lt;h3 id="seojin-system-ess-picks-and-shovels"&gt;Seojin System: ESS picks-and-shovels
&lt;/h3&gt;&lt;p&gt;AI data centers need ESS, UPS and BBU systems, but the cell supplier is not the only beneficiary. Enclosures, racks, containers, thermal design and fire-safety structures can be manufacturing bottlenecks. Seojin System is interesting for that reason. The issue is not revenue growth alone. Working capital, net debt and cash-flow quality must also improve.&lt;/p&gt;
&lt;h3 id="cables-and-bus-ducts"&gt;Cables and bus ducts
&lt;/h3&gt;&lt;p&gt;Iljin Electric, Gaon Cable and LS Eco Energy sit in a less glamorous but important layer. Data centers need not only external grid connections but also dense internal power distribution. Cable bus, bus ducts and certified high-voltage cables can matter more as rack density rises. Copper pass-through and margin discipline are the main risks.&lt;/p&gt;
&lt;h3 id="energy-campus"&gt;Energy campus
&lt;/h3&gt;&lt;p&gt;SK Gas, SK Eternix and Doosan Fuel Cell map to a broader energy-campus thesis. Data centers may value secured power more than cheap power. LNG/LPG generation, renewables, fuel cells and ESS may be bundled into site-level power solutions. SK Gas looks the most investable in this group because Ulsan GPS and the SK Ulsan AI data-center structure are more concrete, while SK Eternix and Doosan Fuel Cell still need direct PPA or commercial-order evidence.&lt;/p&gt;
&lt;h3 id="engine-optionality"&gt;Engine optionality
&lt;/h3&gt;&lt;p&gt;HD Hyundai Heavy Industries&amp;rsquo; U.S. data-center HiMSEN engine order proves that medium-speed engines can be part of the onsite-power solution.&lt;sup id="fnref:13"&gt;&lt;a href="#fn:13" class="footnote-ref" role="doc-noteref"&gt;13&lt;/a&gt;&lt;/sup&gt; But that does not automatically make HD Hyundai Marine Engine or Hanwha Engine direct beneficiaries. Listed-entity-level orders, production roles and margin capture must be confirmed first.&lt;/p&gt;
&lt;h2 id="5-entry-catalysts-and-invalidation"&gt;5. Entry, Catalysts and Invalidation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Group&lt;/th&gt;
 &lt;th&gt;Entry condition&lt;/th&gt;
 &lt;th&gt;Catalyst&lt;/th&gt;
 &lt;th&gt;Invalidation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;GnCenergy&lt;/td&gt;
 &lt;td&gt;Data-center backlog grows, margins hold, valuation stays reasonable&lt;/td&gt;
 &lt;td&gt;LG Uplus Paju AIDC follow-on, Naver or telecom data-center expansion&lt;/td&gt;
 &lt;td&gt;Two quarters of slowing orders, margin collapse, diesel backup demand weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;VinaTech&lt;/td&gt;
 &lt;td&gt;Bloom revenue recognition, module supply, visible path to profitability&lt;/td&gt;
 &lt;td&gt;Bloom data-center power expansion, follow-on purchase orders&lt;/td&gt;
 &lt;td&gt;One-off Bloom order, ongoing losses, no customer diversification&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Seojin System&lt;/td&gt;
 &lt;td&gt;ESS enclosure growth and operating cash-flow improvement together&lt;/td&gt;
 &lt;td&gt;Samsung SDI America, Fluence or Powin-linked ESS expansion&lt;/td&gt;
 &lt;td&gt;Working-capital deterioration, debt pressure, low-margin OEM trap&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sanil Electric&lt;/td&gt;
 &lt;td&gt;New customers beyond Bloom, margin stability, valuation reset&lt;/td&gt;
 &lt;td&gt;Follow-on special-transformer orders&lt;/td&gt;
 &lt;td&gt;One-off Bloom order, margin compression, customer concentration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Iljin Electric and Gaon Cable&lt;/td&gt;
 &lt;td&gt;Repeat data-center cable orders and copper pass-through&lt;/td&gt;
 &lt;td&gt;U.S. and Southeast Asia data-center cable wins&lt;/td&gt;
 &lt;td&gt;Copper spike, certification or installation delays, commodity cable mix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td&gt;More detail on SK/AWS Ulsan power economics&lt;/td&gt;
 &lt;td&gt;Ulsan AI data-center power contract, UGPS stable operation&lt;/td&gt;
 &lt;td&gt;Gas spread pressure, power-market regulation, weak PPA economics&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hanwha Engine and HD Hyundai Marine Engine&lt;/td&gt;
 &lt;td&gt;Listed-entity data-center engine order&lt;/td&gt;
 &lt;td&gt;North American onsite-generation orders&lt;/td&gt;
 &lt;td&gt;No order, unclear group margin capture, stock rises only on expectation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;The strongest screen today is &lt;strong&gt;GnCenergy, VinaTech and Seojin System&lt;/strong&gt;. They are not the obvious large transformer winners. They sit in lower-level bottlenecks inside data-center power infrastructure.&lt;/p&gt;
&lt;p&gt;Sanil Electric, Iljin Electric and Gaon Cable have direct order evidence, but valuation and margin proof matter. SK Gas is not a pure-play, but Ulsan GPS and SK&amp;rsquo;s AI data-center power structure make it a low-valuation infrastructure candidate. Samsung SDI and LG Energy Solution have data-center ESS exposure, but EV battery cycles still dominate their equities.&lt;/p&gt;
&lt;p&gt;Hanwha Engine and HD Hyundai Marine Engine remain stories ahead of evidence. Engine-based onsite power is a real theme, but listed-company revenue and margin capture must be proven.&lt;/p&gt;
&lt;p&gt;The practical stance is not “buy every power stock.” It is to monitor repeat contracts, backlog quality, margin capture and cash flow, then enter selectively when the evidence improves.&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.iea.org/reports/energy-and-ai/executive-summary" target="_blank" rel="noopener"
 &gt;IEA, Energy and AI Executive Summary&lt;/a&gt;&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.iea.org/news/rising-component-prices-and-supply-chain-pressures-are-hindering-the-development-of-transmission-grid-infrastructure" target="_blank" rel="noopener"
 &gt;IEA, Rising component prices and supply chain pressures are hindering transmission grid infrastructure&lt;/a&gt;&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.electimes.com/news/articleView.html?idxno=358993" target="_blank" rel="noopener"
 &gt;Electimes, GnCenergy data-center backup-generator order coverage&lt;/a&gt;&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.vinatech.com/en/sub/pr/news.php?bid=2&amp;amp;idx=3467&amp;amp;mode=view" target="_blank" rel="noopener"
 &gt;VINA Tech, Target AI Data Center and Sales by 2030 Trillion&lt;/a&gt;&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.thelec.net/news/articleView.html?idxno=11780" target="_blank" rel="noopener"
 &gt;The Elec, Seojin System ESS enclosure manufacturing agreement&lt;/a&gt;&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;&lt;a class="link" href="https://en.sedaily.com/finance/2026/04/30/sanil-electric-surges-25-percent-on-503-billion-won-bloom" target="_blank" rel="noopener"
 &gt;Seoul Economic Daily, Sanil Electric Bloom Energy order&lt;/a&gt;&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:7"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.electimes.com/news/articleView.html?idxno=367534" target="_blank" rel="noopener"
 &gt;Electimes, LS ELECTRIC North America data-center power infrastructure&lt;/a&gt;&amp;#160;&lt;a href="#fnref:7" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:8"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.koreajoongangdaily.com/business/hd-hyundai-electric-seals-721-million-deal-with-big-tech-to-power-north-america-data-center/12751608" target="_blank" rel="noopener"
 &gt;Korea JoongAng Daily, HD Hyundai Electric North America data-center deal&lt;/a&gt;&amp;#160;&lt;a href="#fnref:8" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:9"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.bloter.net/news/articleView.html?idxno=664663" target="_blank" rel="noopener"
 &gt;Bloter, Iljin Electric Malaysia data-center power-grid order&lt;/a&gt;&amp;#160;&lt;a href="#fnref:9" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:10"&gt;
&lt;p&gt;&lt;a class="link" href="https://zdnet.co.kr/view/?no=20260604084555" target="_blank" rel="noopener"
 &gt;ZDNet Korea, Gaon Cable U.S. AI data-center cable market&lt;/a&gt;&amp;#160;&lt;a href="#fnref:10" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:11"&gt;
&lt;p&gt;&lt;a class="link" href="https://eng.skgas.co.kr/Business/gas_energy.html" target="_blank" rel="noopener"
 &gt;SK gas, Gas Energy business&lt;/a&gt;&amp;#160;&lt;a href="#fnref:11" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:12"&gt;
&lt;p&gt;&lt;a class="link" href="https://news.samsungsdi.com/ko/articleView?seq=317" target="_blank" rel="noopener"
 &gt;Samsung SDI Newsroom, ESS batteries for data centers and power systems&lt;/a&gt;&amp;#160;&lt;a href="#fnref:12" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:13"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.hyundai-engine.com/en/media/newsdetail/457" target="_blank" rel="noopener"
 &gt;HD Hyundai Heavy Industries Engine &amp;amp; Machinery, U.S. data center HiMSEN engine order&lt;/a&gt;&amp;#160;&lt;a href="#fnref:13" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>VinaTech and Bloom Energy: Who Buffers AI Data Center Power Shocks?</title><link>https://koreainvestinsights.com/post/vinatech-bloom-ai-datacenter-supercapacitor-power-buffer-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 10:30:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/vinatech-bloom-ai-datacenter-supercapacitor-power-buffer-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context
This post follows &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-capex-5p3t-korea-power-substrate-storage-bottleneck-2026-06-05/" &gt;AI Data Center CapEx and Korean bottlenecks&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/mlcc-silicon-capacitor-ai-package-power-integrity-2026-05-22/" &gt;MLCC and silicon capacitors&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/ai-server-passive-components-bottleneck-samsung-electro-mechanics-2026-05-26/" &gt;AI server passive-component bottlenecks&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/warsh-fed-expensive-money-era-forward-guidance-ai-infra-2026-06-19/" &gt;the expensive-money AI infrastructure frame&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;the H1 2026 AI infrastructure postmortem&lt;/a&gt;. The relevant hubs are the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-equipment-ip-hub/" &gt;Korean Semiconductor Value Chain Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;VinaTech is not primarily a &amp;ldquo;company that stores a lot of electricity.&amp;rdquo; The better frame is &lt;strong&gt;a company that can absorb transient power shocks between AI data centers and Bloom Energy&amp;rsquo;s SOFC systems&lt;/strong&gt;. If Bloom is the on-site generator, VinaTech&amp;rsquo;s supercapacitor system is the power buffer between that generator and AI servers.&lt;/p&gt;
&lt;p&gt;VinaTech signed a data-center supercapacitor-system supply contract with Bloom Energy. The contract value is KRW 41.215B, equal to 50.12% of VinaTech&amp;rsquo;s 2025 consolidated revenue of KRW 82.229B. The contract period runs from June 30, 2026 to April 10, 2027.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt; For a small company, that is large enough to change the revenue base.&lt;/p&gt;
&lt;p&gt;The alpha is not the KRW 41.2B contract itself. The real question is whether this is a one-off Bloom project or the beginning of repeat standard content inside Bloom&amp;rsquo;s SOFC data-center package. If follow-on purchase orders and system margins appear, VinaTech can be reclassified from a supercapacitor-cell supplier into an AI data-center power-stability system supplier.&lt;/p&gt;
&lt;p&gt;The technology moat is medium-high. Supercapacitor cells have global competitors. But customer qualification inside Bloom&amp;rsquo;s data-center power architecture, plus the shift from cell supply to module, control and software packaging, is a higher barrier than selling a discrete component.&lt;/p&gt;
&lt;p&gt;The current stance is &lt;strong&gt;Watch, with a possible upgrade to conditional Buy&lt;/strong&gt;. At the July 3, 2026 close of KRW 84,400, implied market capitalization is about KRW 606B. Volatility is high: the stock hit limit-up on July 1, fell 20.1% on July 2, then fell another 1.9% on July 3. Chasing the first contract is less attractive than waiting for Bloom repeat orders, system-margin proof and customer diversification.&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Bottom line&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 VinaTech is not an AI data-center power-generation stock. It is a possible power-quality and transient-load-buffer stock. If Bloom's SOFC units become more common in AI data centers, the key question is whether VinaTech's supercapacitor system becomes repeat standard content.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="0-analysis-setup"&gt;0. Analysis Setup
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Detail&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Company&lt;/td&gt;
 &lt;td&gt;VinaTech&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ticker&lt;/td&gt;
 &lt;td&gt;126340 / KOSDAQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Analysis date&lt;/td&gt;
 &lt;td&gt;July 4, 2026 KST&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price basis&lt;/td&gt;
 &lt;td&gt;July 3, 2026 close of KRW 84,400, pykrx and local Kiwoom DB&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core event&lt;/td&gt;
 &lt;td&gt;KRW 41.2B Bloom Energy data-center supercapacitor-system supply contract&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core question&lt;/td&gt;
 &lt;td&gt;Can VinaTech become a repeat bottleneck supplier in the AI data-center power chain?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stance&lt;/td&gt;
 &lt;td&gt;Watch, with possible upgrade to conditional Buy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This post starts with business position rather than price. Reading VinaTech as an AI data-center power-generation stock is wrong. Bloom Energy&amp;rsquo;s SOFC, or solid oxide fuel cell, generates the power. VinaTech&amp;rsquo;s job is to buffer the shock when AI servers suddenly pull or release power.&lt;/p&gt;
&lt;h2 id="1-the-bottleneck-is-power-speed-not-just-power-volume"&gt;1. The Bottleneck Is Power Speed, Not Just Power Volume
&lt;/h2&gt;&lt;p&gt;AI data-center power has two different problems.&lt;/p&gt;
&lt;p&gt;The first is power volume. The data center must secure enough electricity through grid interconnection, power plants, PPAs, nuclear, gas, fuel cells or BESS.&lt;/p&gt;
&lt;p&gt;The second is power speed and quality. Voltage and current need to remain stable when AI servers abruptly pull more power or drop load. More generation alone does not solve this.&lt;/p&gt;
&lt;p&gt;NVIDIA explains that AI training differs from traditional data-center workloads because thousands of GPUs operate in lockstep. The load does not naturally average out. Workloads shift rapidly between idle and high-power states, creating grid-level fluctuations. NVIDIA says the GB300 NVL72 power shelf includes energy storage that can smooth power spikes and reduce peak grid demand by up to 30%.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;In plain language, a traditional data center is like an apartment block where many residents use electricity at different times. AI training is closer to a building where thousands of heavy elevators start and stop together. The grid problem is not only that the building uses a lot of power. It is that it pulls and releases power abruptly.&lt;/p&gt;
&lt;p&gt;VinaTech sits in that second bottleneck: transient load, power quality, voltage movement and current shock.&lt;/p&gt;
&lt;h2 id="2-bloom-sofc-and-vinatech-supercapacitors-do-different-jobs"&gt;2. Bloom SOFC and VinaTech Supercapacitors Do Different Jobs
&lt;/h2&gt;&lt;p&gt;Bloom Energy announced a strategic partnership with CoreWeave to deploy fuel cells for on-site power at a high-performance AI data center in Volo, Illinois.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt; The point is that AI data centers are trying to secure on-site power rather than wait indefinitely for grid interconnection.&lt;/p&gt;
&lt;p&gt;But SOFC is closer to a steady generator than an infinitely fast load-following device.&lt;/p&gt;
&lt;p&gt;Research on hybrid SOFC-based DC microgrids shows the issue. The RSC paper explains that a stand-alone SOFC system has difficulty with fast load following, and that battery plus supercapacitor storage can help cover transient power and reduce fuel-starvation risk. Supercapacitors provide high current during sudden load increases, reducing stress on the SOFC system.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The division of labor is simple.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Simple analogy&lt;/th&gt;
 &lt;th&gt;Power-system role&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bloom SOFC&lt;/td&gt;
 &lt;td&gt;On-site generator&lt;/td&gt;
 &lt;td&gt;Stable long-duration power supply&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Battery or BESS&lt;/td&gt;
 &lt;td&gt;Large water tank&lt;/td&gt;
 &lt;td&gt;Minute-to-hour energy storage and backup&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;VinaTech supercapacitor&lt;/td&gt;
 &lt;td&gt;Shock absorber&lt;/td&gt;
 &lt;td&gt;Millisecond-to-second peak buffering&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PCS, BMS, control software&lt;/td&gt;
 &lt;td&gt;Valves and automatic controls&lt;/td&gt;
 &lt;td&gt;Voltage, current, charge-discharge and protection logic&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Supercapacitors store less total energy than batteries. But they charge and discharge quickly and have strong cycle life. They are not best understood as long-duration backup power. They are short-duration shock absorbers.&lt;/p&gt;
&lt;p&gt;For Bloom SOFC power to support AI data centers, the output must be smoothed into a form AI servers can consume reliably. That is where VinaTech matters.&lt;/p&gt;
&lt;h2 id="3-did-vinatech-actually-enter-the-bloom-chain"&gt;3. Did VinaTech Actually Enter the Bloom Chain?
&lt;/h2&gt;&lt;p&gt;The confirmed fact is strong.&lt;/p&gt;
&lt;p&gt;VinaTech signed a data-center supercapacitor-system supply contract with Bloom Energy. The contract is KRW 41.215B, equal to 50.12% of 2025 consolidated revenue. The term runs from June 30, 2026 to April 10, 2027. There is no advance payment, and supply will be handled through outsourced production via an overseas subsidiary.&lt;sup id="fnref1:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The important change is from &amp;ldquo;cell supply&amp;rdquo; to &amp;ldquo;system supply.&amp;rdquo; The Bell reported that this is VinaTech&amp;rsquo;s first case of directly supplying an entire supercapacitor system for data centers; prior data-center business was mainly supercapacitor cells.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;VinaTech&amp;rsquo;s own materials point in the same direction. The company says it signed an AI data-center supercapacitor supply contract with Bloom Energy and plans to shift in the first half of 2026 from individual cell sales to modular products. It also describes integrated packages that include PCB and software.&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The product-status change matters more than the headline revenue.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Old frame: supercapacitor-cell company
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Possible new frame: AI data-center power-stability system supplier
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;However, this is not yet a long-duration monopoly. We do not know whether the Bloom contract is tied to one project or whether it is becoming standard content in Bloom&amp;rsquo;s data-center power package.&lt;/p&gt;
&lt;h2 id="4-what-is-and-is-not-confirmed-about-meta"&gt;4. What Is and Is Not Confirmed About Meta
&lt;/h2&gt;&lt;p&gt;This point requires discipline.&lt;/p&gt;
&lt;p&gt;There is no confirmed direct VinaTech-to-Meta supply contract. There is also no confirmed direct Meta-to-VinaTech relationship. The confirmed chain is indirect.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Meta → CoreWeave → Bloom Energy → VinaTech
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;CoreWeave announced a roughly $21B long-term AI cloud capacity expansion with Meta in April 2026.&lt;sup id="fnref:7"&gt;&lt;a href="#fn:7" class="footnote-ref" role="doc-noteref"&gt;7&lt;/a&gt;&lt;/sup&gt; Bloom announced on-site fuel-cell deployment for CoreWeave&amp;rsquo;s AI data center.&lt;sup id="fnref1:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt; VinaTech signed a data-center supercapacitor-system contract with Bloom.&lt;sup id="fnref2:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;So VinaTech is not a &amp;ldquo;Meta direct supplier.&amp;rdquo; The cleaner investment phrase is:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;a transient-power-stability supplier inside Bloom systems when AI data-center power bottlenecks lead to Bloom SOFC adoption&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;Using Meta&amp;rsquo;s logo makes the story sound larger, but it weakens the investment analysis. The real alpha is whether VinaTech gets repeat revenue inside Bloom&amp;rsquo;s system.&lt;/p&gt;
&lt;h2 id="5-vinatechs-real-bottleneck-power-rate-not-energy-amount"&gt;5. VinaTech&amp;rsquo;s Real Bottleneck: Power Rate, Not Energy Amount
&lt;/h2&gt;&lt;p&gt;Investors need to separate energy from power.&lt;/p&gt;
&lt;p&gt;Energy (kWh) is the size of the water tank. Power (kW) is the instantaneous force from the faucet. AI data centers need more than a large tank. They need pipes that do not shake when the faucet opens and closes abruptly.&lt;/p&gt;
&lt;p&gt;VinaTech&amp;rsquo;s supercapacitor system is a transient pressure regulator. When AI servers suddenly need current, the supercapacitor discharges. When server load drops, it absorbs excess energy. Bloom&amp;rsquo;s SOFC then does not need to ramp aggressively up and down.&lt;/p&gt;
&lt;p&gt;The P, Q and C framework looks like this.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Driver&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;th&gt;What to verify&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P, price&lt;/td&gt;
 &lt;td&gt;System products should carry higher ASP than cells&lt;/td&gt;
 &lt;td&gt;Does PCB, software and control packaging lift ASP?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q, volume&lt;/td&gt;
 &lt;td&gt;Bloom SOFC data-center projects must scale&lt;/td&gt;
 &lt;td&gt;Does VinaTech content attach to each new Bloom project?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C, cost&lt;/td&gt;
 &lt;td&gt;System supply adds outsourcing, quality and control cost&lt;/td&gt;
 &lt;td&gt;Does revenue growth translate into operating margin?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The market can miss the shift by seeing VinaTech only as a supercapacitor-cell company. The contract points to a subsystem supplier role. But the market can also over-read the news. One system contract does not prove standardization. Repeat orders must confirm it.&lt;/p&gt;
&lt;h2 id="6-technology-moat-application-qualification-matters-more-than-the-cell"&gt;6. Technology Moat: Application Qualification Matters More Than The Cell
&lt;/h2&gt;&lt;p&gt;Supercapacitor cells are not a monopoly. The global market includes Maxwell, Skeleton Technologies, Panasonic, Murata, Eaton, Nippon Chemi-Con, LS Materials and VinaTech.&lt;sup id="fnref:8"&gt;&lt;a href="#fn:8" class="footnote-ref" role="doc-noteref"&gt;8&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;Therefore, simply making supercapacitors is not a strong moat. VinaTech&amp;rsquo;s moat is elsewhere.&lt;/p&gt;
&lt;p&gt;First, it has qualified a product for Bloom&amp;rsquo;s data-center power system and converted that into a contract. Power-infrastructure products are not swapped like phone accessories. Voltage, current, temperature, lifetime, failure mode, software control, warranty and certification all matter.&lt;/p&gt;
&lt;p&gt;Second, VinaTech is moving from cell to module and system. Once cells are connected in series and parallel, balancing, thermal control, protection circuits and control software become critical.&lt;/p&gt;
&lt;p&gt;Third, customer qualification itself becomes a switching barrier in data-center power. A power-quality problem can become a server outage. Customers do not casually replace verified suppliers with cheaper unproven parts.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Moat element&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Strength&lt;/th&gt;
 &lt;th&gt;Explanation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Supercapacitor cell manufacturing&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;High-output, long-life, low-resistance know-how matters, but competitors exist.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Module and system design&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;Voltage balancing, thermal control, protection and software control raise difficulty.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bloom qualification&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;A supply contract is a strong qualification signal, though exclusivity is not confirmed.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mass production and quality control&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;Large customer supply requires delivery, defect control and traceability.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Proprietary source-material monopoly&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium&lt;/td&gt;
 &lt;td&gt;Public sources do not support a claim of global source-material monopoly.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Overall technology moat score: &lt;strong&gt;7/10&lt;/strong&gt;. Cell technology alone looks closer to a six. Including system design adoption inside Bloom&amp;rsquo;s data-center power system pushes it toward the mid-sevens. This is not an ASML-style equipment monopoly or a TSMC-style process moat. It is an application-qualification moat.&lt;/p&gt;
&lt;h2 id="7-business-moat-standard-content-would-change-the-story"&gt;7. Business Moat: Standard Content Would Change The Story
&lt;/h2&gt;&lt;p&gt;The business moat is more conditional than the technology moat.&lt;/p&gt;
&lt;p&gt;The strongest confirmed fact is the KRW 41.2B supply contract. It equals 50.12% of 2025 revenue. For a small company, that can change the base line.&lt;sup id="fnref3:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;But one contract is not enough to call this a durable moat.&lt;/p&gt;
&lt;p&gt;First, customer concentration matters. As Bloom revenue grows, VinaTech may become more dependent on Bloom. If Bloom dual-sources or pushes price, margins can be pressured.&lt;/p&gt;
&lt;p&gt;Second, system margins are not yet public. System supply has higher revenue, but it also includes PCB, enclosure, outsourcing, quality assurance, logistics and support costs. Gross margin could be better or worse than cell supply.&lt;/p&gt;
&lt;p&gt;Third, repeat orders are the key. If KRW 41.2B is a one-off project order, the multiple is limited. If VinaTech becomes standard content in Bloom&amp;rsquo;s SOFC data-center package, the company changes category.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Business moat item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current score&lt;/th&gt;
 &lt;th&gt;What could lift it&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer access&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8/10&lt;/td&gt;
 &lt;td&gt;Direct Bloom reference is valuable.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Repeat revenue probability&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6/10&lt;/td&gt;
 &lt;td&gt;Follow-on POs could lift this above 8.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pricing power&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5/10&lt;/td&gt;
 &lt;td&gt;Bloom concentration may cap bargaining power.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Switching cost&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8/10&lt;/td&gt;
 &lt;td&gt;Design-in power systems are costly to replace.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer diversification&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4/10&lt;/td&gt;
 &lt;td&gt;Needs proof beyond Bloom.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Overall business moat score: &lt;strong&gt;6.5/10 today&lt;/strong&gt;, with possible upgrade to &lt;strong&gt;8/10&lt;/strong&gt; if repeat orders appear. This is an early strong entry signal, not yet a structural monopoly.&lt;/p&gt;
&lt;h2 id="8-how-vinatech-complements-bloom"&gt;8. How VinaTech Complements Bloom
&lt;/h2&gt;&lt;p&gt;Bloom SOFC&amp;rsquo;s advantage is clear: it can provide on-site power before grid interconnection catches up. The CoreWeave project and Brookfield&amp;rsquo;s expanded $25B AI infrastructure power financing framework show Bloom&amp;rsquo;s link to the &amp;ldquo;time-to-power&amp;rdquo; problem.&lt;sup id="fnref2:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;sup id="fnref:9"&gt;&lt;a href="#fn:9" class="footnote-ref" role="doc-noteref"&gt;9&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;But SOFC is a high-temperature steady operating system. It is strong in stable operation, but AI GPU loads are abrupt. SOFC research highlights a trade-off between fast load following and safe operation because fuel starvation and thermal safety constrain how fast SOFC current should move.&lt;sup id="fnref1:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;VinaTech complements Bloom&amp;rsquo;s product.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Bloom does this&lt;/th&gt;
 &lt;th&gt;VinaTech helps with this&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Generates power on site&lt;/td&gt;
 &lt;td&gt;Buffers transient load so AI servers can consume that power reliably&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operates SOFC stacks stably&lt;/td&gt;
 &lt;td&gt;Reduces the need for abrupt SOFC output changes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shortens grid-interconnection dependence&lt;/td&gt;
 &lt;td&gt;Improves internal power quality and peak handling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Supplies long-duration power&lt;/td&gt;
 &lt;td&gt;Absorbs millisecond-to-second power shocks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;NVIDIA also frames AI-factory power as more than a capacity problem. It says BESS and power controls can manage fast-changing load profiles and improve power quality.&lt;sup id="fnref:10"&gt;&lt;a href="#fn:10" class="footnote-ref" role="doc-noteref"&gt;10&lt;/a&gt;&lt;/sup&gt; VinaTech is not a BESS provider, but it addresses the faster-response layer within the same problem.&lt;/p&gt;
&lt;h2 id="9-price-flow-and-consensus-a-good-story-is-not-a-comfortable-entry"&gt;9. Price, Flow and Consensus: A Good Story Is Not A Comfortable Entry
&lt;/h2&gt;&lt;p&gt;VinaTech closed at KRW 84,400 on July 3, 2026. Both pykrx and the local Kiwoom DB show that close. Foreign ownership was 15.54%. Using 1,115,794 foreign-held shares to infer total shares gives roughly 7.18M shares, implying market capitalization near KRW 606B.&lt;/p&gt;
&lt;p&gt;The near-term chart is unstable.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Return or status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;July 1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 107,700, limit-up&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 86,000, -20.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;July 3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 84,400, -1.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 5 trading days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 10 trading days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-25.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 20 trading days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-43.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20-day closing high&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 132,900 on June 1, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20-day closing low&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 80,200 on June 26, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Flow is mixed.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Retail&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Institution&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Real money&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 3D&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 10.58B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 14.49B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 4.13B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 3.84B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 5D&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 11.17B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 8.32B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 2.10B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 2.78B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 10D&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 6.44B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 12.81B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 7.27B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 5.22B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Last 20D&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 16.59B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 11.93B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 26.82B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 33.06B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Real money here combines investment trusts, private funds, pensions and insurance. The last three days show some real-money inflow, but the 20-day number remains weak. The news reaction was strong; durable long-only accumulation is not yet confirmed.&lt;/p&gt;
&lt;p&gt;Consensus also shows valuation pressure. Local DB Naver consensus shows FY2026 revenue of KRW 145.7B, operating profit of KRW 4.4B, net income of KRW 2.0B, PER of 306.6x and PBR of 8.0x. Whether those estimates fully incorporate the Bloom system contract needs separate verification. Still, the stock is not cheap on reported consensus. This is a repeat-order option, not a value stock.&lt;/p&gt;
&lt;h2 id="10-investment-stance-watch-to-conditional-buy-candidate"&gt;10. Investment Stance: Watch To Conditional Buy Candidate
&lt;/h2&gt;&lt;p&gt;VinaTech is not just a theme stock. It has a real Bloom contract and a product-step change from cell to system. AI data-center power volatility is a real structural issue supported by technical sources.&lt;/p&gt;
&lt;p&gt;But unconditional buying at the current price is weak.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;The stock has already reacted to the first contract.&lt;/li&gt;
&lt;li&gt;Follow-on Bloom orders are not yet confirmed.&lt;/li&gt;
&lt;li&gt;System-product margins remain blocked.&lt;/li&gt;
&lt;li&gt;Customer concentration and outsourced production are real risks.&lt;/li&gt;
&lt;/ol&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Stance&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Company&lt;/td&gt;
 &lt;td&gt;VinaTech&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ticker&lt;/td&gt;
 &lt;td&gt;126340 / KOSDAQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Idea type&lt;/td&gt;
 &lt;td&gt;Idiosyncratic Alpha&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current stance&lt;/td&gt;
 &lt;td&gt;Watch, with possible upgrade to conditional Buy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core thesis&lt;/td&gt;
 &lt;td&gt;Possible reclassification into a supercapacitor-system supplier inside Bloom SOFC data-center packages&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core risks&lt;/td&gt;
 &lt;td&gt;One-off Bloom order, low system margins, customer concentration, dual sourcing, high valuation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="11-entry-conditions"&gt;11. Entry Conditions
&lt;/h2&gt;&lt;h3 id="111-follow-on-bloom-purchase-orders"&gt;11.1 Follow-on Bloom Purchase Orders
&lt;/h3&gt;&lt;p&gt;This is the most important condition. If additional Bloom supply contracts appear in H2 2026 or H1 2027, the KRW 41.2B contract can be read as the start of standard content rather than a one-off.&lt;/p&gt;
&lt;h3 id="112-system-margin-proof"&gt;11.2 System-Margin Proof
&lt;/h3&gt;&lt;p&gt;Revenue growth without margin improvement weakens the thesis. System supply can carry higher ASP, but outsourcing, PCB, control and quality costs also rise. Gross margin and operating margin matter more than headline sales.&lt;/p&gt;
&lt;h3 id="113-customer-diversification"&gt;11.3 Customer Diversification
&lt;/h3&gt;&lt;p&gt;If VinaTech expands beyond Bloom into other SOFC providers, data-center power-infrastructure suppliers, PSU vendors or rack-power systems, concentration risk falls. That would move the thesis from &amp;ldquo;Bloom supplier option&amp;rdquo; to &amp;ldquo;AI data-center power-stability platform.&amp;rdquo;&lt;/p&gt;
&lt;h3 id="114-price-stabilization-after-the-news-spike"&gt;11.4 Price Stabilization After The News Spike
&lt;/h3&gt;&lt;p&gt;The July 1 limit-up and July 2 selloff show very high volatility. Without repeat orders, buying the spike means buying theme beta more than alpha.&lt;/p&gt;
&lt;h2 id="12-catalysts-and-invalidation"&gt;12. Catalysts And Invalidation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Catalyst&lt;/th&gt;
 &lt;th&gt;Checkpoint&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Catalyst 1&lt;/td&gt;
 &lt;td&gt;Additional Bloom system supply contract&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Catalyst 2&lt;/td&gt;
 &lt;td&gt;Bloom revenue recognition in Q4 2026 to Q1 2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Catalyst 3&lt;/td&gt;
 &lt;td&gt;Bloom/Brookfield AI power projects converting from framework to project-level orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Catalyst 4&lt;/td&gt;
 &lt;td&gt;ASP lift from cell to module/system conversion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Catalyst 5&lt;/td&gt;
 &lt;td&gt;AI data-center power quality becoming a standard requirement across PSU, BESS and supercapacitor layers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The invalidation list is more important.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Invalidation&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;No follow-on Bloom PO by H1 2027&lt;/td&gt;
 &lt;td&gt;KRW 41.2B contract may be one-off&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;System sales rise but operating margin does not&lt;/td&gt;
 &lt;td&gt;High-value-system thesis weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bloom dual-sources aggressively&lt;/td&gt;
 &lt;td&gt;VinaTech share and pricing power fall&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SOFC data-center projects shift to other power solutions&lt;/td&gt;
 &lt;td&gt;Bloom chain expansion slows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Outsourcing, quality or delivery issues appear&lt;/td&gt;
 &lt;td&gt;High-reliability data-center trust weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="13-evidence-classification"&gt;13. Evidence Classification
&lt;/h2&gt;&lt;h3 id="fact"&gt;Fact
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;AI data-center power is not just a capacity issue. NVIDIA says synchronized GPU training creates rapid power swings, and GB300 NVL72 energy storage can reduce peak grid demand by up to 30%.&lt;sup id="fnref1:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;Bloom announced fuel-cell deployment for CoreWeave&amp;rsquo;s AI data center.&lt;sup id="fnref3:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;SOFC has limits in fast load following; hybrid SOFC DC microgrid research uses batteries and supercapacitors to cover transient power.&lt;sup id="fnref2:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;VinaTech signed a KRW 41.215B data-center supercapacitor-system contract with Bloom, equal to 50.12% of 2025 revenue.&lt;sup id="fnref4:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;li&gt;The contract was reported as VinaTech&amp;rsquo;s first direct supply of an entire data-center supercapacitor system.&lt;sup id="fnref1:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="inference"&gt;Inference
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;VinaTech&amp;rsquo;s core role is transient power buffering for Bloom SOFC systems serving AI data centers.&lt;/li&gt;
&lt;li&gt;Its moat is stronger in module/system design and Bloom qualification than in the standalone cell.&lt;/li&gt;
&lt;li&gt;Repeat orders could reclassify VinaTech from a supercapacitor-cell company into an AI power-stability system supplier.&lt;/li&gt;
&lt;li&gt;The current stock price already embeds some repeat-order expectation.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="speculation"&gt;Speculation
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;VinaTech supercapacitor systems could become standard content in Bloom SOFC data-center packages.&lt;/li&gt;
&lt;li&gt;VinaTech could expand beyond Bloom to other data-center power-infrastructure customers.&lt;/li&gt;
&lt;li&gt;System ASP uplift could translate into higher operating margin.&lt;/li&gt;
&lt;li&gt;The Meta, CoreWeave, Bloom and VinaTech chain could become project-specific and direct in future disclosures.&lt;/li&gt;
&lt;/ul&gt;
&lt;h3 id="blocked"&gt;Blocked
&lt;/h3&gt;&lt;ul&gt;
&lt;li&gt;Exact Bloom product specs: voltage, capacitance, ESR, cycle life and thermal spec.&lt;/li&gt;
&lt;li&gt;Actual gross margin on the Bloom contract.&lt;/li&gt;
&lt;li&gt;VinaTech&amp;rsquo;s share and exclusivity within Bloom.&lt;/li&gt;
&lt;li&gt;VinaTech content per MW of Bloom SOFC data-center capacity.&lt;/li&gt;
&lt;li&gt;Whether Meta workloads actually run in a Bloom-powered CoreWeave data center.&lt;/li&gt;
&lt;li&gt;Exact power-stage location: SOFC output, DC bus, UPS auxiliary layer or rack-side buffer.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;VinaTech is not an AI data-center power-generation stock. The cleaner description is &lt;strong&gt;a transient-load-buffer system supplier inside Bloom SOFC data-center power systems&lt;/strong&gt;. Its moat comes from customer qualification, design-in status and systemization more than unique raw cell technology.&lt;/p&gt;
&lt;p&gt;The first contract is already in the price. The next alpha must come from follow-on purchase orders and margins. If Bloom repeat orders arrive and system revenue lifts profitability, VinaTech can be reclassified. Without follow-on orders, chasing the stock means buying an expensive theme, not a confirmed compounder.&lt;/p&gt;
&lt;p&gt;The practical conclusion is simple: &lt;strong&gt;the business position has improved; the price has already moved; the next test is repeat Bloom orders and system margins.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="source-ledger"&gt;Source Ledger
&lt;/h2&gt;&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;CBC News Korea, &lt;a class="link" href="https://cbci.co.kr/news/articleView.html?idxno=585647" target="_blank" rel="noopener"
 &gt;VinaTech signs KRW 41.2B data-center supercapacitor contract with Bloom Energy&lt;/a&gt;, July 1, 2026.&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref2:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref3:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref4:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;NVIDIA Technical Blog, &lt;a class="link" href="https://developer.nvidia.com/blog/how-new-gb300-nvl72-features-provide-steady-power-for-ai/" target="_blank" rel="noopener"
 &gt;How New GB300 NVL72 Features Provide Steady Power for AI&lt;/a&gt;, July 28, 2025.&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;Bloom Energy, &lt;a class="link" href="https://investor.bloomenergy.com/press-releases/press-release-details/2024/Bloom-Energy-and-CoreWeave-Partner-to-Revolutionize-AI-Data-Center-Power-Solutions/default.aspx" target="_blank" rel="noopener"
 &gt;Bloom Energy and CoreWeave Partner to Revolutionize AI Data Center Power Solutions&lt;/a&gt;, July 17, 2024.&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref2:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref3:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;Lin Zhang et al., &lt;a class="link" href="https://pubs.rsc.org/en/content/articlehtml/2023/se/d2se01559e" target="_blank" rel="noopener"
 &gt;Optimization of energy management in hybrid SOFC-based DC microgrid&lt;/a&gt;, Sustainable Energy &amp;amp; Fuels, 2023.&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref2:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;The Bell, &lt;a class="link" href="https://www.thebell.co.kr/front/newsview.asp?key=202607011321036760107013" target="_blank" rel="noopener"
 &gt;VinaTech targets AI data-center demand after breaking into Bloom Energy&lt;/a&gt;, July 1, 2026.&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;VINA Tech, &lt;a class="link" href="https://www.vinatech.com/en/sub/pr/news.php?bid=2&amp;amp;idx=3467&amp;amp;mode=view" target="_blank" rel="noopener"
 &gt;Target AI Data Center and Sales by 2030 Trillion&lt;/a&gt;, July 2, 2026.&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:7"&gt;
&lt;p&gt;CoreWeave, &lt;a class="link" href="https://www.coreweave.com/news/coreweave-and-meta-announce-21-billion-expanded-ai-infrastructure-agreement" target="_blank" rel="noopener"
 &gt;CoreWeave and Meta Expand $21B AI Cloud Deal&lt;/a&gt;, April 30, 2026.&amp;#160;&lt;a href="#fnref:7" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:8"&gt;
&lt;p&gt;Global Market Insights, &lt;a class="link" href="https://www.gminsights.com/industry-analysis/supercapacitor-market" target="_blank" rel="noopener"
 &gt;Supercapacitor Market Size &amp;amp; Share 2026-2035&lt;/a&gt;, accessed July 4, 2026.&amp;#160;&lt;a href="#fnref:8" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:9"&gt;
&lt;p&gt;Bloom Energy, &lt;a class="link" href="https://investor.bloomenergy.com/press-releases/press-release-details/2026/Brookfield-and-Bloom-Energy-Expand-AI-Infrastructure-Partnership-to-25-Billion-Fivefold-Increase-to-Build-and-Finance-Rapid-Power-for-AI-Infrastructure/default.aspx" target="_blank" rel="noopener"
 &gt;Brookfield and Bloom Energy Expand AI Infrastructure Partnership to $25 Billion&lt;/a&gt;, June 30, 2026.&amp;#160;&lt;a href="#fnref:9" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:10"&gt;
&lt;p&gt;NVIDIA Technical Blog, &lt;a class="link" href="https://developer.nvidia.com/blog/designing-production-ready-battery-energy-storage-systems-for-ai-factories/" target="_blank" rel="noopener"
 &gt;Designing Production-Ready Battery Energy Storage Systems for AI Factories&lt;/a&gt;, 2026.&amp;#160;&lt;a href="#fnref:10" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>How SK Hynix's ADR Could Change 000660 Price Discovery And Leveraged ETF Flows</title><link>https://koreainvestinsights.com/post/sk-hynix-adr-000660-price-path-leverage-etf-plumbing-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 07:18:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/sk-hynix-adr-000660-price-path-leverage-etf-plumbing-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-f1a-f6-skhy-adr-hbm-capex-2026-07-04/" &gt;SK Hynix F-1/A and F-6: how the SKHY ADR packages HBM capex&lt;/a&gt;. The previous note focused on the filing mechanics. This note focuses on price transmission, ADR-common-share arbitrage and single-stock leveraged ETF flows.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The SK Hynix ADR is structurally positive over the medium term, but it is not a near-term volatility dampener. It widens U.S. investor access and may allow SK Hynix to be compared more directly with U.S.-listed AI memory peers. But right after listing, the key variables are the offer discount, the first week of ADR trading quality, the ADR premium or discount to the KRX common share and the impact on Korea&amp;rsquo;s single-stock leveraged ETF plumbing.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Is the ADR positive for 000660?&lt;/td&gt;
 &lt;td&gt;Medium-term yes, because investor access and peer framing improve.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is it an immediate buy signal?&lt;/td&gt;
 &lt;td&gt;No. Pricing, discount and initial premium or discount must be checked first.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;What is the key mechanism?&lt;/td&gt;
 &lt;td&gt;ADR-common-share arbitrage. If the ADR trades rich to implied parity, it can create KRX common-share buying pressure. If it trades cheap, it can pressure the common share.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Can the ADR absorb Korea&amp;rsquo;s 2x ETF overheating?&lt;/td&gt;
 &lt;td&gt;Not in the short run. The bottleneck is KRX close rebalancing, not U.S. access.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current stance&lt;/td&gt;
 &lt;td&gt;Watch the common share and the ADR. Avoid leveraged wrappers if the goal is HBM exposure.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="1-what-is-confirmed"&gt;1. What Is Confirmed
&lt;/h2&gt;&lt;p&gt;SK Hynix has filed an F-1/A for a planned Nasdaq Global Select Market ADS listing under &lt;code&gt;SKHY&lt;/code&gt;.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt; The F-6 filed by Citibank states that each ADS represents the right to receive one-tenth of one common share.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt; Press reports indicate a maximum new-share issuance of 17.79 million common shares, around 2.5% of existing common shares.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The final ADS count, offer price and net proceeds remain blank in the F-1/A. That matters. The investment signal is not merely that an ADR is coming. The signal will be in the final price, the discount to the Korean common share and the first few days of trading.&lt;/p&gt;
&lt;h2 id="2-access-premium"&gt;2. Access Premium
&lt;/h2&gt;&lt;p&gt;The first effect is access. U.S. investors can buy SK Hynix exposure in dollars, during U.S. hours, without using a Korea trading account or indirect ETF exposure. That can shift the comparison set from “Korean memory stock” to “U.S.-tradable AI memory and HBM supplier.”&lt;/p&gt;
&lt;p&gt;This is positive, but not new information for the stock. Much of the obvious ADR rerating argument has already been anticipated. The remaining alpha is whether U.S. investors pay at or above implied KRX parity after the ADR starts trading.&lt;/p&gt;
&lt;h2 id="3-new-shares-and-dilution"&gt;3. New Shares And Dilution
&lt;/h2&gt;&lt;p&gt;The ADR is linked to new share issuance, not merely a secondary sale by existing holders. The headline maximum is about 2.5% of existing shares. That is dilution in ownership terms, but not necessarily large economic dilution if the offer price is close to market and proceeds are invested into HBM, eSSD, advanced packaging and EUV-related capacity.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Small discount and strong order book&lt;/td&gt;
 &lt;td&gt;U.S. investors validate the common-share price. Positive for 000660.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Large discount and weak demand&lt;/td&gt;
 &lt;td&gt;U.S. market rejects the prevailing valuation. Negative for 000660.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADR premium after listing&lt;/td&gt;
 &lt;td&gt;U.S. demand can support KRX common shares.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Persistent ADR discount&lt;/td&gt;
 &lt;td&gt;U.S. trading becomes a new downside anchor.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="4-adr-common-share-arbitrage"&gt;4. ADR-Common-Share Arbitrage
&lt;/h2&gt;&lt;p&gt;With one ADS representing 0.1 common share, parity is simple:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Calculation&lt;/th&gt;
 &lt;th&gt;Formula&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Implied ADR price from KRX common share&lt;/td&gt;
 &lt;td&gt;&lt;code&gt;KRX common price × 0.1 ÷ USD/KRW&lt;/code&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Implied common-share price from ADR&lt;/td&gt;
 &lt;td&gt;&lt;code&gt;ADR price × USD/KRW × 10&lt;/code&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;If the ADR trades above implied parity, arbitrageurs may buy KRX common shares and sell or create ADR exposure. That creates common-share buying pressure. If the ADR trades below parity, arbitrageurs may buy ADRs and sell common-share exposure, pressuring the KRX line.&lt;/p&gt;
&lt;p&gt;Therefore the first week is not about the absolute ADR move. It is about the premium or discount versus the Korean common share, after conversion costs, FX, settlement and stock-borrow constraints.&lt;/p&gt;
&lt;h2 id="5-us-price-discovery"&gt;5. U.S. Price Discovery
&lt;/h2&gt;&lt;p&gt;The ADR extends SK Hynix price discovery into U.S. trading hours. This helps incorporate Nvidia, Micron, SOX, rates and AI capex news faster. But it also imports U.S. overnight gap risk into the Korean open. The ADR is a price-discovery channel, not automatically a volatility cushion.&lt;/p&gt;
&lt;h2 id="6-leveraged-etf-plumbing"&gt;6. Leveraged ETF Plumbing
&lt;/h2&gt;&lt;p&gt;The ADR does not directly remove Korea&amp;rsquo;s single-stock 2x ETF rebalancing pressure. Korean 2x ETFs must maintain daily target exposure through common shares, futures, swaps and liquidity provider hedging around the KRX close. The Korean Financial Services Commission described the daily 2x structure when single-stock leveraged products were introduced.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt; The Korea Capital Market Institute estimated SK Hynix leveraged ETF AUM at KRW 9.15tn as of June 19 and highlighted how daily rebalancing can amplify volatility.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The core issue is mechanical. When SK Hynix rises, 2x ETF exposure may need to increase. When it falls, exposure may need to decrease. That creates a short-gamma-like flow. An ADR only offsets this if U.S. long-only ADR demand is large enough to replace the domestic leveraged flow. That is possible later, but not proven now.&lt;/p&gt;
&lt;h2 id="7-what-to-watch"&gt;7. What To Watch
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Watch item&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Final offer price and discount&lt;/td&gt;
 &lt;td&gt;Tests U.S. demand quality.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First 3 to 5 days of ADR premium or discount&lt;/td&gt;
 &lt;td&gt;Determines whether U.S. trading supports or pressures 000660.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADR trading value&lt;/td&gt;
 &lt;td&gt;Shows whether SKHY can become a real price-discovery line.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean leveraged ETF AUM&lt;/td&gt;
 &lt;td&gt;Shows whether the domestic short-gamma pipe is cooling.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM ASP, shipment and capex commentary&lt;/td&gt;
 &lt;td&gt;Keeps the ADR story grounded in earnings.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;SKHY is a major capital-market event for SK Hynix. It can broaden the investor base and reframe the company as a U.S.-tradable AI memory supplier. But the ADR does not immediately solve the domestic leveraged ETF problem. The near-term stance is watchlist for the common share and ADR, and avoidance of leveraged wrappers if the goal is clean HBM exposure.&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119312526289345/d32785df1a.htm" target="_blank" rel="noopener"
 &gt;SEC, SK Hynix Form F-1/A, filed June 30, 2026&lt;/a&gt;&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/1472033/000119380526000898/e665622_f6-skhynix.htm" target="_blank" rel="noopener"
 &gt;SEC, SK Hynix Form F-6, filed July 1, 2026&lt;/a&gt;&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.thebell.co.kr/front/newsview.asp?key=202606241720222280104920" target="_blank" rel="noopener"
 &gt;The Bell, SK Hynix Nasdaq ADR listing report&lt;/a&gt;&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.fsc.go.kr/no010107/86981" target="_blank" rel="noopener"
 &gt;Korea Financial Services Commission, single-stock leveraged product notice&lt;/a&gt;&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.kcmi.re.kr/publications/pub_detail_view?cno=6801&amp;amp;syear=2026&amp;amp;zcd=002001016&amp;amp;zno=1922" target="_blank" rel="noopener"
 &gt;Korea Capital Market Institute, single-stock leveraged and inverse ETF flows&lt;/a&gt;&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>SK Gas vs SK Oceanplant: Which Second-Line AI Data-Center Power Candidate Is Cleaner?</title><link>https://koreainvestinsights.com/post/sk-gas-sk-oceanplant-ai-datacenter-power-second-line-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 05:54:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/sk-gas-sk-oceanplant-ai-datacenter-power-second-line-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-capex-5p3t-korea-power-substrate-storage-bottleneck-2026-06-05/" &gt;AI data-center capex and Korea&amp;rsquo;s bottlenecks&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/vinatech-bloom-ai-datacenter-supercapacitor-power-buffer-2026-07-04/" &gt;VinaTech and Bloom Energy&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/gnc-energy-ai-datacenter-backup-generator-sustainable-rally-2026-07-04/" &gt;GNC Energy and AI data-center backup power&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/construction-epc-rerating-ai-power-nuclear-datacenter-reconstruction-2026-06-14/" &gt;construction and EPC rerating&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;H1 2026 AI infrastructure review&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;After SK Group&amp;rsquo;s AI data-center investment announcement, the market did not treat every SK affiliate as a clean winner. It looked for the names that are closest to the actual power bottleneck. On that basis, SK Gas is the more direct second-line AI data-center power candidate than SK Oceanplant.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th&gt;Summary&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td&gt;Directly linked to Ulsan GPS, a 1.2GW LNG/LPG combined-cycle power plant, the KET LNG terminal, fuel trading capability and SK Group&amp;rsquo;s Ulsan AI data-center power structure.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Oceanplant&lt;/td&gt;
 &lt;td&gt;Not a direct power supplier to AI data centers. It is an offshore-wind foundation manufacturer, and works better as a second-line candidate if the power-shortage theme broadens into renewables and offshore-wind infrastructure.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Current priority&lt;/td&gt;
 &lt;td&gt;SK Gas &amp;gt; SK Oceanplant. SK Gas has valuation and asset-revaluation support. SK Oceanplant has early flow-turn signals, but the chart and earnings visibility are still weaker.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="0-basis"&gt;0. Basis
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Detail&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Companies&lt;/td&gt;
 &lt;td&gt;SK Gas, SK Oceanplant&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Topic&lt;/td&gt;
 &lt;td&gt;Second-line power candidates after SK Group&amp;rsquo;s AI data-center plan&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price, consensus and flow basis&lt;/td&gt;
 &lt;td&gt;July 3, 2026 close and recent flow snapshot&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Public-source check&lt;/td&gt;
 &lt;td&gt;July 4, 2026 KST&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Key public events&lt;/td&gt;
 &lt;td&gt;SK Group AI data-center plan, Ulsan AI data-center project, SK Gas sale of 49% of Ulsan GPS, SK Oceanplant&amp;rsquo;s Taiwan offshore-wind foundation delivery&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Conclusion&lt;/td&gt;
 &lt;td&gt;SK Gas first, SK Oceanplant conditional watchlist&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is not a claim that either company has a confirmed AI data-center power contract. The point is to rank exposure, valuation and flow quality.&lt;/p&gt;
&lt;h2 id="1-what-the-market-selected-after-sk-groups-announcement"&gt;1. What The Market Selected After SK Group&amp;rsquo;s Announcement
&lt;/h2&gt;&lt;p&gt;SK Group&amp;rsquo;s June 29, 2026 announcement framed AI infrastructure as a full-stack system that combines semiconductors, data centers, power and telecom. Local reporting said SK is already working with AWS on an Ulsan AI data center, combining SK Telecom, SK Broadband and SK Gas capabilities, and plans to expand similar models toward 15GW of AI data centers.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;Al Jazeera also reported South Korea&amp;rsquo;s broader chip and AI data-center investment drive. The reported plan includes large-scale data-center investment involving SK Group, GS Group and Naver, and mentions a 10GW additional AI data-center target by 2035.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;But stock-market reaction was selective.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Jun 30-Jul 3 return&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.6%&lt;/td&gt;
 &lt;td&gt;Broad market correction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Inc.&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.1%&lt;/td&gt;
 &lt;td&gt;Holding-company and funding discount&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-7.7%&lt;/td&gt;
 &lt;td&gt;Crowded memory leader pulled back&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Telecom&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.2%&lt;/td&gt;
 &lt;td&gt;AI data-center opportunity offset by cost concerns&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.1%&lt;/td&gt;
 &lt;td&gt;NAV discount partly cushioned the move&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.6%&lt;/td&gt;
 &lt;td&gt;Power and gas infrastructure directness stood out&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK REITs&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.6%&lt;/td&gt;
 &lt;td&gt;Infrastructure and asset-value defense&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The signal is not the absolute gain, but the dispersion. The market did not simply buy the SK Group label. It preferred the affiliates tied to power, land, infrastructure and asset value.&lt;/p&gt;
&lt;h2 id="2-core-comparison"&gt;2. Core Comparison
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Gas&lt;/th&gt;
 &lt;th style="text-align: right"&gt;SK Oceanplant&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 218,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 15,320&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jun 30-Jul 3 return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;1-week return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;1-month return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2-month return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-24.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-37.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026F PER&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.9x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;21.6x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026F PBR&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.52x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1.15x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026F ROE&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Consensus target price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 367,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 26,167&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Target-price upside&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+70.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI data-center directness&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Low to medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investment character&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Power and gas infrastructure plus undervaluation&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Offshore-wind infrastructure plus turnaround option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The target-price upside looks similar. That is not enough. The better question is whether the company is close to the bottleneck, whether valuation is already stretched, and whether flow is turning from a damaged chart into real accumulation.&lt;/p&gt;
&lt;p&gt;On those tests, SK Gas is ahead.&lt;/p&gt;
&lt;h2 id="3-sk-gas-the-more-direct-power-candidate"&gt;3. SK Gas: The More Direct Power Candidate
&lt;/h2&gt;&lt;p&gt;SK Gas has a clearer link to SK Group&amp;rsquo;s AI data-center power structure. The Ulsan project reportedly combines SK Telecom, SK Broadband and SK Gas capabilities.&lt;sup id="fnref1:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt; That makes SK Gas more than a generic group affiliate.&lt;/p&gt;
&lt;p&gt;The relevant assets are straightforward.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Asset or capability&lt;/th&gt;
 &lt;th&gt;Investor meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Ulsan GPS&lt;/td&gt;
 &lt;td&gt;1.2GW LNG/LPG dual-fuel combined-cycle power plant, the closest physical power asset in this comparison.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KET LNG terminal&lt;/td&gt;
 &lt;td&gt;LNG procurement, storage and supply infrastructure.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LPG/LNG trading&lt;/td&gt;
 &lt;td&gt;Fuel procurement and pricing capability that can affect generation economics.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;SK Gas describes its LNG and power business on its official site.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt; The Fact reported that Ulsan GPS started commercial operation in December 2024 as a 1.2GW LNG/LPG dual-fuel combined-cycle power plant. On June 30, 2026, SK Gas sold a 49% stake in Ulsan GPS for roughly KRW 1.2242tn while retaining a 51% controlling stake.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;That transaction matters because it confirms external asset value, releases capital and preserves strategic control.&lt;/p&gt;
&lt;h3 id="sk-gas-investment-points"&gt;SK Gas Investment Points
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Point&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;AI data-center power directness&lt;/td&gt;
 &lt;td&gt;Potentially tied to SK Group&amp;rsquo;s Ulsan AI data-center power stack.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Valuation&lt;/td&gt;
 &lt;td&gt;2026F PER 3.9x and PBR 0.52x are not demanding.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Asset revaluation&lt;/td&gt;
 &lt;td&gt;The Ulsan GPS stake sale validated infrastructure value.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shareholder return&lt;/td&gt;
 &lt;td&gt;2026F DPS KRW 10,500, roughly a 4% yield.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flow&lt;/td&gt;
 &lt;td&gt;Latest four trading days: institutions +KRW 0.52bn, real-money-style flow +KRW 0.86bn.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The stock is not yet a strong momentum name. It is above the 20-day moving average, but still 8.9% below the 60-day line, with RSI around 40. This looks more like a low-valuation infrastructure name trying to base after a drawdown than a confirmed trend stock.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Level&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 214,000-218,000&lt;/td&gt;
 &lt;td&gt;First support zone to monitor.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reclaim of KRW 222,000&lt;/td&gt;
 &lt;td&gt;Initial strength confirmation.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Break above KRW 231,500&lt;/td&gt;
 &lt;td&gt;A stronger trend-reversal signal.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Break below KRW 203,000-209,000&lt;/td&gt;
 &lt;td&gt;Weakens the rebound thesis.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The key follow-ups are the second-half strategy update, actual power-supply structure for the Ulsan AI data center, use of the Ulsan GPS sale proceeds and the 2027-2029 shareholder-return policy.&lt;/p&gt;
&lt;h2 id="4-sk-oceanplant-offshore-wind-infrastructure-option"&gt;4. SK Oceanplant: Offshore-Wind Infrastructure Option
&lt;/h2&gt;&lt;p&gt;SK Oceanplant should be framed differently. It does not directly sell electricity to AI data centers. It manufactures offshore-wind foundations. So the right thesis is not &amp;ldquo;direct AI data-center power exposure,&amp;rdquo; but &amp;ldquo;second-line exposure if the power-shortage theme broadens into renewables and offshore-wind infrastructure.&amp;rdquo;&lt;/p&gt;
&lt;p&gt;Recent business news is constructive. Yonhap reported that SK Oceanplant delivered the first foundations for Taiwan&amp;rsquo;s Fengmiao 1 offshore-wind project on July 2, 2026. The project uses 33 15MW turbines for a 495MW wind farm, and SK Oceanplant signed a KRW 393.4bn contract in June 2024 to supply 21 foundations.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The company is also investing KRW 1.153tn in what it describes as the world&amp;rsquo;s largest specialized offshore-wind foundation production base, targeting completion by the end of 2026.&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;h3 id="sk-oceanplant-investment-points"&gt;SK Oceanplant Investment Points
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Point&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Theme&lt;/td&gt;
 &lt;td&gt;Benefits if AI power shortages broaden into generation buildout and offshore wind.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Recent event&lt;/td&gt;
 &lt;td&gt;First Fengmiao 1 delivery gives visibility to the KRW 393.4bn contract.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flow&lt;/td&gt;
 &lt;td&gt;Jun 30-Jul 3: foreigners +KRW 1.67bn, institutions +KRW 3.07bn, retail -KRW 5.02bn.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Program flow&lt;/td&gt;
 &lt;td&gt;Jun 30-Jul 3 program buying of KRW 3.00bn.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Burden&lt;/td&gt;
 &lt;td&gt;2026F PER 21.6x, ROE 5.4% and a still-damaged chart.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The recent four-day flow is constructive, but not enough to call structural accumulation. On a 20-day basis, foreigners plus institutions were still net sellers by KRW 18.69bn.&lt;/p&gt;
&lt;p&gt;The chart is also still weak. The stock is 4.6% below its 20-day moving average, 21.8% below the 60-day line and 48.5% below its two-month high. The July 1-2 rebound faded again on July 3.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Level&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 14,300-14,700&lt;/td&gt;
 &lt;td&gt;Aggressive support zone.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 16,200-16,650&lt;/td&gt;
 &lt;td&gt;First strength-confirmation zone.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Break above KRW 17,800&lt;/td&gt;
 &lt;td&gt;Potential trend reversal if turnover holds.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Break below KRW 13,900&lt;/td&gt;
 &lt;td&gt;Invalidates the early turn.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For SK Oceanplant, the decisive variables are additional offshore-wind orders, the new yard schedule and upward revisions to 2027 earnings estimates.&lt;/p&gt;
&lt;h2 id="5-final-view"&gt;5. Final View
&lt;/h2&gt;&lt;p&gt;SK Gas is the better second-line candidate today.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;It is closer to the AI data-center power stack.&lt;/li&gt;
&lt;li&gt;It trades at a lower 2026F PER and PBR.&lt;/li&gt;
&lt;li&gt;The Ulsan GPS stake sale validated asset value and provided capital.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;SK Oceanplant is still worth watching, but it is not the same kind of exposure. The Taiwan delivery and large production-base investment are meaningful. Still, this is an indirect offshore-wind infrastructure option, not a direct AI data-center power supplier.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Stance&lt;/th&gt;
 &lt;th&gt;What to confirm&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;SK Gas&lt;/td&gt;
 &lt;td&gt;Watchlist / Pullback Buy&lt;/td&gt;
 &lt;td&gt;KRW 214,000-218,000 support, reclaim of KRW 222,000, Ulsan AI data-center power structure, shareholder-return policy.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;SK Oceanplant&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;KRW 14,300-14,700 support, reclaim of KRW 16,200-16,650, new offshore-wind orders and 2027 earnings revisions.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The simple conclusion is this: in the AI data-center power trade, SK Gas is the closer infrastructure asset. SK Oceanplant is a higher-beta candidate for a later phase if the market expands the theme into offshore-wind generation buildout.&lt;/p&gt;
&lt;hr&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;&lt;a class="link" href="https://v.daum.net/v/20260629170816942?f=p" target="_blank" rel="noopener"
 &gt;MoneyToday via Daum, SK Group AI data-center plan&lt;/a&gt;&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&amp;#160;&lt;a href="#fnref1:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.aljazeera.com/news/2026/6/29/south-korea-announces-more-than-1-trillion-ai-chip-investment-drive" target="_blank" rel="noopener"
 &gt;Al Jazeera, “South Korea announces more than $1 trillion AI, chip investment drive”&lt;/a&gt;&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;&lt;a class="link" href="https://skgas.co.kr/Business/Gas_Energy.html" target="_blank" rel="noopener"
 &gt;SK Gas business overview&lt;/a&gt;&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;&lt;a class="link" href="https://v.daum.net/v/20260630160302257" target="_blank" rel="noopener"
 &gt;The Fact via Daum, SK Gas sale of 49% of Ulsan GPS&lt;/a&gt;&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;&lt;a class="link" href="https://v.daum.net/v/20260702150210493" target="_blank" rel="noopener"
 &gt;Yonhap via Daum, SK Oceanplant Fengmiao 1 delivery&lt;/a&gt;&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;&lt;a class="link" href="https://www.skoceanplant.com/bbs/board.php?bo_table=05_01&amp;amp;page=1&amp;amp;wr_id=59" target="_blank" rel="noopener"
 &gt;SK Oceanplant press release on offshore-wind production base&lt;/a&gt;&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>GNC Energy Rally Analysis: AI Data-Center Backup Power Stock Or Short-Term Overheat?</title><link>https://koreainvestinsights.com/post/gnc-energy-ai-datacenter-backup-generator-sustainable-rally-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 05:42:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/gnc-energy-ai-datacenter-backup-generator-sustainable-rally-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-capex-5p3t-korea-power-substrate-storage-bottleneck-2026-06-05/" &gt;AI data-center capex and Korea&amp;rsquo;s bottlenecks&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/vinatech-bloom-ai-datacenter-supercapacitor-power-buffer-2026-07-04/" &gt;VinaTech and Bloom Energy&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/construction-epc-rerating-ai-power-nuclear-datacenter-reconstruction-2026-06-14/" &gt;construction and EPC rerating&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/warsh-fed-expensive-money-era-forward-guidance-ai-infra-2026-06-19/" &gt;the era of expensive money&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/" &gt;H1 2026 AI infrastructure review&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;GNC Energy is not just a loose AI theme stock. It is a potential earnings-backed name in AI data-center backup power infrastructure. The company&amp;rsquo;s core business is generator EPC for emergency and commercial power systems. In AI data centers, that places it in the backup-power layer that keeps servers running when the grid or the power path fails.&lt;/p&gt;
&lt;p&gt;The latest rally has four drivers.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Driver&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Uplus AIDC cooperation&lt;/td&gt;
 &lt;td&gt;Media reports say cooperation started with emergency generators for the Paju AIDC and could broaden across AIDC power infrastructure.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung SDS contract&lt;/td&gt;
 &lt;td&gt;A reported KRW 29.748bn generator supply contract for Samsung SDS&amp;rsquo;s Gumi AI data center, equal to 11.3% of FY2025 revenue.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI data-center construction cycle&lt;/td&gt;
 &lt;td&gt;Data-center capex is moving from servers into power, cooling, backup power and construction.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Institutional flow&lt;/td&gt;
 &lt;td&gt;Latest five-day flow: institutions +KRW 7.66bn, including investment trusts +KRW 4.14bn and private funds +KRW 3.09bn.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The business thesis is valid, but the entry is not clean after a 45.5% five-day rally. Foreigners sold KRW 5.85bn over the same period, and foreign ownership fell from 5.12% on June 29 to 4.06% on July 3. On July 3, the stock traded as high as KRW 30,850 but closed at KRW 27,800, leaving a wide upper wick.&lt;/p&gt;
&lt;p&gt;My stance is Watchlist / Buy on pullback. The first support zone is KRW 25,400-26,000. The more comfortable zone is KRW 23,400-22,700. A reclaim of KRW 28,400 would be a 60-day moving-average recovery. A breakout above KRW 30,850 with stronger turnover and broader buying would revive the momentum case.&lt;/p&gt;
&lt;h2 id="0-basis"&gt;0. Basis
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Detail&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Company&lt;/td&gt;
 &lt;td&gt;GNC Energy&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ticker&lt;/td&gt;
 &lt;td&gt;119850 / KOSDAQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;As of&lt;/td&gt;
 &lt;td&gt;2026-07-04 KST&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price and flow basis&lt;/td&gt;
 &lt;td&gt;2026-07-03 close and recent flow snapshot&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Key events&lt;/td&gt;
 &lt;td&gt;LG Uplus AIDC cooperation report, Samsung SDS Gumi AI data-center generator contract report, data-center backup-generator demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stance&lt;/td&gt;
 &lt;td&gt;Watchlist / Buy on pullback&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The Samsung SDS KRW 29.748bn contract is treated based on media and public-disclosure reports. Contract text, revenue-recognition schedule and margin details still need confirmation. The right interpretation is not &amp;ldquo;all revenue immediately recognized,&amp;rdquo; but &amp;ldquo;order visibility improved.&amp;rdquo;&lt;/p&gt;
&lt;h2 id="1-what-gnc-energy-does"&gt;1. What GNC Energy Does
&lt;/h2&gt;&lt;p&gt;GNC Energy supplies diesel generators, gas turbine generators, gas generators and biogas power systems.&lt;sup id="fnref:1"&gt;&lt;a href="#fn:1" class="footnote-ref" role="doc-noteref"&gt;1&lt;/a&gt;&lt;/sup&gt; For AI data centers, the relevant business is emergency backup generation.&lt;/p&gt;
&lt;p&gt;Data centers do not only need more electricity. They need uninterrupted electricity. That is why a data center power stack includes grid connection, transformers, UPS, batteries, emergency generators and control systems.&lt;/p&gt;
&lt;p&gt;GNC Energy sits near the last line of defense. If Bloom Energy or fuel cells supply on-site generation, and VinaTech-type supercapacitor systems buffer transient power movement, GNC Energy is in the backup-power layer that allows a data center to keep operating when grid power is disrupted.&lt;/p&gt;
&lt;p&gt;That distinction matters. This is not a generic &amp;ldquo;AI power&amp;rdquo; story. It is more precise to describe GNC Energy as an AI data-center backup-power EPC supplier.&lt;/p&gt;
&lt;h2 id="2-why-the-stock-rallied"&gt;2. Why The Stock Rallied
&lt;/h2&gt;&lt;h3 id="lg-uplus-aidc"&gt;LG Uplus AIDC
&lt;/h3&gt;&lt;p&gt;ETNews reported that LG Uplus and GNC Energy started cooperation with emergency generators for the Paju AIDC and intend to broaden the cooperation across AIDC power infrastructure.&lt;sup id="fnref:2"&gt;&lt;a href="#fn:2" class="footnote-ref" role="doc-noteref"&gt;2&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;This matters because data-center power infrastructure requires qualification, reliability, service support and delivery capability. Once a supplier is inside the customer reference base, repeat opportunities become more plausible.&lt;/p&gt;
&lt;h3 id="samsung-sds-ai-data-center-contract"&gt;Samsung SDS AI Data-Center Contract
&lt;/h3&gt;&lt;p&gt;According to Herald Economy reporting, GNC Energy signed a generator supply contract with Samsung SDS for the Gumi AI data center. The amount is KRW 29.748bn, equivalent to 11.3% of FY2025 revenue, with a contract period reportedly running until May 31, 2029.&lt;sup id="fnref:3"&gt;&lt;a href="#fn:3" class="footnote-ref" role="doc-noteref"&gt;3&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;This is the most direct number behind the rally. It is a meaningful single project relative to annual revenue. The long contract period means investors should not pull the entire amount into a near-term quarter, but it does strengthen backlog and customer reference quality.&lt;/p&gt;
&lt;h3 id="ai-data-center-construction-cycle"&gt;AI Data-Center Construction Cycle
&lt;/h3&gt;&lt;p&gt;Asia Business Daily discussed GNC Energy alongside construction names such as Hyundai Engineering &amp;amp; Construction, GS E&amp;amp;C and DL E&amp;amp;C in the context of AI data-center construction investment.&lt;sup id="fnref:4"&gt;&lt;a href="#fn:4" class="footnote-ref" role="doc-noteref"&gt;4&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;The implication is that data centers are not only a construction or server story. The project stack includes power equipment, backup generation and delivery schedules. GNC Energy is smaller than the EPC primes, but it can carry higher beta to the backup-power node.&lt;/p&gt;
&lt;h3 id="institutional-buying"&gt;Institutional Buying
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Investor&lt;/th&gt;
 &lt;th style="text-align: right"&gt;5-day net flow&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Retail&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 2.47bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 5.85bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Institutions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 7.66bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investment trusts&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 4.14bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Private funds&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 3.09bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The constructive signal is that domestic institutions absorbed retail supply. The weak point is that foreigners were sellers. This is not yet clean foreign-plus-institution accumulation.&lt;/p&gt;
&lt;h2 id="3-earnings-level-up-and-margin-risk"&gt;3. Earnings Level-Up And Margin Risk
&lt;/h2&gt;&lt;p&gt;GNC Energy already produced real earnings in FY2025.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2025&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 262.6bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 49.4bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 40.0bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ROE&lt;/td&gt;
 &lt;td style="text-align: right"&gt;21.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;At the July 3 close, market cap was roughly KRW 458bn. Against FY2025 net income, that is about 11.5x trailing earnings. For an AI power-infrastructure theme stock with realized earnings, that is not obviously excessive.&lt;/p&gt;
&lt;p&gt;Samsung Securities argued in February 2026 that the company&amp;rsquo;s revenue step-up into the KRW 200bn-plus range was driven by a sharp increase in domestic data-center emergency-generator backlog. It also highlighted the move toward 80-100MW data centers and higher order value per MW.&lt;sup id="fnref:5"&gt;&lt;a href="#fn:5" class="footnote-ref" role="doc-noteref"&gt;5&lt;/a&gt;&lt;/sup&gt;&lt;/p&gt;
&lt;p&gt;But 1Q26 was weaker.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 55.4bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.4bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 12.3bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.95%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Electimes reported that FX-driven cost pressure on older projects weighed on operating profit.&lt;sup id="fnref:6"&gt;&lt;a href="#fn:6" class="footnote-ref" role="doc-noteref"&gt;6&lt;/a&gt;&lt;/sup&gt; This is the key risk. The next proof point is not only revenue growth. The company needs to recover double-digit operating margin in 2Q or 3Q.&lt;/p&gt;
&lt;h2 id="4-price-and-flow-setup"&gt;4. Price And Flow Setup
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Close&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 27,800&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daily return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.02%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Intraday high&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 30,850&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Intraday low&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 25,400&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Volume&lt;/td&gt;
 &lt;td style="text-align: right"&gt;625,004 shares&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5-day return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+45.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20-day return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+34.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;60-day return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Drawdown from 60-day high&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-29.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RSI14&lt;/td&gt;
 &lt;td style="text-align: right"&gt;61.3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 25,590&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 22,326&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;60-day MA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 28,410&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The chart is strong but not fully repaired. The stock rallied sharply, but it still closed below the 60-day moving average after failing to hold the intraday move above it.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Institutions&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Retail&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;3 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 4.15bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 5.43bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 2.01bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 5.85bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 7.66bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 2.47bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;10 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 6.39bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 8.85bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 3.03bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20 days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.95bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 5.22bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 4.01bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This looks like domestic institutional discovery, not a global long-only accumulation story yet.&lt;/p&gt;
&lt;h2 id="5-what-is-durable-and-what-is-fragile"&gt;5. What Is Durable And What Is Fragile
&lt;/h2&gt;&lt;p&gt;The durable part is the AI data-center power bottleneck. Higher power density, stricter uptime requirements and longer equipment lead times all support demand for backup power infrastructure.&lt;/p&gt;
&lt;p&gt;The company also has evidence. FY2025 showed a real earnings step-up, and the Samsung SDS contract is large enough to matter relative to revenue.&lt;/p&gt;
&lt;p&gt;The fragile part is timing and margin. The stock is up 45.5% in five sessions. Foreigners are selling. 1Q26 margin fell to 7.95%. The Samsung SDS contract extends to 2029, so it is not a near-term revenue lump.&lt;/p&gt;
&lt;h2 id="6-investment-view"&gt;6. Investment View
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Level&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;th&gt;Action&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 25,400-26,000&lt;/td&gt;
 &lt;td&gt;First pullback zone, near July 3 low and 5-day MA&lt;/td&gt;
 &lt;td&gt;Watch for pilot entry if institutional flow holds&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 23,400-22,700&lt;/td&gt;
 &lt;td&gt;More comfortable pullback zone, near the breakout base and 20-day MA&lt;/td&gt;
 &lt;td&gt;Better risk-reward&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 28,400 reclaim&lt;/td&gt;
 &lt;td&gt;60-day MA recovery&lt;/td&gt;
 &lt;td&gt;Trend repair signal&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 30,850 breakout&lt;/td&gt;
 &lt;td&gt;Recent intraday high&lt;/td&gt;
 &lt;td&gt;Momentum entry only with volume and broader buying&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Below KRW 23,400&lt;/td&gt;
 &lt;td&gt;Breakout failure risk&lt;/td&gt;
 &lt;td&gt;Stand aside or cut&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;GNC Energy belongs on the AI infrastructure watchlist. It is not a pure narrative stock: FY2025 earnings, data-center generator backlog, LG Uplus cooperation and the Samsung SDS contract all point in the same direction.&lt;/p&gt;
&lt;p&gt;The stock, however, has already moved. After a 45.5% five-day rally, foreign selling and a large upper wick argue against chasing. Watch the KRW 25,400-26,000 support zone first, and the KRW 23,400-22,700 zone for a more comfortable entry. A reclaim of KRW 28,400 and a breakout above KRW 30,850 would reopen the momentum case.&lt;/p&gt;
&lt;div class="footnotes" role="doc-endnotes"&gt;
&lt;hr&gt;
&lt;ol&gt;
&lt;li id="fn:1"&gt;
&lt;p&gt;GNC Energy, generator business page, &lt;a class="link" href="https://www.gncenergy.co.kr/business/generation-diesel.php" target="_blank" rel="noopener"
 &gt;company website&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:1" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:2"&gt;
&lt;p&gt;ETNews, LG Uplus AIDC cooperation report, June 28, 2026, &lt;a class="link" href="https://www.etnews.com/20260628000007" target="_blank" rel="noopener"
 &gt;article&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:2" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:3"&gt;
&lt;p&gt;Herald Economy, Samsung SDS Gumi AI data-center generator contract report, July 2, 2026, &lt;a class="link" href="https://heraldk.com/2026/07/01/%EC%A7%80%EC%97%94%EC%94%A8%EC%97%90%EB%84%88%EC%A7%80-297%EC%96%B5-%EA%B7%9C%EB%AA%A8-ai-%EB%8D%B0%EC%9D%B4%ED%84%B0-%EB%B0%9C%EC%A0%84%EA%B8%B0-%EA%B3%B5%EA%B8%89%EA%B3%84/" target="_blank" rel="noopener"
 &gt;article&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:3" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:4"&gt;
&lt;p&gt;Asia Business Daily, AI data-center construction cycle, July 1, 2026, &lt;a class="link" href="https://view.asiae.co.kr/article/2026070106070250006" target="_blank" rel="noopener"
 &gt;article&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:4" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:5"&gt;
&lt;p&gt;Samsung Securities, GNC Energy report, February 3, 2026, &lt;a class="link" href="https://www.samsungpop.com/common.do?cmd=down&amp;amp;contentType=application%2Fpdf&amp;amp;fileName=2010%2F2026020314503243K_02_04.pdf&amp;amp;inlineYn=Y&amp;amp;saveKey=research.pdf" target="_blank" rel="noopener"
 &gt;PDF&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:5" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;li id="fn:6"&gt;
&lt;p&gt;Electimes, GNC Energy 1Q26 report, May 15, 2026, &lt;a class="link" href="https://www.electimes.com/news/articleView.html?idxno=368103" target="_blank" rel="noopener"
 &gt;article&lt;/a&gt;.&amp;#160;&lt;a href="#fnref:6" class="footnote-backref" role="doc-backlink"&gt;&amp;#x21a9;&amp;#xfe0e;&lt;/a&gt;&lt;/p&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;/div&gt;</description></item><item><title>SK Hynix F-1/A and F-6: What The SKHY ADR Says About HBM Capex</title><link>https://koreainvestinsights.com/post/sk-hynix-f1a-f6-skhy-adr-hbm-capex-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 05:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/sk-hynix-f1a-f6-skhy-adr-hbm-capex-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/kr-deep-dive-sk-hynix-2026-04-16/" &gt;SK Hynix: the HBM leader in the AI revolution&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-hbm-market-share-ai-memory-demand-2026/" &gt;SK Hynix HBM market share 2026&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/sk-hynix-vs-micron-hbm-premium-ai-memory-platform-2026-05-31/" &gt;SK Hynix vs Micron&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 earnings&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;SK Hynix&amp;rsquo;s SEC filings are not just an ADR wrapper. The sequence from the June 24 F-1, the June 30 F-1/A and the July 1 F-6 shows the company packaging its HBM leadership, eSSD, Custom HBM and Korea capex roadmap for U.S. AI-memory investors.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Confirmed&lt;/th&gt;
 &lt;th&gt;Investor interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;F-1/A&lt;/td&gt;
 &lt;td&gt;Filed June 30, 2026. Nasdaq Global Select Market application under &lt;code&gt;SKHY&lt;/code&gt;. ADS count, price and net proceeds are still blank.&lt;/td&gt;
 &lt;td&gt;Not final terms yet. A further amendment or final prospectus is required.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;F-6&lt;/td&gt;
 &lt;td&gt;Filed July 1, 2026. Citibank is depositary. Each ADS represents one-tenth of one common share.&lt;/td&gt;
 &lt;td&gt;This fills the key ADR ratio that remained blank in the F-1/A cover.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;New-share ceiling&lt;/td&gt;
 &lt;td&gt;Board authorization for up to 17,790,000 new common shares, 2.50% of 712,702,365 existing shares.&lt;/td&gt;
 &lt;td&gt;Maximum dilution is limited, but actual issuance remains unknown.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The most important caveat: the F-6 registration of 1,780,000,000 ADSs is not the offering size. Form F-6 registers the ADR facility and the related depositary-fee calculation. The actual equity offering size depends on the final ADS count and offering price in the F-1/A or final prospectus.&lt;/p&gt;
&lt;p&gt;The stock read-through is Watchlist. This is an access and ownership-base event, not a fresh discovery of the HBM thesis. The alpha question is whether U.S. long-only and AI-memory investors use &lt;code&gt;SKHY&lt;/code&gt; as a cleaner vehicle for SK Hynix exposure.&lt;/p&gt;
&lt;h2 id="1-why-the-june-24-f-1-is-not-enough"&gt;1. Why The June 24 F-1 Is Not Enough
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Filing date&lt;/th&gt;
 &lt;th&gt;Form&lt;/th&gt;
 &lt;th&gt;Document&lt;/th&gt;
 &lt;th&gt;What changed&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-06-24&lt;/td&gt;
 &lt;td&gt;F-1&lt;/td&gt;
 &lt;td&gt;Registration No. 333-296987&lt;/td&gt;
 &lt;td&gt;Initial ADS offering and Nasdaq listing preparation; key offering terms blank&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-06-30&lt;/td&gt;
 &lt;td&gt;F-1/A&lt;/td&gt;
 &lt;td&gt;Amendment No. 1&lt;/td&gt;
 &lt;td&gt;Offering structure, new-share ceiling, use of proceeds and capex tables updated; ADS count and price still blank&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-07-01&lt;/td&gt;
 &lt;td&gt;F-6&lt;/td&gt;
 &lt;td&gt;Registration No. 333-297185&lt;/td&gt;
 &lt;td&gt;Citibank depositary and one ADS = one-tenth common share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The company is SK hynix Inc. Its common shares trade on KOSPI under &lt;code&gt;000660&lt;/code&gt;. The ADSs are expected to trade on the Nasdaq Global Select Market under &lt;code&gt;SKHY&lt;/code&gt;, subject to effectiveness and listing approval.&lt;/p&gt;
&lt;h2 id="2-offering-mechanics"&gt;2. Offering Mechanics
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Latest status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Issuer&lt;/td&gt;
 &lt;td&gt;SK hynix Inc.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADS ticker&lt;/td&gt;
 &lt;td&gt;&lt;code&gt;SKHY&lt;/code&gt;, Nasdaq Global Select Market application&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Depositary&lt;/td&gt;
 &lt;td&gt;Citibank, N.A.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Global coordinators&lt;/td&gt;
 &lt;td&gt;BofA Securities, Citigroup, Goldman Sachs, J.P. Morgan&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADS ratio&lt;/td&gt;
 &lt;td&gt;F-6: 1 ADS = 0.1 common share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ADS count&lt;/td&gt;
 &lt;td&gt;Blank in F-1/A&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Offering price&lt;/td&gt;
 &lt;td&gt;Blank in F-1/A&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net proceeds&lt;/td&gt;
 &lt;td&gt;Blank in F-1/A&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Lock-up&lt;/td&gt;
 &lt;td&gt;90-day lock-up may apply to the company and certain affiliates, subject to exceptions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The board authorized up to 17,790,000 new common shares. Against 712,702,365 existing shares, that is 2.50%. If fully issued, new shares would be about 2.44% of the post-issue share count. This is a ceiling, not the actual offering size.&lt;/p&gt;
&lt;h2 id="3-f-6-ads-registration-is-not-equity-proceeds"&gt;3. F-6 ADS Registration Is Not Equity Proceeds
&lt;/h2&gt;&lt;p&gt;The F-6 registers 1,780,000,000 ADSs, each representing one-tenth of one common share. It also uses a fee calculation of $5.00 per 100 ADSs, giving a proposed maximum aggregate fee-basis amount of $89,000,000 and a registration fee of $12,290.90.&lt;/p&gt;
&lt;p&gt;That fee basis is not the gross equity offering. Investors should not read 1.78 billion ADSs as the number of ADSs that will be sold in this offering. The real offering size remains blocked until the final ADS count and offer price are filed.&lt;/p&gt;
&lt;h2 id="4-use-of-proceeds-yongin-pt7-and-euv"&gt;4. Use Of Proceeds: Yongin, P&amp;amp;T7 And EUV
&lt;/h2&gt;&lt;p&gt;The F-1/A frames proceeds around general corporate purposes, including capital expenditures. The filing specifically discusses KRW 45.5 trillion of Korea production-facility capex and approximately KRW 11.9 trillion of EUV scanner cost, with EUV delivery expected by December 2027.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Project&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Target completion of planned investments&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Total expected cost&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Invested by May 31, 2026&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Additional planned investment&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Yongin Fab 1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;End of 2030&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 31.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.4T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 26.6T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cheongju P&amp;amp;T7&lt;/td&gt;
 &lt;td style="text-align: right"&gt;End of 2030&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 19.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 0.1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 18.9T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total&lt;/td&gt;
 &lt;td style="text-align: right"&gt;End of 2030&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 50.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 4.5T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 45.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Project&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2027&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2028&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2029&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2030&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Total&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Yongin Fab 1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 7.4T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 10.1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 6.6T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.5T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 0.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 26.6T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cheongju P&amp;amp;T7&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 0.5T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.7T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 5.8T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 7.8T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 18.9T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 7.9T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 12.2T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 9.3T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 8.3T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 7.8T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 45.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The business section and capex table use different milestones. The business section points to the first Yongin cleanroom opening in 1Q27 and P&amp;amp;T7 construction completion by end-2027. The capex table targets completion of planned investments by end-2030. That is not necessarily a contradiction; it separates phase-opening from full investment completion and equipment installation.&lt;/p&gt;
&lt;h2 id="5-the-narrative-hbm-custom-hbm-and-essd"&gt;5. The Narrative: HBM, Custom HBM And eSSD
&lt;/h2&gt;&lt;p&gt;The F-1/A does not present SK Hynix as a plain DRAM-cycle company. It emphasizes HBM, server DRAM and eSSD as structural AI-infrastructure products.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Market&lt;/th&gt;
 &lt;th&gt;IDC-based position in 1Q26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM including HBM&lt;/td&gt;
 &lt;td&gt;No. 2 by revenue, 29.1% share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM&lt;/td&gt;
 &lt;td&gt;No. 1 by revenue, 56.4% share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND flash&lt;/td&gt;
 &lt;td&gt;No. 2 by revenue, 18.5% share&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Custom HBM is especially important. The filing defines it as HBM that integrates certain GPU or ASIC functions into the HBM base die according to customer configuration. That shifts the story from memory pricing alone toward memory becoming part of the accelerator design stack.&lt;/p&gt;
&lt;h2 id="6-recent-financials"&gt;6. Recent Financials
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q25&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Change&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 17.639T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 52.576T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+198.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Profit for the period&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 8.108T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 40.346T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~5.0x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total assets&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 176.108T at 2025-end&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 222.829T at Mar. 31, 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total equity&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 120.667T at 2025-end&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 164.380T at Mar. 31, 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+36.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Product&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26 revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Share&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 40.659T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;77.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 11.574T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;22.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Other&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 0.343T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The U.S. accounted for KRW 33.999T, or 64.7% of 1Q26 revenue. That makes the ADR event strategically coherent: the investor access point is moving closer to the customer-revenue geography.&lt;/p&gt;
&lt;h2 id="7-shareholders-dividends-and-governance"&gt;7. Shareholders, Dividends And Governance
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Shareholder&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Shares&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Ownership&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square&lt;/td&gt;
 &lt;td style="text-align: right"&gt;146,100,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20.50%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;National Pension Service&lt;/td&gt;
 &lt;td style="text-align: right"&gt;57,439,774&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8.06%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;BlackRock&lt;/td&gt;
 &lt;td style="text-align: right"&gt;36,407,157&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5.11%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Capital Research and Management&lt;/td&gt;
 &lt;td style="text-align: right"&gt;25,149,374&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.53%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Others&lt;/td&gt;
 &lt;td style="text-align: right"&gt;445,979,195&lt;/td&gt;
 &lt;td style="text-align: right"&gt;62.58%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Treasury shares&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,626,865&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0.23%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The filing says the maximum offering size reflects the Korean Monopoly Regulation and Fair Trade Act requirement that SK Square maintain at least 20% ownership.&lt;/p&gt;
&lt;p&gt;The 2025-2027 base dividend policy is KRW 1,500 per share per year, paid quarterly at KRW 375. The 2025 total dividend was KRW 3,000 per share after additional annual distribution. This ADR story is not a dividend story; it is a capex and AI-memory access story.&lt;/p&gt;
&lt;h2 id="8-risks"&gt;8. Risks
&lt;/h2&gt;&lt;p&gt;The main risk factors remain memory cyclicality, HBM manufacturing complexity, AI infrastructure slowdown, customer concentration, U.S.-China restrictions, raw materials, water, power, equipment, foreign-exchange volatility and ADS-specific rights. ADS holders exercise voting and dividend rights through the depositary and can face conversion, cancellation or redeposit limitations.&lt;/p&gt;
&lt;h2 id="9-investment-view"&gt;9. Investment View
&lt;/h2&gt;&lt;p&gt;SKHY can make SK Hynix easier to own for U.S. AI-memory investors. It can also make the comparison with Micron&amp;rsquo;s U.S.-listed AI-memory premium more direct.&lt;/p&gt;
&lt;p&gt;But this is not yet an immediate buy signal. The final ADS count, offering price, net proceeds, effective date and first trading date remain unavailable. The event belongs on the watchlist until those terms are filed.&lt;/p&gt;
&lt;h2 id="watch-list"&gt;Watch List
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Watch item&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Next F-1/A or final prospectus&lt;/td&gt;
 &lt;td&gt;ADS count, offering price and net proceeds&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SEC effectiveness&lt;/td&gt;
 &lt;td&gt;Start of real offering timeline&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First &lt;code&gt;SKHY&lt;/code&gt; trading date&lt;/td&gt;
 &lt;td&gt;U.S. investor access becomes observable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRX common vs ADS parity&lt;/td&gt;
 &lt;td&gt;Premium, discount or arbitrage pressure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Offer price relative to KRX&lt;/td&gt;
 &lt;td&gt;Demand strength and near-term flow risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. AI-memory ETF inclusion&lt;/td&gt;
 &lt;td&gt;Potential structural demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q/3Q HBM ASP and shipments&lt;/td&gt;
 &lt;td&gt;Whether the filing narrative converts into earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Custom HBM and eSSD commentary&lt;/td&gt;
 &lt;td&gt;Whether SK Hynix becomes a memory platform, not only an HBM supplier&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Yongin, P&amp;amp;T7 and Indiana milestones&lt;/td&gt;
 &lt;td&gt;Whether capex turns into supply capacity on schedule&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119312526289345/d32785df1a.htm" target="_blank" rel="noopener"
 &gt;SK hynix Form F-1/A, June 30, 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119312526289345/0001193125-26-289345-index.htm" target="_blank" rel="noopener"
 &gt;SK hynix Form F-1/A filing index&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119380526000898/e665622_f6-skhynix.htm" target="_blank" rel="noopener"
 &gt;SK hynix Form F-6, July 1, 2026&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119380526000898/0001193805-26-000898-index.htm" target="_blank" rel="noopener"
 &gt;SK hynix Form F-6 filing index&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.sec.gov/Archives/edgar/data/2120882/000119312526280172/d32785df1.htm" target="_blank" rel="noopener"
 &gt;SK hynix initial Form F-1, June 24, 2026&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>FADU 2Q26 Earnings Preview: Likely Beat, Not a Mega Surprise</title><link>https://koreainvestinsights.com/post/fadu-2q26-earnings-preview-essd-controller-moderate-beat-2026-07-04/</link><pubDate>Sat, 04 Jul 2026 05:00:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/fadu-2q26-earnings-preview-essd-controller-moderate-beat-2026-07-04/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/fadu-sandisk-ai-storage-korea-beta-jeju-semiconductor-2026-05-31/" &gt;FADU as Korea&amp;rsquo;s Sandisk beta&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/fadu-ai-infra-storage-bottleneck-p-q-new-segment-2026-06-02/" &gt;FADU&amp;rsquo;s P, Q and new-segment test&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-semiconductor-top50-two-month-alpha-candidates-samsung-hynix-2026-06-13/" &gt;Korean semiconductor Top 50 two-month alpha candidates&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/post/hbf-hbc-commercialization-stock-trigger-catalyst-calendar-2026-06-28/" &gt;HBF/HBC commercialization calendar&lt;/a&gt;. FADU should be analyzed as an eSSD controller and AI-storage bottleneck name, not as a confirmed HBF beneficiary.&lt;/p&gt;

 &lt;/blockquote&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Result update, July 22, 2026
FADU&amp;rsquo;s preliminary Q2 result was KRW 73.16bn of revenue, KRW 16.27bn of operating profit and a 22.2% operating margin. Revenue fell just below this preview&amp;rsquo;s KRW 75bn bull threshold, while operating profit exceeded the KRW 15.4bn bull-case upper bound. The post-mortem and Gen6/QLC moat assessment continue in &lt;a class="link" href="https://koreainvestinsights.com/post/fadu-2q26-earnings-surprise-gen6-essd-controller-narrow-moat-2026-07-22/" &gt;FADU&amp;rsquo;s Q2 Surprise and Its Real Moat&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;My FADU 2Q26 estimate is KRW 69-72B revenue and KRW 11-12.5B operating profit. The midpoint is KRW 70.5B revenue, KRW 11.8B operating profit and 16.7% operating margin.&lt;/p&gt;
&lt;p&gt;The local consensus snapshot as of July 3, 2026 shows 2Q26 revenue of KRW 60.5B and operating profit of KRW 9.3B. The midpoint estimate is 16.5% above revenue consensus and 26.9% above operating-profit consensus. A beat is likely, but this is not a mega-surprise bar.&lt;/p&gt;
&lt;p&gt;The key is not the gross amount of 2Q contract disclosures. Major supply contracts disclosed during the quarter total about KRW 150.8B. A mechanical pro-rata allocation would imply roughly KRW 35.4B of 2Q exposure. But amended disclosures in June pushed some contract end dates and payment terms later. Treating the full pro-rata amount as incremental 2Q revenue would be too aggressive. A more reasonable 2Q incremental recognition range is KRW 9.5-12.5B.&lt;/p&gt;
&lt;p&gt;The stock call is Wait. FADU closed at KRW 79,600 on July 3, 2026, implying roughly KRW 3.98T market cap. Against FY2026E revenue of KRW 301.5B and operating profit of KRW 44.8B, that is about 13.2x sales and 89x operating profit. At this valuation, a good 2Q print is not enough. Investors need evidence of second-half delivery normalization, backlog conversion and controller-mix durability.&lt;/p&gt;
&lt;h2 id="0-setup"&gt;0. Setup
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Detail&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Company&lt;/td&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Ticker&lt;/td&gt;
 &lt;td&gt;440110 / KOSDAQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Analysis date&lt;/td&gt;
 &lt;td&gt;2026-07-04 KST&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price basis&lt;/td&gt;
 &lt;td&gt;July 3, 2026 close, KRW 79,600, local KR database&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Consensus basis&lt;/td&gt;
 &lt;td&gt;July 3, 2026 local consensus snapshot, 2Q26 revenue KRW 60.5B, OP KRW 9.3B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Main question&lt;/td&gt;
 &lt;td&gt;How much of the 2Q contract disclosures can actually be recognized as 2Q revenue?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stance&lt;/td&gt;
 &lt;td&gt;Wait; conditional entry after proof&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This preview uses public filings, public media reports and local market data. It does not use private management color.&lt;/p&gt;
&lt;h2 id="1-baseline-from-1q26"&gt;1. Baseline From 1Q26
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 59.54B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 7.69B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 10.21B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SSD controller revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 47.9B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SSD controller mix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;80.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SSD finished-product revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 9.9B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SSD finished-product mix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The important point is mix. FADU earns a higher-quality multiple only if it is seen as an eSSD controller and firmware bottleneck company, not as a finished-SSD assembler. The 1Q controller mix of 80.4% is therefore more important than the headline profit turn.&lt;/p&gt;
&lt;p&gt;The reported 1Q gross profit was KRW 34.65B. Against revenue of KRW 59.54B, this implies a 58.2% gross margin. Subtracting KRW 7.69B operating profit gives implied SG&amp;amp;A of about KRW 26.96B.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Formula&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q gross margin&lt;/td&gt;
 &lt;td&gt;34.65 / 59.54&lt;/td&gt;
 &lt;td style="text-align: right"&gt;58.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Implied 1Q SG&amp;amp;A&lt;/td&gt;
 &lt;td&gt;34.65 - 7.69&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 26.96B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q operating margin&lt;/td&gt;
 &lt;td&gt;7.69 / 59.54&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-backlog-and-2026-delivery-base"&gt;2. Backlog And 2026 Delivery Base
&lt;/h2&gt;&lt;p&gt;FADU&amp;rsquo;s 1Q-end backlog was reported at $116.208M, roughly KRW 175.9B. The 2026 delivery portion was $95.131M, roughly KRW 144.0B. Adding 1Q revenue gives a visible annual revenue base above KRW 200B, before the rest of the 2Q disclosures.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Amount&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q-end backlog&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$116.208M&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW equivalent&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 175.9B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 delivery portion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$95.131M&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW equivalent&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 144.0B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q26 revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 59.54B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The caveat is timing. A 2026 delivery commitment does not mean 2Q recognition.&lt;/p&gt;
&lt;h2 id="3-2q-contract-disclosures"&gt;3. 2Q Contract Disclosures
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Disclosure timing&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Product&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Contract amount&lt;/th&gt;
 &lt;th&gt;Contract period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Pro-rata 2Q exposure&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Apr 13&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Enterprise SSD controller&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 15.196B&lt;/td&gt;
 &lt;td&gt;Apr 10-Oct 9&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 6.8B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May 6&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Enterprise SSD controller&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 50.012B&lt;/td&gt;
 &lt;td&gt;May 5-Dec 8 after amendment&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 13.1B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May 22&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Enterprise SSD controller&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 28.684B&lt;/td&gt;
 &lt;td&gt;May 21-Dec 11&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 5.7B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May 28&lt;/td&gt;
 &lt;td style="text-align: right"&gt;eSSD controller&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 46.5B&lt;/td&gt;
 &lt;td&gt;May 28-Jan 1, 2027&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 7.2B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May 29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Finished SSD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 10.373B&lt;/td&gt;
 &lt;td&gt;May 28-Oct 11&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 2.6B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 150.8B&lt;/td&gt;
 &lt;td&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 35.4B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The June amendments matter. The KRW 50B contract end date moved from Nov. 13 to Dec. 8, and payment terms moved from the next Friday after delivery to within 30 days after delivery. A separate KRW 20.3B controller contract was also extended from June 12 to July 17. These do not prove demand weakness, but they do weaken the 2Q cutoff.&lt;/p&gt;
&lt;h2 id="4-revenue-model"&gt;4. Revenue Model
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Component&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Estimate&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;1Q revenue baseline&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 59.5B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Incremental 2Q recognition from new contracts&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 9.5-12.5B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Finished products and other movement&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+/- KRW 1.0B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q26 revenue estimate&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 69-72B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Midpoint&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 70.5B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is more conservative than the earlier June framework that looked for KRW 90-105B of 2Q revenue. The reason is new information: amended disclosures increased the risk that part of the revenue recognition slips into 3Q.&lt;/p&gt;
&lt;h2 id="5-operating-profit-model"&gt;5. Operating Profit Model
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Segment&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue assumption&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Gross margin assumption&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Gross profit&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Controllers&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 57.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;61%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 35.1B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Finished SSD / modules&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 11.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.7B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Other&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;50%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1.0B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Total&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 70.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~53.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 37.7B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Using KRW 25.9B of SG&amp;amp;A, the midpoint operating profit is about KRW 11.8B, or a 16.7% margin. The implied 2Q gross margin is below 1Q&amp;rsquo;s 58.2% to reflect finished-product mix and cutoff risk.&lt;/p&gt;
&lt;h2 id="6-scenario-table"&gt;6. Scenario Table
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Operating profit&lt;/th&gt;
 &lt;th style="text-align: right"&gt;OPM&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bear&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 62-65B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 9-10.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;14.5-16.2%&lt;/td&gt;
 &lt;td&gt;In line to slight beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 69-72B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 11-12.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15.9-17.4%&lt;/td&gt;
 &lt;td&gt;Clear beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bull&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 78-83B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 14.5-17B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18.6-20.5%&lt;/td&gt;
 &lt;td&gt;Major surprise&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Base is the central case. Bull requires faster shipment, limited impact from amendments and controller mix near 80%.&lt;/p&gt;
&lt;h2 id="7-consensus-bar"&gt;7. Consensus Bar
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Consensus&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Midpoint estimate&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Delta&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Delta %&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 60.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 70.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 10.0B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 9.3B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 11.8B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 2.5B&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;OPM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;16.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.3ppt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A meaningful beat needs revenue above KRW 70B and OP above KRW 11.5B. A true estimate-reset print likely needs revenue above KRW 75B and OP above KRW 14B.&lt;/p&gt;
&lt;h2 id="8-price-flow-and-valuation"&gt;8. Price, Flow And Valuation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;July 3 close&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 79,600&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5D return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-27.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;60D return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+57.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Estimated market cap&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 3.98T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D foreign net buying&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW 227.3B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D institutional net selling&lt;/td&gt;
 &lt;td style="text-align: right"&gt;~KRW -117.6B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign ownership&lt;/td&gt;
 &lt;td style="text-align: right"&gt;24.11% on Jun 8 to 28.25% on Jul 3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Foreign flow is the positive signal. Institutional confirmation is not there yet. On valuation, FY2026E revenue of KRW 301.5B and OP of KRW 44.8B imply about 13.2x sales and 89x operating profit. This is not cheap. It requires continued proof.&lt;/p&gt;
&lt;h2 id="9-entry-catalysts-and-falsifiers"&gt;9. Entry, Catalysts And Falsifiers
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Action rule&lt;/th&gt;
 &lt;th&gt;Requirement&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Earnings-confirmation entry&lt;/td&gt;
 &lt;td&gt;2Q revenue above KRW 70B, OP above KRW 11.5B, controller mix near 80%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Strong-beat entry&lt;/td&gt;
 &lt;td&gt;Revenue above KRW 75B, OP above KRW 14B&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Valuation entry&lt;/td&gt;
 &lt;td&gt;If the print is only moderate and the stock derates toward below KRW 3.0T market cap or below 10x FY2026E sales&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Catalysts are the 2Q result, 2Q-end backlog, 2026 delivery update, controller mix, normalization of amended contracts and 3Q revenue acceleration.&lt;/p&gt;
&lt;p&gt;Falsifiers are revenue below KRW 65B, OP below KRW 10B, controller mix below 75%, gross margin below 50%, further delays in the KRW 50B or KRW 20.3B contracts, no 3Q backlog-conversion guide, or continued dependence on two customers near 80%.&lt;/p&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;FADU probably beats consensus in 2Q26. My midpoint is KRW 70.5B revenue and KRW 11.8B operating profit. But this is a moderate beat, not a clean buy signal at any price.&lt;/p&gt;
&lt;p&gt;FADU remains a valid AI-storage bottleneck candidate. The stock needs stricter evidence now: 2Q revenue above KRW 70B, controller mix near 80%, and proof that delayed contracts convert into 2H revenue.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://v.daum.net/v/20260703070207071" target="_blank" rel="noopener"
 &gt;Daum News&lt;/a&gt;, &lt;a class="link" href="https://www.thelec.kr/news/articleView.html?idxno=58121" target="_blank" rel="noopener"
 &gt;The Elec&lt;/a&gt;, &lt;a class="link" href="https://www.dailyinvest.kr/news/articleView.html?idxno=70694" target="_blank" rel="noopener"
 &gt;DailyInvest&lt;/a&gt;, &lt;a class="link" href="https://kind.krx.co.kr/common/disclsviewer.do?acptno=20260512000812&amp;amp;docno=&amp;amp;method=search&amp;amp;viewerhost=" target="_blank" rel="noopener"
 &gt;KIND filing&lt;/a&gt;, &lt;a class="link" href="https://www.thebell.co.kr/front/newsview.asp?key=202605130923581040107666" target="_blank" rel="noopener"
 &gt;The Bell&lt;/a&gt;, &lt;a class="link" href="https://www.datatooza.com/article/20260413101307147852ef3fffef_80" target="_blank" rel="noopener"
 &gt;Datatooza&lt;/a&gt;, local KR price, flow and consensus databases as of July 3, 2026.&lt;/p&gt;</description></item><item><title>H1 2026 Review: AI Infrastructure Bottlenecks Broadened, but the Market Did Not</title><link>https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/</link><pubDate>Tue, 30 Jun 2026 22:35:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/h1-2026-ai-infra-bottleneck-korea-narrow-market-postmortem-2026-06-30/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context
This is a half-year follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-2q26-preview-micron-surprise-erased-core-op-hbm-2026-06-29/" &gt;Samsung 2Q26 preview&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 earnings&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/nvidia-earnings-elasticity-hbm-cycle-samsung-hynix-2026-06-28/" &gt;the NVIDIA elasticity framework for Samsung and SK Hynix&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/kospi-benchmark-hard-to-beat-narrow-market-monte-carlo-2026-06-20/" &gt;how hard it was to beat the KOSPI benchmark&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/kospi-disparity60-overheat-framework-partial-trim-2026-06-21/" &gt;the KOSPI 60-day disparity framework&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/real-money-flow-framework-korea-institution-quality-2026-06-03/" &gt;the Real Money Flow framework&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/ai-supercycle-midgame-rate-risk-yellow-not-red-2026-06-06/" &gt;the AI supercycle midgame note&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;H1 2026 looked like a broad risk-on period from the index level. Under the surface, it was much narrower. Korea was a KOSPI megacap crowding market. The U.S. was an AI infrastructure market where leadership moved from megacap AI platforms into memory, storage, servers, semiconductor equipment and power. Neither market was a broad “everything went up” market.&lt;/p&gt;
&lt;p&gt;The headline Korean index was extraordinary. KOSPI rose from 4,214.17 on December 30, 2025 to 8,476.48 on June 30, 2026, up 101.1%. KOSDAQ fell from 925.47 to 916.18, down 1.0%. Across roughly 2,700 Korean stocks, 856 rose and 1,788 fell. The median stock return was -12.6%. The index doubled, but the average stock was in a weak market.&lt;/p&gt;
&lt;p&gt;The U.S. story was also more specific than the index suggests. The S&amp;amp;P 500 gained 9.2% and Nasdaq gained 11.1%, but the real story was AI infrastructure bottlenecks. SanDisk, Micron, Western Digital, Dell, Marvell, ARM, Applied Materials and Lam Research were much stronger than NVIDIA. The market was buying the places where the AI factory actually bottlenecks.&lt;/p&gt;
&lt;p&gt;The H2 strategy is not simply “find cheap stocks.” The better rule is to buy proven bottlenecks with real revenue and contracts on pullbacks. Option-like equipment, gaming and biotech event names should be sized smaller. Continuous-revenue bottlenecks in memory, eSSD, MLCC, substrates, power, shipbuilding, defense and large biotech platforms deserve more attention.&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;One-Line Conclusion&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 H1 was risk-on, but it was not broad. Korea was dominated by Samsung Electronics and SK Hynix crowding, while the U.S. rewarded the physical bottlenecks beneath AI compute. H2 is about separating recurring-revenue bottlenecks from option-like stories and leveraged flow risk.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="0-data-basis-and-reconciliation"&gt;0. Data Basis and Reconciliation
&lt;/h2&gt;&lt;p&gt;The data basis is:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Area&lt;/th&gt;
 &lt;th&gt;Basis&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea indices and stocks&lt;/td&gt;
 &lt;td&gt;June 30, 2026 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. indices and stocks&lt;/td&gt;
 &lt;td&gt;June 29, 2026 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea flows&lt;/td&gt;
 &lt;td&gt;KR local DB, Kiwoom amount basis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. stock returns&lt;/td&gt;
 &lt;td&gt;US local DB and yfinance adjusted close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Secondary inputs&lt;/td&gt;
 &lt;td&gt;Naver Finance, local screeners and session flow analysis&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Some figures differ slightly by dataset and reference date. For example, the final KOSPI return is +101.1% using the June 30 close, while some local basket snapshots show +96.7%, and the June 23 interim review showed +90.4%. U.S. AI stock returns also differ depending on adjusted-close treatment and vendor handling.&lt;/p&gt;
&lt;p&gt;This post reconciles those differences as follows.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Data type&lt;/th&gt;
 &lt;th&gt;How it is used&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;June 30 Korea close&lt;/td&gt;
 &lt;td&gt;Main H1 performance benchmark&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;June 29 U.S. close&lt;/td&gt;
 &lt;td&gt;Main U.S. H1 performance benchmark&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;June 23 interim snapshot&lt;/td&gt;
 &lt;td&gt;Supporting evidence for breadth and buyer structure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted-close versus local DB differences&lt;/td&gt;
 &lt;td&gt;Used directionally when conclusions are unchanged&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Less-verified single figures&lt;/td&gt;
 &lt;td&gt;Treated as H2 watch items, not as core proof&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The point is not false precision. The conclusion is stable: indices were strong, breadth was weak, and AI infrastructure returns moved from GPUs into memory, storage, servers, equipment and power.&lt;/p&gt;
&lt;h2 id="1-index-performance-kospi-doubled-kosdaq-fell"&gt;1. Index Performance: KOSPI Doubled, KOSDAQ Fell
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Market&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Basis&lt;/th&gt;
 &lt;th style="text-align: right"&gt;H1 return&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4,214.17 on 2025-12-30 to 8,476.48 on 2026-06-30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+101.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;925.47 to 916.18&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;S&amp;amp;P 500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-12-31 to 2026-06-29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nasdaq&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-12-31 to 2026-06-29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Russell 2000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-12-31 to 2026-06-29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SMH semiconductor ETF&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-12-31 to 2026-06-29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+75.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SOXX semiconductor ETF&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025-12-31 to 2026-06-29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+104.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;At the index level, Korea was spectacular. KOSPI doubled in six months. But KOSDAQ fell. Within the same Korean market, large-cap AI infrastructure and growth stocks lived in different worlds.&lt;/p&gt;
&lt;p&gt;The U.S. index story also understates the internal rotation. The S&amp;amp;P 500 and Nasdaq were up around low double digits, while semiconductor ETFs were up far more. SOXX gained 104.2%, close to the KOSPI’s gain, showing how concentrated the global AI hardware trade became.&lt;/p&gt;
&lt;h2 id="2-korea-the-index-doubled-the-median-stock-fell"&gt;2. Korea: The Index Doubled, the Median Stock Fell
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Observed Korean stocks&lt;/td&gt;
 &lt;td style="text-align: right"&gt;roughly 2,700&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;H1 advancers&lt;/td&gt;
 &lt;td style="text-align: right"&gt;856&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;H1 decliners&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,788&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Median stock return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-12.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI H1 return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+101.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ H1 return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is an unusual combination. The index was in a historic bull market, while the median stock was weak. That is why the lived investor experience was so different from the index headline.&lt;/p&gt;
&lt;p&gt;The main H1 leaders were:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;H1 return&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electro-Mechanics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+756%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jusung Engineering&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+626%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+361%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+307%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jeju Semiconductor&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+294%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daeduck Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+226%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Simmtech&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+201%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+179%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The Korean market’s structure was:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Samsung Electronics and SK Hynix
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ Samsung Electro-Mechanics, MLCC and FC-BGA
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ substrates, PCB and semi equipment
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ power equipment, industrials and some financials
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;KOSDAQ, many biotech names, platforms and parts of secondary batteries lagged. Korea was strong, but most Korean stocks were not.&lt;/p&gt;
&lt;h2 id="3-korea-was-not-a-foreign-buying-market"&gt;3. Korea Was Not a Foreign-Buying Market
&lt;/h2&gt;&lt;p&gt;YTD flow data on the local Kiwoom amount basis:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Investor group&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD cumulative&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreigners&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 142T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Institutions&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 45T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Retail&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 77.7T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Financial investment&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 75.4T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investment trusts&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.7T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Private funds&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 3.8T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pension funds and others&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 7.0T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The key point is that Korea was not a simple foreign-buying market. Foreigners sold Samsung Electronics and SK Hynix heavily. Domestic retail, financial investment accounts and institutions absorbed the flow.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD return&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Foreign cumulative flow&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;around +160%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 72.6T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;around +291%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 57.1T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Square&lt;/td&gt;
 &lt;td style="text-align: right"&gt;around +333%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 7.6T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics preferred&lt;/td&gt;
 &lt;td style="text-align: right"&gt;around +125%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 3.2T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Foreigners sold the core megacaps while those stocks rallied. That is not a normal foreign-led rally. It was a domestic absorption market.&lt;/p&gt;
&lt;p&gt;Foreigners did not sell everything. They rotated into select bottlenecks and global-fund-friendly assets such as Samsung Electro-Mechanics, Doosan Enerbility, Samsung SDI, FADU, Hanwha Ocean and Sanil Electric. The right description is not “foreigners bought Korea.” It is “foreigners sold the largest memory megacaps, domestic liquidity absorbed them, and foreigners reallocated into select second-layer bottlenecks.”&lt;/p&gt;
&lt;h2 id="4-the-hidden-buyer-was-financial-investment-not-long-only-real-money"&gt;4. The Hidden Buyer Was Financial Investment, Not Long-Only Real Money
&lt;/h2&gt;&lt;p&gt;Institutional buying was not the same as long-only accumulation. The strongest institutional subcategory was financial investment, not investment trusts, private funds or pension funds.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Flow category&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Financial investment&lt;/td&gt;
 &lt;td&gt;ETF, derivatives, program and rebalance-linked flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Investment trusts&lt;/td&gt;
 &lt;td&gt;More likely to represent long-only fund demand, but weak in H1&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Private funds&lt;/td&gt;
 &lt;td&gt;Mixed thematic, long-short and event-driven flow, but weak cumulatively&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pension funds and others&lt;/td&gt;
 &lt;td&gt;Long-term real money, but net selling in H1&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Retail&lt;/td&gt;
 &lt;td&gt;Key absorber of megacap AI infrastructure flow and leverage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This matters. Real money can sustain a trend. Financial-investment and program flow can be powerful, but it is more vulnerable to futures expiry, ETF caps, derivative positioning and leverage mechanics.&lt;/p&gt;
&lt;p&gt;H1 Korea was both a fundamental rerating market and a derivatives/ETF market. Confusing the two leads to repeated mistakes.&lt;/p&gt;
&lt;h2 id="5-kosdaq-lacked-buyers-not-stories"&gt;5. KOSDAQ Lacked Buyers, Not Stories
&lt;/h2&gt;&lt;p&gt;KOSDAQ had plenty of stories: biotech, robotics, CPO, SOCAMM, semiconductor equipment, gaming and AI software. The index still lagged because the buyer base was weak.&lt;/p&gt;
&lt;p&gt;The June 23 interim snapshot was already clear:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Index&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD&lt;/th&gt;
 &lt;th style="text-align: right"&gt;From high&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+90.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-10.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-27.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The final June 30 structure was the same: KOSPI +101.1%, KOSDAQ -1.0%.&lt;/p&gt;
&lt;p&gt;KOSDAQ struggled because:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Retail capital was tied up absorbing Samsung Electronics, SK Hynix and large AI infrastructure names.&lt;/li&gt;
&lt;li&gt;Institutional real money did not buy KOSDAQ broadly.&lt;/li&gt;
&lt;li&gt;Foreign buying was narrow, focused on names such as FADU, Jeju Semiconductor and Simmtech.&lt;/li&gt;
&lt;li&gt;Rates and FX pressure weighed on biotech and unprofitable growth multiples.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The condition for KOSDAQ recovery is:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;KOSDAQ recovery =
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;retail capital returning
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ institutional real-money buying
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ Samsung Electronics and SK Hynix moving sideways rather than falling
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ KOSDAQ earnings revisions stabilizing
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;If Samsung and SK Hynix crash, money does not automatically rotate into KOSDAQ. Risk-off comes first. The best environment for KOSDAQ is not megacap collapse. It is megacap consolidation.&lt;/p&gt;
&lt;h2 id="6-the-us-market-was-not-just-nvidia"&gt;6. The U.S. Market Was Not Just NVIDIA
&lt;/h2&gt;&lt;p&gt;The real U.S. signal was not “AI went up.” It was:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;The physical bottlenecks after the GPU outperformed the GPU leader.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;p&gt;Representative H1 return ranges:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;H1 return range&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;SanDisk&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+645-764%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+263-301%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Western Digital&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+247-279%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Intel&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+235-257%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Dell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+224-232%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Marvell&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+211-227%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ARM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+200-214%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Applied Materials&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+171%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AMD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+152%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Lam Research&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+141%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Big Tech was mixed:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;H1 return&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Google&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+13%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Amazon&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Apple&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3-5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Broadcom&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Meta&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-15%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Microsoft&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-22% to -23%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Oracle&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-25%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Palantir&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-31%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The market bought the points where the AI factory bottlenecks:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;GPU
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ HBM and DRAM
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ eSSD and storage
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ servers and racks
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ optical and interconnect
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ semiconductor equipment
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ power and cooling
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;This is directly relevant to Korea. HBM maps to Samsung and SK Hynix. eSSD and NAND map to Samsung, FADU and Jeju Semiconductor. MLCC and FC-BGA map to Samsung Electro-Mechanics and substrate names. Power maps to Doosan Enerbility and electrical equipment. Equipment maps to HPSP, TES, VM and Jusung.&lt;/p&gt;
&lt;h2 id="7-macro-was-not-a-pure-tailwind"&gt;7. Macro Was Not a Pure Tailwind
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Asset or indicator&lt;/th&gt;
 &lt;th style="text-align: right"&gt;H1 move&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Dollar index&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;USD/KRW&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. 10-year yield&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4.16% to 4.39%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;WTI oil&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+24.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Gold&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bitcoin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-33.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This was not a clean multiple-expansion macro environment. The dollar rose, the won weakened, U.S. yields rose and oil rose. Gold and Bitcoin fell.&lt;/p&gt;
&lt;p&gt;Stocks rose anyway because the AI infrastructure earnings cycle overwhelmed macro pressure. That is important for H2. If rates, the dollar and oil all rise together again, the most crowded bottleneck stocks can suffer multiple compression even if demand remains strong.&lt;/p&gt;
&lt;p&gt;The realistic risk is:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;AI demand stays strong
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ earnings remain good
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ rates, dollar and oil remain firm
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ valuation and leveraged flows crack first
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;That is the same stock-elasticity problem discussed in the NVIDIA/Samsung/SK Hynix framework.&lt;/p&gt;
&lt;h2 id="8-non-consensus-truth-1-leadership-moved-from-compute-to-memory"&gt;8. Non-Consensus Truth 1: Leadership Moved From Compute to Memory
&lt;/h2&gt;&lt;p&gt;The first non-consensus point is U.S. AI leadership. NVIDIA is still the center of AI. But incremental stock performance moved into Micron, SanDisk, Western Digital, Dell, Marvell and ARM.&lt;/p&gt;
&lt;p&gt;That does not mean compute no longer matters. It means the marginal return moved down the stack.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Is this company a cost item in AI capex?&lt;/td&gt;
 &lt;td&gt;It can be pressured when customers optimize costs.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is this company a bottleneck item in AI capex?&lt;/td&gt;
 &lt;td&gt;It can reprice if customers need supply.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is this company only a narrative?&lt;/td&gt;
 &lt;td&gt;It is vulnerable during catalyst gaps.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Cost items get pressured. Bottlenecks rerate. Narratives require timing discipline.&lt;/p&gt;
&lt;h2 id="9-non-consensus-truth-2-the-real-top-came-before-the-index-high"&gt;9. Non-Consensus Truth 2: The Real Top Came Before the Index High
&lt;/h2&gt;&lt;p&gt;Breadth weakened before the index did. The June index high was driven by Samsung Electronics, SK Hynix and a small set of AI infrastructure names. KOSDAQ and the median stock had already weakened earlier.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Weak median stock
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ weak KOSDAQ
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;+ strong KOSPI memory megacaps
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;= index high and weak lived market at the same time
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;For H2, the key indicators are breadth, ADR, KOSDAQ/KOSPI relative strength, megacap memory trading-value share, and KOSDAQ retail/institutional joint buying.&lt;/p&gt;
&lt;h2 id="10-non-consensus-truth-3-multiples-moved-before-fundamentals-fully-arrived"&gt;10. Non-Consensus Truth 3: Multiples Moved Before Fundamentals Fully Arrived
&lt;/h2&gt;&lt;p&gt;Samsung, SK Hynix, Micron and Samsung Electro-Mechanics all had strong fundamentals. But stock returns moved faster than near-term earnings. The market changed the multiple.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Old frame:
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;memory = commodity cycle = peak-earnings discount
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;H1 frame:
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;memory = AI infrastructure bottleneck = supply shortage plus contract premium
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The H2 risk is not just bad earnings. It is good earnings with lower multiples if the market starts treating memory as commodity again. Investors should watch EPS revision speed, beat magnitude and FY2027 visibility, not just the next quarterly beat.&lt;/p&gt;
&lt;h2 id="11-non-consensus-truth-4-leverage-plumbing-amplified-tail-risk"&gt;11. Non-Consensus Truth 4: Leverage Plumbing Amplified Tail Risk
&lt;/h2&gt;&lt;p&gt;Korean memory megacaps were affected not only by fundamentals, but also by leverage plumbing: 2x and 3x Samsung/SK Hynix ETFs listed offshore, domestic leverage and inverse products, futures and program flow.&lt;/p&gt;
&lt;p&gt;Simplified upside loop:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;stock up
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ leveraged ETF rebalancing buy
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ futures and program support
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ more megacap upside
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ more momentum capital
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Downside loop:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;stock down
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ leveraged ETF rebalancing sell
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ futures and program pressure
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ retail margin and forced-selling risk
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;→ larger drawdown
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Korea can therefore move more violently than Micron even when the fundamental direction is similar.&lt;/p&gt;
&lt;h2 id="12-non-consensus-truth-5-supply-moved-from-strong-hands-to-weaker-hands"&gt;12. Non-Consensus Truth 5: Supply Moved From Strong Hands to Weaker Hands
&lt;/h2&gt;&lt;p&gt;The important question is not only who sold. It is who bought.&lt;/p&gt;
&lt;p&gt;H1 saw foreigners and some institutional accounts reduce exposure in memory megacaps while retail and financial-investment accounts absorbed the flow.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Positive read&lt;/th&gt;
 &lt;th&gt;Negative read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Domestic liquidity is strong&lt;/td&gt;
 &lt;td&gt;Shares may have moved into weaker leveraged hands&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI held despite foreign selling&lt;/td&gt;
 &lt;td&gt;Forced-selling risk rises if margin unwinds&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Large-cap buyer base widened&lt;/td&gt;
 &lt;td&gt;It may also represent distribution from better-informed accounts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;H2 monitoring should combine foreign cash flow, retail margin, deposits, ETF balances, and program flow.&lt;/p&gt;
&lt;h2 id="13-non-consensus-truth-6-foreigners-bought-globally-explainable-korean-assets"&gt;13. Non-Consensus Truth 6: Foreigners Bought Globally Explainable Korean Assets
&lt;/h2&gt;&lt;p&gt;Foreigners did not buy every Korean theme. The stocks they bought were globally explainable.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;YTD foreign net buying&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electro-Mechanics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 2.28T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Doosan Enerbility&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1.71T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Celltrion&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1.52T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung SDI&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1.33T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Doosan&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1.11T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;APR&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 1.08T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai Rotem&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.85T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LG Energy Solution&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.85T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.80T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hanwha Ocean&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.76T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Common features:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;power, nuclear, shipbuilding and defense
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;global quality industrials
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;large battery names
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;large biotech platforms
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;direct AI infrastructure bottlenecks
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;For H2, foreign long-only capital is more likely to buy assets it can explain in an investment committee than obscure small-cap stories.&lt;/p&gt;
&lt;h2 id="14-non-consensus-truth-7-revenue-pattern-beat-company-quality"&gt;14. Non-Consensus Truth 7: Revenue Pattern Beat Company Quality
&lt;/h2&gt;&lt;p&gt;In H1, the quality of the revenue pattern mattered more than abstract company quality.&lt;/p&gt;
&lt;p&gt;Weak revenue-pattern names:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Type&lt;/th&gt;
 &lt;th&gt;Weakness&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Equipment&lt;/td&gt;
 &lt;td&gt;Order gaps can create earnings gaps.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Gaming&lt;/td&gt;
 &lt;td&gt;Revenue visibility breaks around launch cycles.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Biotech event stocks&lt;/td&gt;
 &lt;td&gt;Trial and licensing gaps create narrative gaps.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Some robotics and AI software&lt;/td&gt;
 &lt;td&gt;Expectations move before revenue.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Strong revenue-pattern names:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Type&lt;/th&gt;
 &lt;th&gt;Strength&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory&lt;/td&gt;
 &lt;td&gt;Pricing and volume improved together.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;eSSD and storage&lt;/td&gt;
 &lt;td&gt;AI server storage bottleneck rerating.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MLCC and substrates&lt;/td&gt;
 &lt;td&gt;Server and package bottlenecks.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Power equipment&lt;/td&gt;
 &lt;td&gt;Data-center power demand plus policy capital.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shipbuilding&lt;/td&gt;
 &lt;td&gt;Long backlog visibility.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Defense&lt;/td&gt;
 &lt;td&gt;Export backlog and government budgets.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Large biotech platforms&lt;/td&gt;
 &lt;td&gt;Contracts, production and platform value are explainable.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The H2 rule is simple:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Size option-like names smaller. Size recurring-revenue bottlenecks larger.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="15-bonus-hypothesis-korea-became-a-leveraged-us-ai-proxy"&gt;15. Bonus Hypothesis: Korea Became a Leveraged U.S. AI Proxy
&lt;/h2&gt;&lt;p&gt;Korea’s market identity changed in H1. Historically, Samsung and SK Hynix were proxies for EM, China and global IT inventory cycles. In H1 2026, the linkage to U.S. AI capex became much stronger.&lt;/p&gt;
&lt;p&gt;Session analysis suggested a stronger relationship between U.S. prior-day AI/semi moves and Korea next-day memory megacaps, including an interim correlation estimate of 0.42. That figure needs formal revalidation, but the price action is consistent with the direction.&lt;/p&gt;
&lt;p&gt;The issue is diversification. If Korea behaves like a leveraged U.S. AI capex proxy, domestic buffers weaken when the global AI factor cools.&lt;/p&gt;
&lt;p&gt;Another watch item is capex. Session analysis classified large SK Hynix capex commitment as a peak-confidence checkpoint. The exact amount and scope should be rechecked against filings before it becomes a core argument. But historically, peak capex announcements often arrive when demand looks strongest.&lt;/p&gt;
&lt;p&gt;Key H2 capex questions:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Are customer contracts supporting capex?&lt;/td&gt;
 &lt;td&gt;Lowers supply-glut risk.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Does capex impair FCF?&lt;/td&gt;
 &lt;td&gt;Earnings can look good while cash flow weakens.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Does supply arrive near demand peak?&lt;/td&gt;
 &lt;td&gt;Raises 2027 cycle risk.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is capex HBM and AI-memory focused?&lt;/td&gt;
 &lt;td&gt;Must be separated from commodity DRAM oversupply risk.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="16-h2-checklist"&gt;16. H2 Checklist
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Checkpoint&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26 Core OP&lt;/td&gt;
 &lt;td&gt;Core profit power matters more than reported headline OP.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix 2Q and 3Q HBM allocation&lt;/td&gt;
 &lt;td&gt;Determines whether the crowded winner can reaccelerate.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron FY4Q26 guide delivery&lt;/td&gt;
 &lt;td&gt;Tests whether U.S. AI memory premium can persist.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM and NAND pricing&lt;/td&gt;
 &lt;td&gt;Confirms margin defense beyond HBM.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ retail and institutional buying&lt;/td&gt;
 &lt;td&gt;Needed for KOSDAQ recovery.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Financial-investment and program flow&lt;/td&gt;
 &lt;td&gt;Real plumbing behind megacap volatility.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;USD/KRW, U.S. 10-year yields and oil&lt;/td&gt;
 &lt;td&gt;Put a ceiling on AI bottleneck multiples.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI capex contracts and FCF&lt;/td&gt;
 &lt;td&gt;Shows whether demand is paid for by cash flow, not just narrative.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="17-h2-strategy"&gt;17. H2 Strategy
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Area&lt;/th&gt;
 &lt;th&gt;Strategy&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;AI infrastructure core&lt;/td&gt;
 &lt;td&gt;Keep exposure, but buy pullbacks rather than chasing.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory megacaps&lt;/td&gt;
 &lt;td&gt;Samsung is the unreflected-alpha candidate. SK Hynix is the reacceleration candidate after pullbacks.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM lower-stack bottlenecks&lt;/td&gt;
 &lt;td&gt;Buy only where orders, margins and customer qualification are visible.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ&lt;/td&gt;
 &lt;td&gt;Do not buy the whole index first. Start with semiconductor equipment, test and profitable growth names.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Option-like equipment, gaming and biotech&lt;/td&gt;
 &lt;td&gt;Smaller sizing due to catalyst-gap risk.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Global quality cyclicals&lt;/td&gt;
 &lt;td&gt;Continue to monitor shipbuilding, defense, power, nuclear and large biotech platforms.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The H2 strategy in one sentence:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Keep AI infrastructure core exposure, reduce option-like positions, and shift toward recurring-revenue bottlenecks and Korean quality cyclicals that global capital can explain.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="18-final-view"&gt;18. Final View
&lt;/h2&gt;&lt;p&gt;The crowd may remember H1 2026 as the best Korea/semiconductor supercycle year. That is not wrong. KOSPI doubled and memory, substrates, equipment and power leaders exploded.&lt;/p&gt;
&lt;p&gt;But the more important story is:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;leadership became extremely narrow
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;multiples moved before fundamentals fully arrived
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;foreigners sold the core megacaps while domestic liquidity absorbed them
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;financial-investment and ETF/program flows amplified the market
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;KOSDAQ lacked buyers, not stories
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;U.S. AI leadership moved from GPU into memory, storage, servers and power
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The H2 question is not only “how much more upside is left?” The better questions are:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Does stock elasticity remain positive even after strong earnings?&lt;/li&gt;
&lt;li&gt;Can domestic liquidity keep absorbing foreign selling?&lt;/li&gt;
&lt;li&gt;Do KOSDAQ buyers return?&lt;/li&gt;
&lt;li&gt;Do AI bottlenecks convert from narrative into revenue and contracts?&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;H1 was risk-on, but it was not broad. It was a compressed leadership market. H2 requires more precision: real bottlenecks over cheap stocks, recurring revenue over one-off catalysts, and Core OP plus FCF over headline earnings.&lt;/p&gt;
&lt;h2 id="data-note"&gt;Data Note
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Korea data: KR local DB, Kiwoom investor flow data and Naver Finance support as of the June 30, 2026 close.&lt;/li&gt;
&lt;li&gt;U.S. data: US local DB and yfinance adjusted close as of the June 29, 2026 close.&lt;/li&gt;
&lt;li&gt;Interim snapshots: June 23 KOSPI/KOSDAQ relative performance and session flow analysis.&lt;/li&gt;
&lt;li&gt;U.S. June 30 regular-session results are not included.&lt;/li&gt;
&lt;li&gt;Less-verified capex single figures are treated as H2 monitoring items rather than core proof.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>Samsung 2Q26 Preview: Micron’s Surprise Has Been Erased, Core OP Is the Real Test</title><link>https://koreainvestinsights.com/post/samsung-2q26-preview-micron-surprise-erased-core-op-hbm-2026-06-29/</link><pubDate>Mon, 29 Jun 2026 14:35:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/samsung-2q26-preview-micron-surprise-erased-core-op-hbm-2026-06-29/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Related context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/" &gt;Micron FY3Q26 earnings&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/nvidia-earnings-elasticity-hbm-cycle-samsung-hynix-2026-06-28/" &gt;the NVIDIA elasticity framework for Samsung and SK Hynix&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-hynix-micron-forward-per-parity-memory-catch-up-2026-06-03/" &gt;Samsung-Hynix-Micron parity&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-electronics-shareholder-return-ds-bonus-buyback-flow-2026-06-22/" &gt;Samsung shareholder return and DS bonus buyback flow&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-sk-hynix-hbm4e-12h-sample-allocation-battle-2026-06-18/" &gt;Samsung versus SK Hynix in HBM4E 12-high&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;The number that matters in Samsung Electronics’ 2Q26 preview is not simply reported operating profit. The market needs to separate reported OP from &lt;strong&gt;Core OP, defined as reported OP plus the catch-up accrual for performance-based compensation&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;At current prices, Micron’s FY3Q26 surprise has almost disappeared from Korean memory equities. Micron retained an 8.0% gain versus its pre-earnings price as of June 26. Samsung Electronics is down 5.9% versus the June 24 close based on June 29 intraday data, while SK Hynix is down 0.5%. The June 25 reaction was real, but most of it was erased by June 26 and June 29.&lt;/p&gt;
&lt;p&gt;Samsung’s 2Q26 DRAM pricing is not the downgrade factor. Conventional DRAM contract prices are estimated to be up 58-63% QoQ, while mobile LPDDR is up 70-83% QoQ. A blended Samsung DRAM ASP assumption in the high-50% QoQ range is therefore not aggressive.&lt;/p&gt;
&lt;p&gt;The reasonable range for 2Q26 reported OP is KRW 87-92 trillion, with a midpoint of KRW 89.5 trillion. Adding back KRW 15-17 trillion of catch-up bonus accrual implies Core OP of roughly KRW 104-106 trillion. In other words, the correct reading is not “earnings fell from the KRW 100 trillion expectation to KRW 89 trillion.” It is closer to “the memory core business reached more than KRW 100 trillion of profit power, but Samsung is recognizing excess-profit sharing costs in 2Q.”&lt;/p&gt;
&lt;p&gt;The rating is &lt;strong&gt;Watchlist / Conditional Buy&lt;/strong&gt;. Conditions are clear: reported OP should be at least KRW 87 trillion, Core OP should be at least KRW 102 trillion, the downgrade should be explained by accruals rather than DRAM/NAND price or shipment weakness, and 3Q26 OP durability around KRW 100 trillion should remain credible.&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Main Point&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Micron’s surprise was reflected in Samsung and SK Hynix for one day on June 25, but it has mostly been erased. Samsung is back to a price that barely embeds the Micron read-through. The real test is not headline 2Q OP, but Core OP above KRW 100 trillion and the durability of 3Q memory earnings.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-price-reaction-is-microns-surprise-still-in-korea-memory"&gt;1. Price reaction: is Micron’s surprise still in Korea memory?
&lt;/h2&gt;&lt;p&gt;Micron reported FY3Q26 results after the U.S. close on June 24, 2026. For Korea, the clean baseline is the June 24 close.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Jun 24 close&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Jun 25 close&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Jun 26 close&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Jun 29 intraday&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current vs Jun 24&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 340,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 358,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 339,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 320,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-5.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,580,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,917,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,673,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 2,568,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td style="text-align: right"&gt;USD 1,048.51&lt;/td&gt;
 &lt;td style="text-align: right"&gt;USD 1,213.56&lt;/td&gt;
 &lt;td style="text-align: right"&gt;USD 1,132.33&lt;/td&gt;
 &lt;td style="text-align: right"&gt;latest DB value Jun 26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+8.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Source: KII local daily price database for Korea and U.S. stocks; June 29 intraday Korea prices from Naver Finance API at 13:19 KST.&lt;/p&gt;
&lt;p&gt;SK Hynix reacted most directly on June 25, rising 13.1%. But it has since nearly returned to its pre-event level. Samsung reacted 5.3% on June 25, then fell below the pre-event price. Micron still held an 8.0% gain as of June 26.&lt;/p&gt;
&lt;p&gt;The conclusion is simple: at current prices, Korean memory has not retained Micron’s earnings read-through. It was reflected for one day, then erased by macro pressure, KOSPI overheat adjustment and company-specific uncertainty. The unreflected alpha is more visible in Samsung. SK Hynix is more of a pullback-and-reacceleration candidate.&lt;/p&gt;
&lt;h2 id="2-what-micron-confirmed"&gt;2. What Micron confirmed
&lt;/h2&gt;&lt;p&gt;Micron FY3Q26 was a strong benchmark for the memory cycle. Revenue was USD 41.456 billion, non-GAAP EPS was USD 25.11, non-GAAP gross margin was 84.9%, and FY4Q26 guidance called for USD 50.0 billion revenue and USD 31.00 non-GAAP EPS. These are official company numbers. &lt;a class="link" href="https://www.globenewswire.com/news-release/2026/06/24/3317151/14450/en/micron-technology-inc-reports-record-results-for-the-third-quarter-of-fiscal-2026.html" target="_blank" rel="noopener"
 &gt;Micron FY3Q26 earnings release&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The key read-throughs are:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Micron signal&lt;/th&gt;
 &lt;th&gt;Meaning for Samsung and SK Hynix&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;USD 41.456B revenue and 84.9% non-GAAP GM&lt;/td&gt;
 &lt;td&gt;Memory pricing, mix and shortage-driven operating leverage are very strong.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY4Q guidance of USD 50B revenue and USD 31 EPS&lt;/td&gt;
 &lt;td&gt;Pricing and shipments are unlikely to collapse immediately after 2Q.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Data center revenue above USD 25B&lt;/td&gt;
 &lt;td&gt;AI memory demand is broader than HBM alone.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4 and HBM4E roadmap&lt;/td&gt;
 &lt;td&gt;2027 allocation and customer qualification remain the next battleground.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Strategic Customer Agreements&lt;/td&gt;
 &lt;td&gt;Long-term contracts may reduce the old memory cyclicality discount.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This matters because Samsung now has to prove the same point with its own numbers: whether core memory profitability is already in the KRW 100 trillion range.&lt;/p&gt;
&lt;h2 id="3-reported-op-versus-core-op"&gt;3. Reported OP versus Core OP
&lt;/h2&gt;&lt;p&gt;The working formula is:&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Reported OP = accounting operating profit
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Core OP = Reported OP + catch-up accrual for performance-based compensation
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;Base case:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Base case&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reported OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 89.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Catch-up bonus accrual&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 15-17T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 104-106T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The important point is that the expected OP cut is not driven by weaker DRAM or NAND assumptions. In the research input, Kiwoom’s 2Q26 OP estimate fell from KRW 100T in May to KRW 89T on June 29. But over the same period, DRAM ASP assumptions moved from +55% QoQ to +58% QoQ, and NAND from +72% QoQ to +75% QoQ. The downgrade is therefore better explained by bonus accruals, non-memory weakness, DX cost pressure, and continued Foundry/S.LSI losses.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;May estimate&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Late-June estimate&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q26 OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 100T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 89T&lt;/td&gt;
 &lt;td&gt;Headline lower&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM ASP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55% QoQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+58% QoQ&lt;/td&gt;
 &lt;td&gt;Price assumption stronger&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND ASP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+72% QoQ&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+75% QoQ&lt;/td&gt;
 &lt;td&gt;Price assumption stronger&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Downgrade driver&lt;/td&gt;
 &lt;td style="text-align: right"&gt;not applicable&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Bonus, non-memory, DX, Foundry/S.LSI&lt;/td&gt;
 &lt;td&gt;Not a memory pricing problem&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="4-official-1q26-base"&gt;4. Official 1Q26 base
&lt;/h2&gt;&lt;p&gt;Samsung’s 1Q26 official results provide the base:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1Q26&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Company revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 133.9T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Company operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 57.2T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DS revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 81.7T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DS operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 53.7T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Source: &lt;a class="link" href="https://news.samsungsemiconductor.com/global/samsung-electronics-announces-first-quarter-2026-results/" target="_blank" rel="noopener"
 &gt;Samsung Electronics 1Q26 results&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;DS operating profit of KRW 53.7T is the key base for the 2Q bonus accrual discussion. If 1Q did not fully accrue the new profit-sharing structure, 2Q can carry a catch-up cost.&lt;/p&gt;
&lt;h2 id="5-dram-pricing-is-supportive"&gt;5. DRAM pricing is supportive
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Product group&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Estimated 2Q26 price trend&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Conventional DRAM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+58-63% QoQ&lt;/td&gt;
 &lt;td&gt;Kiwoom’s +58% assumption is near the lower end of the market range.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PC DDR5&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+43-48% QoQ&lt;/td&gt;
 &lt;td&gt;PC demand is not strong, but pricing still rose.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PC DDR4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+35-40% QoQ&lt;/td&gt;
 &lt;td&gt;Lower than DDR5, but still strong.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mobile LPDDR4X&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+70-75% QoQ&lt;/td&gt;
 &lt;td&gt;Mobile DRAM helps pull blended ASP higher.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mobile LPDDR5X&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+78-83% QoQ&lt;/td&gt;
 &lt;td&gt;Positive for Samsung’s blended DRAM ASP.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Consumer DRAM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+45-50% QoQ&lt;/td&gt;
 &lt;td&gt;Customer price resistance exists, but pricing is still up.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The core conclusion is that a high-50% QoQ blended DRAM ASP assumption is reasonable. If reported OP comes in near KRW 89T, it should not be interpreted as DRAM price weakness unless Samsung’s own disclosures say otherwise.&lt;/p&gt;
&lt;h2 id="6-bonus-accrual-treatment"&gt;6. Bonus accrual treatment
&lt;/h2&gt;&lt;p&gt;The existence of DS performance-based compensation is not new. The unsettled question is scale, timing and disclosure quality.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Analytical purpose&lt;/th&gt;
 &lt;th&gt;How to treat bonus accrual&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2Q earnings quality&lt;/td&gt;
 &lt;td&gt;Add it back to estimate Core OP&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3Q run-rate&lt;/td&gt;
 &lt;td&gt;Exclude 1Q catch-up, include normal ongoing bonus expense&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY26 EPS&lt;/td&gt;
 &lt;td&gt;Include it&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY27 valuation&lt;/td&gt;
 &lt;td&gt;Include it if recurring&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shareholder return and FCF&lt;/td&gt;
 &lt;td&gt;Include it&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Short-term event trade&lt;/td&gt;
 &lt;td&gt;Watch both reported OP and Core OP&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The two mistakes to avoid are: treating all bonus expense as a one-off, or treating all of it as structural deterioration. The catch-up portion matters for 2Q interpretation. The recurring profit-sharing structure matters for FY26/FY27 EPS and valuation.&lt;/p&gt;
&lt;h2 id="7-sell-side-op-distribution"&gt;7. Sell-side OP distribution
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Source&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2Q26 OP estimate&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;BNK&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 77.7T&lt;/td&gt;
 &lt;td&gt;Most conservative bonus retroactive accrual assumption&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Securities&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 81.3T&lt;/td&gt;
 &lt;td&gt;DRAM ASP +46%, NAND ASP +65%, DS bonus reflected&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mirae Asset&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 84.1T&lt;/td&gt;
 &lt;td&gt;Strong pre-accrual cost scenario near KRW 35T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FnGuide consensus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 86.2T&lt;/td&gt;
 &lt;td&gt;Early-June average, may not fully reflect latest prices and accruals&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Kiwoom Jun 29&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 89.0T&lt;/td&gt;
 &lt;td&gt;Higher DRAM/NAND price assumptions, but bonus/non-memory costs reflected&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KB&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 90.0-90.2T&lt;/td&gt;
 &lt;td&gt;Memory price up more than 50%, DS OP in the high-KRW 80T range&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daishin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 91.0T&lt;/td&gt;
 &lt;td&gt;Memory price strength with partial cost reflection&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hana&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 92.0T&lt;/td&gt;
 &lt;td&gt;Stronger pricing and shipment assumptions&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Investment framing:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Reported OP range&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Below KRW 84T&lt;/td&gt;
 &lt;td&gt;Bear case. Costs and non-memory drag are very large.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 87-92T&lt;/td&gt;
 &lt;td&gt;Base cluster. This is the realistic range.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Above KRW 94T&lt;/td&gt;
 &lt;td&gt;Bull case. Costs are limited or price/shipments are even stronger.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="8-scenario-model"&gt;8. Scenario model
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Bear&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Base&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Bull&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM blended ASP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+50%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+56-58%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+62-63%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM bit growth&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FX effect&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 86-88T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 91-93T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 95-97T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM OP margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;71-73%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;73-75%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;76-78%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 62-64T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 67-70T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 72-75T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 18-20T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 20-22T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 22-24T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry/S.LSI OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.5T to -KRW 1.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.0T to 0&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0 to KRW 0.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Display/DX/Harman combined&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 0.5T to KRW 1.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;0 to KRW 1.0T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 1-2T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reported OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 82-86T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 87-92T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 94-98T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Base midpoint:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Midpoint&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 181T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reported OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 89.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;OPM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about 49.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core OP&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 104-106T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="9-sensitivities"&gt;9. Sensitivities
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Variable&lt;/th&gt;
 &lt;th style="text-align: right"&gt;OP sensitivity&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM ASP +1ppt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.4-0.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DRAM bit growth +1ppt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.4-0.5T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAND ASP +1ppt&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 0.1-0.2T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bonus/accrual +KRW 1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.0T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DX cost +KRW 1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.0T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry/S.LSI loss +KRW 1T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.0T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The OP downgrade from KRW 100T to KRW 89T despite stronger DRAM/NAND price assumptions implies roughly KRW 12-13T of cost, accrual and non-memory drag.&lt;/p&gt;
&lt;h2 id="10-reported-op-and-core-op-matrix"&gt;10. Reported OP and Core OP matrix
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Reported OP&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Bonus accrual&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Core OP&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;th&gt;Action&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 92T or higher&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 15T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 107T or higher&lt;/td&gt;
 &lt;td&gt;Clear beat&lt;/td&gt;
 &lt;td&gt;Buy possible&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 89-92T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 15T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 104-107T&lt;/td&gt;
 &lt;td&gt;In-line positive&lt;/td&gt;
 &lt;td&gt;Buy weakness&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 87-89T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 15T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 102-104T&lt;/td&gt;
 &lt;td&gt;Consensus area&lt;/td&gt;
 &lt;td&gt;Staged entry possible&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 84-87T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 15T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 99-102T&lt;/td&gt;
 &lt;td&gt;Weak miss&lt;/td&gt;
 &lt;td&gt;Needs explanation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 84-87T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 25-35T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 109-122T&lt;/td&gt;
 &lt;td&gt;Headline miss, but neutral to positive if full-year pre-accrual&lt;/td&gt;
 &lt;td&gt;Verify cost nature&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;KRW 80-84T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;about KRW 15T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 95-99T&lt;/td&gt;
 &lt;td&gt;Cannot be explained by bonus alone&lt;/td&gt;
 &lt;td&gt;Wait or reduce&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;Below KRW 80T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;any&lt;/td&gt;
 &lt;td style="text-align: right"&gt;not meaningful&lt;/td&gt;
 &lt;td&gt;Thesis damage likely&lt;/td&gt;
 &lt;td&gt;Avoid&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="11-catalysts"&gt;11. Catalysts
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Event&lt;/th&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Importance&lt;/th&gt;
 &lt;th&gt;Watch item&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26 preliminary results&lt;/td&gt;
 &lt;td&gt;likely early July 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Very high&lt;/td&gt;
 &lt;td&gt;Reported OP at least KRW 87T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung 2Q26 final results and call&lt;/td&gt;
 &lt;td&gt;likely late July 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Very high&lt;/td&gt;
 &lt;td&gt;Bonus accrual, DS core profit, 3Q outlook&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3Q26 DRAM/NAND contract prices&lt;/td&gt;
 &lt;td&gt;July-August 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Very high&lt;/td&gt;
 &lt;td&gt;Flat to up QoQ&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4E, SOCAMM2, eSSD customer comments&lt;/td&gt;
 &lt;td&gt;2H26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;Evidence of share gain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bonus-share buyback disclosure&lt;/td&gt;
 &lt;td&gt;TBD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;Distinguish compensation buyback from cancellation buyback&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron FY4Q26 results&lt;/td&gt;
 &lt;td&gt;likely late September 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;High&lt;/td&gt;
 &lt;td&gt;Whether USD 50B revenue and USD 31 EPS are delivered&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="12-valuation"&gt;12. Valuation
&lt;/h2&gt;&lt;p&gt;The Kiwoom June 29 estimates summarized in the research input are:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2026F&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2027F&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 723.6T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 820.1T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 372.9T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 421.3T&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 43,429&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 48,177&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PER&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.8x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7.0x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PBR&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.5x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2.3x&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ROE&lt;/td&gt;
 &lt;td style="text-align: right"&gt;53.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;39.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;A KRW 430,000 target price implies 9.9x 2026F EPS and 8.9x 2027F EPS. This is valid only if 3Q26 OP remains above KRW 100T, 2027 EPS stays near KRW 48,000, bonus and stock compensation are already embedded in EPS, HBM4/eSSD contributions are visible, and DRAM/NAND prices do not fall sharply.&lt;/p&gt;
&lt;h2 id="13-red-team-and-invalidation"&gt;13. Red Team and invalidation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;th&gt;What would invalidate the thesis&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Macro failure&lt;/td&gt;
 &lt;td&gt;Rates and credit spreads slow AI capex, hyperscaler capex guidance falls, AI server orders are delayed.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micro failure&lt;/td&gt;
 &lt;td&gt;Samsung does not gain HBM4/eSSD share, HBM4 qualification is delayed, Foundry/S.LSI losses do not shrink.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Accounting failure&lt;/td&gt;
 &lt;td&gt;Reported OP falls below KRW 84T without enough bonus/accrual explanation, or Core OP falls below KRW 100T.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Capital allocation failure&lt;/td&gt;
 &lt;td&gt;Compensation buybacks are mistaken for shareholder return, while recurring profit-sharing erodes FY27 EPS and FCF.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Hard invalidation conditions:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reported OP below KRW 84T and not explained by at least KRW 25T of pre-accrual&lt;/td&gt;
 &lt;td&gt;More than timing noise&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core OP below KRW 100T&lt;/td&gt;
 &lt;td&gt;Core memory profit power weaker than expected&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3Q26 OP outlook below KRW 100T&lt;/td&gt;
 &lt;td&gt;Peak-out risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;3Q DRAM/NAND prices fall QoQ&lt;/td&gt;
 &lt;td&gt;Pricing thesis damaged&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;No HBM4/eSSD share gains&lt;/td&gt;
 &lt;td&gt;2027 EPS and multiple expansion limited&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foundry/S.LSI loss persists or widens&lt;/td&gt;
 &lt;td&gt;Samsung discount persists&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final view
&lt;/h2&gt;&lt;p&gt;At current prices, Micron’s surprise is almost gone from Samsung and SK Hynix. Samsung has returned to a price that barely embeds the Micron read-through. SK Hynix has almost returned to pre-event levels after a strong June 25 reaction.&lt;/p&gt;
&lt;p&gt;Samsung’s opportunity is now specific: if 2Q preliminary results show that the headline OP decline is mostly bonus-accrual timing, while Core OP is above KRW 100T and 3Q durability remains intact, the stock can reprice.&lt;/p&gt;
&lt;p&gt;The relative conclusion is straightforward:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Unreflected alpha if Core OP and 3Q durability are confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Pullback and reacceleration alpha, not a “cheap” stock&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td&gt;Cleanest proof of the AI memory cycle, but already repriced&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;One sentence: Micron’s earnings were reflected in Korean memory names for one day, but most of that effect has been erased. For Samsung, the next battle is not headline 2Q OP but Core OP above KRW 100 trillion and 3Q durability.&lt;/p&gt;</description></item><item><title>HBF And HBC Commercialization Calendar: Milestones, Stock Triggers, And Misclassified Baskets</title><link>https://koreainvestinsights.com/post/hbf-hbc-commercialization-stock-trigger-catalyst-calendar-2026-06-28/</link><pubDate>Sun, 28 Jun 2026 23:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/hbf-hbc-commercialization-stock-trigger-catalyst-calendar-2026-06-28/</guid><description>&lt;h1 id="hbf-and-hbc-commercialization-calendar-milestones-stock-triggers-and-misclassified-baskets"&gt;HBF And HBC Commercialization Calendar: Milestones, Stock Triggers, And Misclassified Baskets
&lt;/h1&gt;&lt;p&gt;This note follows the earlier HBM, HBF and HBC technology comparison. The goal here is not to restate the technology. The goal is to separate timing, evidence, equity exposure and misclassified stock baskets.&lt;/p&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;HBF and HBC are real directions, but both are early.&lt;/li&gt;
&lt;li&gt;HBF should be read as second-half 2026 sample, early 2027 first inference-device sample, and 2028 or later revenue potential.&lt;/li&gt;
&lt;li&gt;HBC starts with Qualcomm AI250. Qualcomm AI200 is the pre-HBC product. AI250 with HBC Gen 1 is targeted for commercial sampling in mid-2027. Revenue relevance is more likely 2028 or later.&lt;/li&gt;
&lt;li&gt;The closest 2026 trigger is not revenue. It is physical evidence: HBF sample delivery and Qualcomm AI200 shipment.&lt;/li&gt;
&lt;li&gt;HBF equity exposure is narrow. SanDisk is the most direct listed option, but the stock is still primarily a NAND and data-center SSD cycle stock. SK Hynix has HBF exposure, but its stock is still driven by HBM. FADU is not a confirmed HBF beneficiary.&lt;/li&gt;
&lt;li&gt;HBC exposure is also narrow. Qualcomm is the home asset. In Korea, Samsung Electro-Mechanics is the most direct confirmed component exposure through FC-BGA substrates for Qualcomm AI200, based on media reports. Hanmi Semiconductor is an HBM and HBF equipment name, not an HBC name.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="why-the-distinction-matters"&gt;Why The Distinction Matters
&lt;/h2&gt;&lt;p&gt;HBF and HBC sound similar, but they are not the same layer.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;What it is&lt;/th&gt;
 &lt;th&gt;Current state&lt;/th&gt;
 &lt;th&gt;Key question&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM&lt;/td&gt;
 &lt;td&gt;High-bandwidth DRAM memory&lt;/td&gt;
 &lt;td&gt;Mass production&lt;/td&gt;
 &lt;td&gt;Can 2027 pricing and allocation hold?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBF&lt;/td&gt;
 &lt;td&gt;NAND-based high-bandwidth flash layer&lt;/td&gt;
 &lt;td&gt;Standardization and pre-sample&lt;/td&gt;
 &lt;td&gt;Can samples and customer adoption appear?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBC&lt;/td&gt;
 &lt;td&gt;Qualcomm accelerator architecture&lt;/td&gt;
 &lt;td&gt;Roadmap and design targets&lt;/td&gt;
 &lt;td&gt;Can Qualcomm win data-center customers?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;SanDisk and SK Hynix have announced HBF standardization work under OCP. SanDisk has targeted first HBF samples in the second half of 2026 and first AI inference devices with HBF in early 2027. Qualcomm, meanwhile, describes HBC as a near-memory compute architecture inside its Dragonfly data-center roadmap, with HBC Gen 1 commercial sampling targeted with AI250 in mid-2027.&lt;/p&gt;
&lt;h2 id="commercialization-clock"&gt;Commercialization Clock
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Milestone&lt;/th&gt;
 &lt;th&gt;HBF&lt;/th&gt;
 &lt;th&gt;HBC&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Current state&lt;/td&gt;
 &lt;td&gt;Pre-sample. Most numbers are simulations or targets&lt;/td&gt;
 &lt;td&gt;No shipped HBC product. Public numbers are design targets&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First physical evidence&lt;/td&gt;
 &lt;td&gt;HBF sample delivery in 2H26&lt;/td&gt;
 &lt;td&gt;AI200 shipment in 2H26, but AI200 is not HBC&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Defining point&lt;/td&gt;
 &lt;td&gt;First HBF inference-device sample in early 2027&lt;/td&gt;
 &lt;td&gt;AI250 with HBC Gen 1 sample in mid-2027&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue timing&lt;/td&gt;
 &lt;td&gt;2028 or later, likely 2028-2030&lt;/td&gt;
 &lt;td&gt;2028 or later&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Largest swing factor&lt;/td&gt;
 &lt;td&gt;NVIDIA or major AI-platform adoption&lt;/td&gt;
 &lt;td&gt;Qualcomm data-center customer orders and revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Direct listed exposure&lt;/td&gt;
 &lt;td&gt;SanDisk, with NAND caveat&lt;/td&gt;
 &lt;td&gt;Qualcomm; Samsung Electro-Mechanics as Korea component exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The practical conclusion is simple: 2026 is a sample and shipment watch, 2027 is a validation watch, and 2028 or later is the revenue watch.&lt;/p&gt;
&lt;h2 id="hbf-stock-exposure"&gt;HBF Stock Exposure
&lt;/h2&gt;&lt;p&gt;SanDisk is the cleanest HBF equity option because it is directly involved in HBF standardization and sample timing. But it is not a pure HBF stock. The core drivers remain NAND pricing, data-center SSD demand and AI storage backlog. HBF is an asymmetric option layered on top of that.&lt;/p&gt;
&lt;p&gt;SK Hynix is involved in standardization, but HBF is not the main stock driver. HBM4 allocation, NVIDIA demand, DRAM pricing and memory operating margin remain far more important.&lt;/p&gt;
&lt;p&gt;FADU should not be put into the HBF basket without evidence of HBF products, contracts or logic-die supply. Its real discussion belongs under eSSD controllers, AI storage, ICMS and KV-cache offload. The company also carries legal uncertainty from the IPO-related case reported in late 2025 and 2026.&lt;/p&gt;
&lt;h2 id="hbc-stock-exposure"&gt;HBC Stock Exposure
&lt;/h2&gt;&lt;p&gt;Qualcomm is the home asset for HBC. But its data-center business is still small relative to its smartphone, automotive and broader QCT base. The main triggers are AI200 shipment, AI250 sampling, customer purchase orders, data-center revenue and independent benchmark validation.&lt;/p&gt;
&lt;p&gt;Samsung Electro-Mechanics is the most direct Korean component exposure. Media reports say it has begun producing FC-BGA substrates for Qualcomm AI200 at its Busan plant. This is exposure to Qualcomm&amp;rsquo;s data-center accelerator roadmap. It is not yet direct HBC Gen 1 exposure because AI200 is not HBC.&lt;/p&gt;
&lt;p&gt;Samsung Electronics, SK Hynix and Micron may have LPDDR relevance, but HBC-linked LPDDR is not as economically powerful as HBM. For Korean memory majors, HBM remains the main thesis.&lt;/p&gt;
&lt;p&gt;Hanmi Semiconductor is misclassified as an HBC beneficiary. Its TC bonder exposure belongs to HBM and potentially HBF, not Qualcomm HBC.&lt;/p&gt;
&lt;h2 id="priority-trigger-calendar"&gt;Priority Trigger Calendar
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Priority&lt;/th&gt;
 &lt;th&gt;Theme&lt;/th&gt;
 &lt;th&gt;Trigger&lt;/th&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;HBF&lt;/td&gt;
 &lt;td&gt;First HBF sample delivery&lt;/td&gt;
 &lt;td&gt;2H26&lt;/td&gt;
 &lt;td&gt;First physical proof&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;HBC&lt;/td&gt;
 &lt;td&gt;AI200 shipment&lt;/td&gt;
 &lt;td&gt;2H26&lt;/td&gt;
 &lt;td&gt;First proof of Qualcomm data-center accelerator execution&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;HBC&lt;/td&gt;
 &lt;td&gt;Data-center revenue visibility&lt;/td&gt;
 &lt;td&gt;FY27&lt;/td&gt;
 &lt;td&gt;Most objective proof&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;HBF&lt;/td&gt;
 &lt;td&gt;NVIDIA stance&lt;/td&gt;
 &lt;td&gt;Unknown&lt;/td&gt;
 &lt;td&gt;Biggest two-way wildcard&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;HBC&lt;/td&gt;
 &lt;td&gt;AI250/HBC Gen 1 sample&lt;/td&gt;
 &lt;td&gt;Mid-2027&lt;/td&gt;
 &lt;td&gt;The defining HBC milestone&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;HBF&lt;/td&gt;
 &lt;td&gt;Anchor customer qualification&lt;/td&gt;
 &lt;td&gt;2027&lt;/td&gt;
 &lt;td&gt;From technology claim to customer adoption&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;Both&lt;/td&gt;
 &lt;td&gt;Independent benchmarks&lt;/td&gt;
 &lt;td&gt;2027+&lt;/td&gt;
 &lt;td&gt;Validation of design targets&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;8&lt;/td&gt;
 &lt;td&gt;HBC&lt;/td&gt;
 &lt;td&gt;Customer PO and deployment&lt;/td&gt;
 &lt;td&gt;2027-2028&lt;/td&gt;
 &lt;td&gt;Narrative turns into revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-take"&gt;Final Take
&lt;/h2&gt;&lt;p&gt;HBF and HBC both address a real AI memory bottleneck. But the equity exposure is narrower than the theme suggests. HBF equals SanDisk option, with SK Hynix as a side branch. HBC equals Qualcomm, with Samsung Electro-Mechanics as the most direct Korean component exposure. FADU in HBF and Hanmi Semiconductor in HBC are weak classifications.&lt;/p&gt;
&lt;p&gt;The right framework is to buy confirmation, not vocabulary. In 2026, watch samples and shipments. In 2027, watch qualification and AI250. In 2028 and beyond, watch revenue.&lt;/p&gt;
&lt;p&gt;Sources include Qualcomm&amp;rsquo;s June 2026 data-center roadmap, SanDisk and SK Hynix HBF announcements, Samsung Electro-Mechanics product materials, and media reports on Samsung Electro-Mechanics and FADU. This is catalyst analysis, not investment advice.&lt;/p&gt;</description></item><item><title>What NVIDIA's Earnings Elasticity Says About Samsung And SK Hynix</title><link>https://koreainvestinsights.com/post/nvidia-earnings-elasticity-hbm-cycle-samsung-hynix-2026-06-28/</link><pubDate>Sun, 28 Jun 2026 22:00:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/nvidia-earnings-elasticity-hbm-cycle-samsung-hynix-2026-06-28/</guid><description>&lt;h1 id="what-nvidias-earnings-elasticity-says-about-samsung-and-sk-hynix"&gt;What NVIDIA&amp;rsquo;s Earnings Elasticity Says About Samsung And SK Hynix
&lt;/h1&gt;&lt;p&gt;This is a follow-up to the Micron FY3Q26 review, the Samsung-Hynix-Micron parity series, the HBM4E allocation note, and the Samsung catch-up work. The question is simple: if Samsung Electronics and SK Hynix still have strong earnings, can their stocks keep rising with the same elasticity?&lt;/p&gt;
&lt;p&gt;The NVIDIA case says no. Stocks do not react only to the level of earnings. They react to the rate of change, the second derivative of growth, and whether estimates are still being revised up faster than the stock has already moved.&lt;/p&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;NVIDIA&amp;rsquo;s first share-price elasticity slowdown started around calendar 3Q24. The more structural plateau began around calendar 4Q25 to 1Q26.&lt;/li&gt;
&lt;li&gt;Revenue kept exploding, but the stock started responding less to absolute growth and more to the durability of surprise.&lt;/li&gt;
&lt;li&gt;The Samsung and SK Hynix earnings cycle has not broken yet. Recent SK Hynix, Samsung, and Micron results still point to strong AI-memory demand.&lt;/li&gt;
&lt;li&gt;The stock-cycle risk is different. Samsung and SK Hynix entered a first overheated zone around June 2026.&lt;/li&gt;
&lt;li&gt;The main decision window is 4Q26 to 1Q27. Watch 2027 HBM contract pricing and volumes, HBM4 share, NVIDIA and hyperscaler AI capex, and memory operating-margin peak-out.&lt;/li&gt;
&lt;li&gt;My equity call is conditional buy for Samsung Electronics and wait for SK Hynix. SK Hynix is the higher-quality HBM leader, but it is now a crowded winner. Samsung still has HBM4 and Vera Rubin catch-up optionality.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="the-nvidia-lesson"&gt;The NVIDIA Lesson
&lt;/h2&gt;&lt;p&gt;NVIDIA was the cleanest listed-equity case of the AI semiconductor cycle. The stock moved first, then the earnings caught up. But once the market fully believed the earnings path, excellent results no longer produced the same stock response.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;NVIDIA reporting quarter&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Revenue growth&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Data-center growth&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Calendar-quarter stock move&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FY23 Q4, reported Feb 2023&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY -21%, QoQ +2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;n/a&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2023 Q1 +90.1%&lt;/td&gt;
 &lt;td&gt;AI expectations moved before earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY24 Q1, reported May 2023&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY -13%, QoQ +19%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;DC revenue $4.28bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2023 Q2 +52.3%&lt;/td&gt;
 &lt;td&gt;Guidance shock&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY24 Q2, reported Aug 2023&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +101%, QoQ +88%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +171%, QoQ +141%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2023 Q3 +2.8%&lt;/td&gt;
 &lt;td&gt;First digestion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY24 Q4, reported Feb 2024&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +265%, QoQ +22%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +409%, QoQ +27%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2024 Q1 +82.5%&lt;/td&gt;
 &lt;td&gt;Blackwell confidence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY25 Q2, reported Aug 2024&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +122%, QoQ +15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +154%, QoQ +16%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2024 Q3 -1.7%&lt;/td&gt;
 &lt;td&gt;First inflection&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY26 Q3, reported Nov 2025&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +62%, QoQ +22%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +66%, QoQ +25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2025 Q4 about 0.0%&lt;/td&gt;
 &lt;td&gt;Structural plateau begins&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY26 Q4, reported Feb 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +73%, QoQ +20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +75%, QoQ +22%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026 Q1 -6.5%&lt;/td&gt;
 &lt;td&gt;Strong earnings, weak elasticity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY27 Q1, reported May 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +85%, QoQ +20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;YoY +92%, QoQ +21%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026 Q2 to Jun 26 +10.4%&lt;/td&gt;
 &lt;td&gt;Earnings follow-through, limited multiple expansion&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;NVIDIA FY27 Q1 revenue was $81.6bn and data-center revenue was $75.2bn. The FY27 Q2 guide was $91.0bn, plus or minus 2%. Those are still exceptional numbers. Yet from September 2025 to June 2026, the stock barely moved compared with its earlier rise.&lt;/p&gt;
&lt;p&gt;The lesson for Korean memory equities is clear: strong earnings are not enough. The stock has to be cheaper than the next estimate revision path.&lt;/p&gt;
&lt;h2 id="why-samsung-and-sk-hynix-have-not-broken-yet"&gt;Why Samsung And SK Hynix Have Not Broken Yet
&lt;/h2&gt;&lt;p&gt;The earnings cycle is still intact.&lt;/p&gt;
&lt;p&gt;SK Hynix reported 1Q26 revenue of KRW 52.5763tn, operating profit of KRW 37.6103tn, and net income of KRW 40.3459tn. Samsung reported 1Q26 consolidated revenue of KRW 133.9tn and operating profit of KRW 57.2tn, with DS operating profit at KRW 53.7tn. Micron&amp;rsquo;s FY3Q26 was also very strong, with revenue of $41.456bn, non-GAAP EPS of $25.11, and non-GAAP gross margin of 84.9%.&lt;/p&gt;
&lt;p&gt;That does not look like a demand collapse. It looks like a profitable shortage.&lt;/p&gt;
&lt;p&gt;The problem is not the earnings cycle. The problem is the stock cycle. SK Hynix became South Korea&amp;rsquo;s most valuable company in June 2026, and Reuters reported that the stock was up more than 340% year-to-date. At that point, the market no longer needs proof that SK Hynix is an HBM winner. It needs proof that 2027 EPS estimates still have room to move up.&lt;/p&gt;
&lt;h2 id="the-timing-framework"&gt;The Timing Framework
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stage&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Expected timing&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;th&gt;Key signal&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;First stock-elasticity slowdown&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Already around June 2026&lt;/td&gt;
 &lt;td&gt;Overheated zone&lt;/td&gt;
 &lt;td&gt;SK Hynix market-cap leadership, retail FOMO, target-price chase&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EPS-revision peak&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4Q26 to 1Q27&lt;/td&gt;
 &lt;td&gt;Most important window&lt;/td&gt;
 &lt;td&gt;2027 HBM contract price and volume disappoint or confirm&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Earnings peak-out&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1H27 to 2H27&lt;/td&gt;
 &lt;td&gt;Base-case fundamental risk window&lt;/td&gt;
 &lt;td&gt;HBM ASP growth slows and memory margin rolls over&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Full memory downcycle&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2H27 to 2028&lt;/td&gt;
 &lt;td&gt;Not yet confirmed&lt;/td&gt;
 &lt;td&gt;Supply catches demand and inventories build&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The stock can peak before earnings. That is the central point.&lt;/p&gt;
&lt;h2 id="five-break-conditions"&gt;Five Break Conditions
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Break condition&lt;/th&gt;
 &lt;th&gt;What it means&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM ASP flattens or declines&lt;/td&gt;
 &lt;td&gt;The first sign is likely price momentum slowing, not demand disappearing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix margin falls for two straight quarters&lt;/td&gt;
 &lt;td&gt;Peak earnings risk becomes observable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung and Micron reduce SK Hynix&amp;rsquo;s HBM premium&lt;/td&gt;
 &lt;td&gt;The monopoly premium becomes a leadership premium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA data-center QoQ growth falls into single digits for two quarters&lt;/td&gt;
 &lt;td&gt;AI capex is no longer accelerating fast enough&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2027 supply catches demand&lt;/td&gt;
 &lt;td&gt;HBM4 and HBM4E ramp creates a new pricing regime&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="stock-implications"&gt;Stock Implications
&lt;/h2&gt;&lt;p&gt;Samsung Electronics is the cleaner conditional buy. The point is not that Samsung&amp;rsquo;s memory business is good. The point is that its HBM laggard discount can narrow if HBM4 and SOCAMM2 revenue show up in DS margin.&lt;/p&gt;
&lt;p&gt;SK Hynix remains the quality leader, but new buying is less attractive. The stock now needs more than a strong 2026. It needs 2027 HBM contracts to prove that EPS can still be revised up.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Conditional buy&lt;/td&gt;
 &lt;td&gt;HBM4 and Vera Rubin catch-up can reduce the laggard discount&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;Best HBM asset, but crowded and closer to peak-expectation risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NVIDIA&lt;/td&gt;
 &lt;td&gt;Watchlist&lt;/td&gt;
 &lt;td&gt;AI leader, but earnings growth and share-price elasticity have separated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM equipment and substrates&lt;/td&gt;
 &lt;td&gt;Selective watch&lt;/td&gt;
 &lt;td&gt;Alpha can move down the bottleneck stack if megacap memory stalls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;Samsung and SK Hynix are not in an earnings breakdown. They are in a stock-elasticity test. The right question is no longer whether HBM is good. The right question is whether 2027 EPS revisions can outrun the stock price.&lt;/p&gt;
&lt;p&gt;That makes Samsung Electronics the more asymmetric Korean memory expression today. SK Hynix is higher quality, but more fully believed.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2027" target="_blank" rel="noopener"
 &gt;NVIDIA FY27 Q1&lt;/a&gt;, &lt;a class="link" href="https://nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-fourth-quarter-and-fiscal-2026" target="_blank" rel="noopener"
 &gt;NVIDIA FY26 Q4&lt;/a&gt;, &lt;a class="link" href="https://news.samsungsemiconductor.com/global/samsung-electronics-announces-first-quarter-2026-results/" target="_blank" rel="noopener"
 &gt;Samsung 1Q26&lt;/a&gt;, &lt;a class="link" href="https://news.skhynix.com/q1-2026-business-results/" target="_blank" rel="noopener"
 &gt;SK Hynix 1Q26&lt;/a&gt;, &lt;a class="link" href="https://investors.micron.com/news-releases/news-release-details/micron-technology-inc-reports-record-results-third-quarter" target="_blank" rel="noopener"
 &gt;Micron FY3Q26&lt;/a&gt;, &lt;a class="link" href="https://www.reuters.com/world/asia-pacific/sk-hynix-overtakes-samsung-become-koreas-most-valuable-company-2026-06-22/" target="_blank" rel="noopener"
 &gt;Reuters on SK Hynix&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;small&gt;Disclaimer: This is research and information only, not investment advice.&lt;/small&gt;&lt;/p&gt;</description></item><item><title>Micron FY3Q26 Earnings: AI Memory Bottlenecks, 84.9% Gross Margin, And A $50B Guide</title><link>https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/</link><pubDate>Thu, 25 Jun 2026 15:20:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/micron-fy3q26-ai-memory-sca-fcf-hold-2026-06-25/</guid><description>&lt;h1 id="micron-fy3q26-earnings-ai-memory-bottlenecks-849-gross-margin-and-a-50b-guide"&gt;Micron FY3Q26 Earnings: AI Memory Bottlenecks, 84.9% Gross Margin, And A $50B Guide
&lt;/h1&gt;&lt;p&gt;This note follows the Samsung-Hynix-Micron parity series, the AI chip and memory P/E map, the SK Hynix versus Micron comparison, and the recent memory-price-risk work around CXMT. The central question is no longer whether Micron is an AI-memory proxy. FY3Q26 showed that the proxy now has earnings, gross margin, FCF, and customer commitments behind it.&lt;/p&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;Micron&amp;rsquo;s FY3Q26 was not just an earnings beat. Revenue reached $41.456bn, non-GAAP EPS was $25.11, non-GAAP gross margin was 84.9%, and FY4Q26 guidance called for $50.0bn of revenue and $31.00 of non-GAAP EPS. These figures are from Micron&amp;rsquo;s official release.&lt;/p&gt;
&lt;p&gt;The key change is threefold.&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Micron can no longer be treated as a plain commodity-memory equity. Management framed memory bandwidth and capacity as structural bottlenecks for AI systems, and data-center revenue exceeded $25bn in the quarter.&lt;/li&gt;
&lt;li&gt;The company has signed 16 strategic customer agreements, or SCAs. For 14 of those contracts, Micron pointed to roughly $100bn of minimum revenue at floor prices and about $22bn of customer deposits and financing commitments. Take-or-pay terms and price floors reduce, but do not eliminate, the earnings-collapse risk that normally defines memory cycles.&lt;/li&gt;
&lt;li&gt;The stock already reacted. After the release, after-hours quotes showed a double-digit rally, roughly in the mid-teens depending on the data vendor. The numbers were excellent, but the margin of safety narrowed quickly.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;My conclusion is &lt;strong&gt;HOLD, with new buying better left for pullbacks&lt;/strong&gt;. The revised 12-month target is $1,350. Versus the roughly $1,214 after-hours reference price used in this note, that implies about 11% upside. The upside looks larger versus the regular close of $1,048.51, but that price did not fully include the earnings release.&lt;/p&gt;
&lt;h2 id="event-check"&gt;Event Check
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Status&lt;/th&gt;
 &lt;th style="text-align: right"&gt;KST Timing&lt;/th&gt;
 &lt;th&gt;Evidence&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Recent earnings release&lt;/td&gt;
 &lt;td&gt;Confirmed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026-06-25 05:01&lt;/td&gt;
 &lt;td&gt;Micron released FY3Q26 results at 2026-06-24 16:01 ET. FY3Q26 ended 2026-05-28.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Earnings call&lt;/td&gt;
 &lt;td&gt;Confirmed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026-06-25 05:30&lt;/td&gt;
 &lt;td&gt;Micron scheduled its Third Quarter 2026 Financial Call at 4:30 PM ET.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Prepared remarks and deck&lt;/td&gt;
 &lt;td&gt;Confirmed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;After release&lt;/td&gt;
 &lt;td&gt;Micron posted the FY3Q26 presentation and prepared remarks on its IR site.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Post-earnings analyst call&lt;/td&gt;
 &lt;td&gt;Confirmed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2026-06-25 07:00&lt;/td&gt;
 &lt;td&gt;Reflected in the event schedule used for this review.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Next earnings date&lt;/td&gt;
 &lt;td&gt;Not officially confirmed&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Not fixed&lt;/td&gt;
 &lt;td&gt;Vendor estimates differ around late September 2026.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The FY3Q26 event is confirmed. FY4Q26 is not. Until Micron publishes the next date, late September after the US close is only an estimate.&lt;/p&gt;
&lt;h2 id="fy3q26-numbers"&gt;FY3Q26 Numbers
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY3Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2Q26&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY3Q25&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$41.456bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$23.860bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$9.301bn&lt;/td&gt;
 &lt;td&gt;+73.7% QoQ, +345.7% YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP net income / EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$28.243bn / $24.67&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$13.785bn / $12.07&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.885bn / $1.68&lt;/td&gt;
 &lt;td&gt;Extreme operating leverage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP net income / EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$28.857bn / $25.11&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$14.021bn / $12.20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.181bn / $1.91&lt;/td&gt;
 &lt;td&gt;Non-GAAP EPS +105.8% QoQ and +1,214.7% YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;84.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;74.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;37.7%&lt;/td&gt;
 &lt;td&gt;Pricing, mix, and shortage effect&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;84.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;74.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;39.0%&lt;/td&gt;
 &lt;td&gt;+10.0pp QoQ and +45.9pp YoY&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP operating income / margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$33.681bn / 81.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$16.455bn / 69.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.490bn / 26.8%&lt;/td&gt;
 &lt;td&gt;More than 13 times the year-ago level&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating cash flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$25.39bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$11.90bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$4.61bn&lt;/td&gt;
 &lt;td&gt;Cash generation followed earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Capex, net&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$7.084bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5.004bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.660bn&lt;/td&gt;
 &lt;td&gt;Investment rising with demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Adjusted free cash flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$18.304bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$6.899bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.949bn&lt;/td&gt;
 &lt;td&gt;Record free cash flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cash, marketable investments, restricted cash&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$30.2bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;not provided&lt;/td&gt;
 &lt;td style="text-align: right"&gt;not provided&lt;/td&gt;
 &lt;td&gt;Buffer for capex and capital return&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The quality of the beat matters. GAAP EPS and non-GAAP EPS moved together. Operating income, operating cash flow, and adjusted FCF also moved together. This was not a one-line adjustment story.&lt;/p&gt;
&lt;h2 id="segment-data"&gt;Segment Data
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Segment&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY3Q26 Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2Q26 Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY3Q25 Revenue&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY3Q26 Gross Margin&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY3Q26 Operating Margin&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Cloud Memory&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$13.769bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$7.749bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$3.386bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;83%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;78%&lt;/td&gt;
 &lt;td&gt;Core HBM and high-end DRAM growth engine&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Core Data Center&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$11.524bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$5.687bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.530bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;87%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;83%&lt;/td&gt;
 &lt;td&gt;Data-center DRAM and SSD demand was very strong&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Mobile and Client&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$11.521bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$7.711bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$3.255bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;87%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;86%&lt;/td&gt;
 &lt;td&gt;PC and mobile also benefited from pricing and mix&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Automotive and Embedded&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$4.634bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$2.708bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1.127bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;79%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;75%&lt;/td&gt;
 &lt;td&gt;Auto and embedded also participated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Cloud Memory plus Core Data Center reached about $25.3bn. That matches management&amp;rsquo;s comment that data-center revenue exceeded $25bn, or more than a $100bn annualized run-rate. This point is crucial: the quarter was not only about HBM. DRAM, NAND, SSD, client, mobile, auto, and embedded all showed pricing power.&lt;/p&gt;
&lt;h2 id="fy4q26-guidance"&gt;FY4Q26 Guidance
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Company Guidance&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$50.0bn ± $1.0bn&lt;/td&gt;
 &lt;td&gt;$49bn to $51bn range&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 86%&lt;/td&gt;
 &lt;td&gt;Record-level margin sustained&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 86%&lt;/td&gt;
 &lt;td&gt;Up again from 84.9% in FY3Q&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GAAP EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$30.73 ± $1.00&lt;/td&gt;
 &lt;td&gt;Close to non-GAAP EPS&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$31.00 ± $1.00&lt;/td&gt;
 &lt;td&gt;$30 to $32 range&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP operating expenses&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $1.65bn&lt;/td&gt;
 &lt;td&gt;Strong operating leverage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Diluted shares&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 1.15bn&lt;/td&gt;
 &lt;td&gt;EPS denominator&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Reuters reported that the $50bn revenue guide was well above the LSEG consensus near $43.58bn. The guide says that FY3Q was not necessarily the peak. Still, management also noted that the rate of price increases can moderate, even if pricing remains favorable.&lt;/p&gt;
&lt;h2 id="consensus-beat"&gt;Consensus Beat
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Actual Or Guide&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Consensus / Estimate&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Difference&lt;/th&gt;
 &lt;th&gt;Reading&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FY3Q revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$41.46bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$35.69bn, Investing.com&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+16.2%&lt;/td&gt;
 &lt;td&gt;Beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY3Q adjusted EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$25.11&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$20.49, Investing.com&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+22.6%&lt;/td&gt;
 &lt;td&gt;Beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY3Q revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$41.46bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$35.91bn, MarketBeat&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+$5.54bn&lt;/td&gt;
 &lt;td&gt;Beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY3Q EPS&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$25.11&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$20.98, MarketBeat&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+$4.13&lt;/td&gt;
 &lt;td&gt;Beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-GAAP gross margin&lt;/td&gt;
 &lt;td style="text-align: right"&gt;84.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;81.6%, LSEG preview&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.3pp&lt;/td&gt;
 &lt;td&gt;Beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY4Q revenue guide&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$50.0bn ± $1.0bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$43.58bn, LSEG via Reuters&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Large beat&lt;/td&gt;
 &lt;td&gt;Beat and raise&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY4Q EPS guide&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$31.00 ± $1.00&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$25.84, Reuters preview&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Above&lt;/td&gt;
 &lt;td&gt;Estimates need to move up&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Consensus sources differ, but the conclusion is stable: Micron beat on revenue, EPS, gross margin, and the next-quarter guide.&lt;/p&gt;
&lt;h2 id="the-strategic-customer-agreement-point"&gt;The Strategic Customer Agreement Point
&lt;/h2&gt;&lt;p&gt;The most important non-earnings item was the SCA structure.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;SCA Item&lt;/th&gt;
 &lt;th&gt;Management Detail&lt;/th&gt;
 &lt;th&gt;Investment Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Number of agreements&lt;/td&gt;
 &lt;td&gt;16&lt;/td&gt;
 &lt;td&gt;Broad, not one-customer dependent&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Typical duration&lt;/td&gt;
 &lt;td&gt;Mostly 2026 to 2030, with shorter automotive structures&lt;/td&gt;
 &lt;td&gt;Longer visibility than a normal memory cycle&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Volume coverage&lt;/td&gt;
 &lt;td&gt;About 20% of DRAM volume and one-third of NAND volume&lt;/td&gt;
 &lt;td&gt;Already material&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Potential revenue coverage&lt;/td&gt;
 &lt;td&gt;Could approach more than half of revenue once fully signed&lt;/td&gt;
 &lt;td&gt;Commodity discount can shrink&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pricing&lt;/td&gt;
 &lt;td&gt;Take-or-pay, floors, ceilings, some fixed or market-based pricing&lt;/td&gt;
 &lt;td&gt;Downside protection and some upside limits&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RPO&lt;/td&gt;
 &lt;td&gt;More than $5bn at FY3Q end, with about $100bn for 14 signed SCAs expected in filings&lt;/td&gt;
 &lt;td&gt;Minimum revenue and floor-price framing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Deposits and commitments&lt;/td&gt;
 &lt;td&gt;Around $22bn, including about $18bn of cash deposits&lt;/td&gt;
 &lt;td&gt;Strong customer commitment&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The SCA changes the debate. It gives Micron a floor under part of the cycle. It also proves that customers care enough about supply assurance to put money and obligations behind the contracts. But it is not a free option. Price ceilings can cap part of the upside when spot markets rise sharply, and customer deposits sit in financing cash flow rather than free cash flow.&lt;/p&gt;
&lt;h2 id="what-management-and-analysts-focused-on"&gt;What Management And Analysts Focused On
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Topic&lt;/th&gt;
 &lt;th&gt;Core Message&lt;/th&gt;
 &lt;th&gt;Why It Matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Supply tightness&lt;/td&gt;
 &lt;td&gt;DRAM and NAND demand is above industry supply, with tightness expected beyond 2027&lt;/td&gt;
 &lt;td&gt;The bull case is supply discipline and capacity bottlenecks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4&lt;/td&gt;
 &lt;td&gt;High-volume shipments for a lead customer, more qualification samples, and more than $1bn of HBM4 revenue&lt;/td&gt;
 &lt;td&gt;HBM4 is already commercial, not just a future narrative&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM4E&lt;/td&gt;
 &lt;td&gt;Development for calendar 2027 volume production&lt;/td&gt;
 &lt;td&gt;The next allocation battle is already starting&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Deposits&lt;/td&gt;
 &lt;td&gt;Around $22bn of customer commitments, about $18bn in cash deposits&lt;/td&gt;
 &lt;td&gt;Customer behavior confirms scarcity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Capex&lt;/td&gt;
 &lt;td&gt;FY4Q capex around $10bn, FY2026 capex around $27bn, FY2027 quarterly capex above FY4Q&lt;/td&gt;
 &lt;td&gt;Growth needs capital and can pressure FCF later&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Gross margin&lt;/td&gt;
 &lt;td&gt;86% guide, but price increases may moderate&lt;/td&gt;
 &lt;td&gt;The next issue is sustainability, not the absolute beat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Buybacks&lt;/td&gt;
 &lt;td&gt;Capital return remains centered on repurchases, but no precise cadence was committed&lt;/td&gt;
 &lt;td&gt;FCF conversion and buyback scale are the next equity catalysts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The call was not mainly about whether FY3Q beat. It was about how durable the new profit level is.&lt;/p&gt;
&lt;h2 id="upside-and-downside-triggers"&gt;Upside And Downside Triggers
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Upside Trigger&lt;/th&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;th&gt;Mechanism&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FY4Q revenue above $51bn or EPS above $32&lt;/td&gt;
 &lt;td&gt;FY4Q26 earnings&lt;/td&gt;
 &lt;td&gt;FY2027 run-rate estimates move up&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;More FY2027 HBM3E/HBM4 allocation fixed, faster HBM4E qualification&lt;/td&gt;
 &lt;td&gt;FY4Q call and second-half product updates&lt;/td&gt;
 &lt;td&gt;Longer shortage duration supports higher margins&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SCA revenue coverage moves toward half or more of revenue&lt;/td&gt;
 &lt;td&gt;Next one to two quarters&lt;/td&gt;
 &lt;td&gt;Cyclicality discount shrinks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Large FY4Q FCF increase plus clearer buyback scale&lt;/td&gt;
 &lt;td&gt;FY4Q and 10-K&lt;/td&gt;
 &lt;td&gt;EPS accretion and FCF yield re-rating&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Non-HBM DRAM and NAND pricing remains strong&lt;/td&gt;
 &lt;td&gt;ASP and segment margin updates&lt;/td&gt;
 &lt;td&gt;Reduces HBM concentration risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Data-center revenue run-rate stays above $100bn&lt;/td&gt;
 &lt;td&gt;FY4Q and FY2027&lt;/td&gt;
 &lt;td&gt;Micron is reclassified as an AI infrastructure beneficiary&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Downside Trigger&lt;/th&gt;
 &lt;th&gt;Timing&lt;/th&gt;
 &lt;th&gt;Mechanism&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FY4Q revenue or EPS misses the midpoint of guidance&lt;/td&gt;
 &lt;td&gt;FY4Q26 earnings&lt;/td&gt;
 &lt;td&gt;Beat-and-raise thesis weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Gross margin rolls over from 86%&lt;/td&gt;
 &lt;td&gt;FY4Q and FY2027 guide&lt;/td&gt;
 &lt;td&gt;Memory supercycle multiple compresses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2027 capex consumes too much FCF&lt;/td&gt;
 &lt;td&gt;FY2026 10-K and FY2027 capex guide&lt;/td&gt;
 &lt;td&gt;FCF multiple and buyback hopes fall&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SCA ceilings cap upside&lt;/td&gt;
 &lt;td&gt;Contract and margin commentary&lt;/td&gt;
 &lt;td&gt;Agreements look like caps, not floors&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI customers shift to cheaper memory structures&lt;/td&gt;
 &lt;td&gt;Accelerator product cycles&lt;/td&gt;
 &lt;td&gt;Premium DRAM and HBM pricing power weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Chinese memory makers penetrate higher-end products&lt;/td&gt;
 &lt;td&gt;Late 2026 to 2027&lt;/td&gt;
 &lt;td&gt;Supply and price competition return&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="stock-reaction"&gt;Stock Reaction
&lt;/h2&gt;&lt;p&gt;FY3Q26 was released after the US close on 24 June 2026. Therefore the first regular session that fully reflects the news is 25 June 2026. At the original analysis time, that session had not opened.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Window&lt;/th&gt;
 &lt;th style="text-align: right"&gt;MU Price / Return&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026-06-24 regular session&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1,048.51, -0.31% using the supplied reference&lt;/td&gt;
 &lt;td&gt;Pre-release trading, not the event reaction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MarketBeat regular close&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1,037.93, -1.32%&lt;/td&gt;
 &lt;td&gt;Vendor difference&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;After-hours supplied reference&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1,213.97, +15.78%&lt;/td&gt;
 &lt;td&gt;Strong immediate reaction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MarketBeat after-hours&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1,221.62, +17.70%&lt;/td&gt;
 &lt;td&gt;Another vendor reference&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;1D regular session&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Pending&lt;/td&gt;
 &lt;td&gt;Available after 25 June close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2D regular session&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Pending&lt;/td&gt;
 &lt;td&gt;Available after 26 June close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;5D regular session&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Pending&lt;/td&gt;
 &lt;td&gt;Available next week&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The reaction was driven by more than EPS. The FY4Q revenue guide, EPS guide, SCA commitments, long-duration supply tightness, and the FCF profile all mattered.&lt;/p&gt;
&lt;h2 id="valuation-and-recommendation"&gt;Valuation And Recommendation
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Rating&lt;/td&gt;
 &lt;td&gt;HOLD&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;New buying&lt;/td&gt;
 &lt;td&gt;Wait for pullback&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;12-month target&lt;/td&gt;
 &lt;td&gt;$1,350&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reference price&lt;/td&gt;
 &lt;td&gt;About $1,213.97 after hours&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Upside&lt;/td&gt;
 &lt;td&gt;About +11.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Upside versus regular close&lt;/td&gt;
 &lt;td&gt;About +28.8%, but that price did not fully reflect the release&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The valuation frame is forward P/E. FY2026 non-GAAP EPS can be approximated as Q1 $4.78 plus Q2 $12.20 plus Q3 $25.11 plus the FY4Q guide midpoint of $31.00, or $73.09. For the base case, I use FY2027 EPS of about $100. That is below the FY4Q26 run-rate to reflect FY2027 capex, start-up costs, and price normalization risk. The base multiple is 13.5 times.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Scenario&lt;/th&gt;
 &lt;th style="text-align: right"&gt;EPS Assumption&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Multiple&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Target Range&lt;/th&gt;
 &lt;th&gt;Key Condition&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Bear&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$70-80&lt;/td&gt;
 &lt;td style="text-align: right"&gt;11-12x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$800-950&lt;/td&gt;
 &lt;td&gt;Supply improves, ASP falls, capex and FCF pressure return&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Base&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About $100&lt;/td&gt;
 &lt;td style="text-align: right"&gt;13.5x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1,350&lt;/td&gt;
 &lt;td&gt;FY4Q guide is achieved, HBM/SCA remain strong, FCF holds&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bull&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$110-120&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;$1,650-1,800&lt;/td&gt;
 &lt;td&gt;2027 supply sold-out extends, buybacks accelerate, SCA coverage rises&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="what-would-change-the-view"&gt;What Would Change The View
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Change&lt;/th&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Upgrade to BUY&lt;/td&gt;
 &lt;td&gt;Stock pulls back below $1,100, or FY4Q revenue exceeds $51bn and EPS exceeds $32 while FY2027 FCF visibility remains intact&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Downgrade to SELL / reduce&lt;/td&gt;
 &lt;td&gt;Gross margin falls quickly below 80%, HBM4/HBM4E qualification is delayed, SCA additions disappoint, or customers shift toward cheaper memory architectures&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Raise target&lt;/td&gt;
 &lt;td&gt;SCA-linked revenue approaches more than half of total revenue and FY2027 EPS visibility moves above $110&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Lower target&lt;/td&gt;
 &lt;td&gt;FY2027 capex materially constrains cash return and the expected FY4Q FCF step-up does not materialize&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="korean-semiconductor-read-through"&gt;Korean Semiconductor Read-Through
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Korean Investor Question&lt;/th&gt;
 &lt;th&gt;Micron FY3Q26 Answer&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Did Samsung Electronics and SK Hynix EPS risk rise?&lt;/td&gt;
 &lt;td&gt;No. Micron&amp;rsquo;s DRAM, NAND, HBM, and SSD numbers point to strong memory demand and pricing.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is the Micron premium more justified now?&lt;/td&gt;
 &lt;td&gt;Partly yes. 84.9% non-GAAP gross margin, the 86% FY4Q guide, and SCA structures reduce the commodity-memory discount.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Does this hurt SK Hynix&amp;rsquo;s exclusivity premium?&lt;/td&gt;
 &lt;td&gt;It raises competitive pressure, but the HBM total addressable market is also expanding.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;What does it mean for Samsung Electronics catch-up?&lt;/td&gt;
 &lt;td&gt;Samsung must prove HBM4E qualification, DS margin recovery, and customer allocation if it wants the discount to close.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;What does it mean for Korean suppliers?&lt;/td&gt;
 &lt;td&gt;HBM, SSD, SOCAMM, high-capacity DDR5, test, packaging, and power-integrity demand remain supported.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The implication is not one-sided. Micron proved that the US-listed AI-memory premium has real earnings behind it. At the same time, if Korean memory EPS is not impaired, the relative discount of Samsung Electronics and SK Hynix versus Micron may remain too wide.&lt;/p&gt;
&lt;h2 id="next-quarter-checklist"&gt;Next Quarter Checklist
&lt;/h2&gt;&lt;ol&gt;
&lt;li&gt;Does FY4Q26 revenue reach the $50bn guide and EPS reach the $31 guide?&lt;/li&gt;
&lt;li&gt;Do HBM3E and HBM4 remain fully booked through 2027?&lt;/li&gt;
&lt;li&gt;Does HBM4E qualification stay on schedule?&lt;/li&gt;
&lt;li&gt;How much revenue visibility do SCAs actually create?&lt;/li&gt;
&lt;li&gt;Does the price floor protect margin downside more than the ceiling limits upside?&lt;/li&gt;
&lt;li&gt;Can FY2027 capex rise while FCF and buybacks remain credible?&lt;/li&gt;
&lt;li&gt;Does non-HBM DRAM and NAND pricing remain strong?&lt;/li&gt;
&lt;li&gt;Do AI customers continue to demand more memory capacity and bandwidth, or do they shift toward cheaper memory architectures?&lt;/li&gt;
&lt;li&gt;Do Chinese memory suppliers or large customers weaken Micron pricing power?&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="sources-and-limits"&gt;Sources And Limits
&lt;/h2&gt;&lt;p&gt;Core sources include Micron&amp;rsquo;s FY3Q26 official release, SEC Exhibit 99.1, the FY3Q26 presentation, prepared remarks, the Micron IR events page, Reuters coverage, MarketWatch live coverage, and the MarketBeat/Quartr transcript mirror. The Q&amp;amp;A transcript was treated as a web fallback and cross-checked against official materials where possible.&lt;/p&gt;
&lt;p&gt;Regular-session 1D, 2D, and 5D event returns were not available at the analysis time because the release came after the US close. FY4Q26&amp;rsquo;s official date was also not yet confirmed. Data vendors showed different late-September estimates.&lt;/p&gt;
&lt;h2 id="machine-readable-summary"&gt;Machine-Readable Summary
&lt;/h2&gt;&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-json" data-lang="json"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;{
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;as_of_kst&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;2026-06-25 13:59&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;company&amp;#34;&lt;/span&gt;: {
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;name&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;Micron Technology, Inc.&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;ticker&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;MU&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;exchange&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;NASDAQ&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;currency&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;USD&amp;#34;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; },
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;recent_earnings_release&amp;#34;&lt;/span&gt;: {
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;release_datetime_kst&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;2026-06-25 05:01&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;confirmed&amp;#34;&lt;/span&gt;: &lt;span style="color:#66d9ef"&gt;true&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; },
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;recent_call&amp;#34;&lt;/span&gt;: {
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;exists&amp;#34;&lt;/span&gt;: &lt;span style="color:#66d9ef"&gt;true&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;call_datetime_kst&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;2026-06-25 05:30&amp;#34;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; },
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;upcoming_earnings_release&amp;#34;&lt;/span&gt;: {
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;release_datetime_kst&amp;#34;&lt;/span&gt;: &lt;span style="color:#66d9ef"&gt;null&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;confirmed&amp;#34;&lt;/span&gt;: &lt;span style="color:#66d9ef"&gt;false&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; },
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;recommendation&amp;#34;&lt;/span&gt;: {
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;rating&amp;#34;&lt;/span&gt;: &lt;span style="color:#e6db74"&gt;&amp;#34;HOLD&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;target_price&amp;#34;&lt;/span&gt;: &lt;span style="color:#ae81ff"&gt;1350&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;horizon_months&amp;#34;&lt;/span&gt;: &lt;span style="color:#ae81ff"&gt;12&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;upside_pct&amp;#34;&lt;/span&gt;: &lt;span style="color:#ae81ff"&gt;11.2&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; },
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;key_triggers_up&amp;#34;&lt;/span&gt;: [
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;FY4Q26 revenue above $51B or non-GAAP EPS above $32&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;Additional FY2027 HBM3E/HBM4 allocation locked in&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;SCA revenue coverage approaches half or more of company revenue&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;FY4Q free cash flow rises sharply and buyback scale becomes explicit&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;Non-HBM DRAM and NAND pricing remains strong&amp;#34;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ],
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#f92672"&gt;&amp;#34;key_triggers_down&amp;#34;&lt;/span&gt;: [
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;FY4Q26 results miss the midpoint of company guidance&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;Gross margin rolls over materially from the 86% guide&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;FY2027 capex reduces free cash flow and limits buybacks&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;SCA ceiling prices cap upside more than floor prices protect downside&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;AI customers shift toward cheaper memory architectures&amp;#34;&lt;/span&gt;,
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; &lt;span style="color:#e6db74"&gt;&amp;#34;Supply catches up earlier than expected and pricing power weakens&amp;#34;&lt;/span&gt;
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt; ]
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;}
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;h2 id="portfolio-manager-comment"&gt;Portfolio-Manager Comment
&lt;/h2&gt;&lt;p&gt;This is not a normal memory rebound. It is pricing power attached to an AI infrastructure bottleneck. The numbers are excellent, but the market recognized that immediately. SCA makes Micron a better business than the old pure commodity-memory model, but memory remains memory: if supply returns too quickly, the stock can still be brutal. I prefer holding or buying on pullbacks rather than chasing the first after-hours reaction.&lt;/p&gt;
&lt;p&gt;Sources: &lt;a class="link" href="https://www.globenewswire.com/news-release/2026/06/24/3317151/14450/en/micron-technology-inc-reports-record-results-for-the-third-quarter-of-fiscal-2026.html" target="_blank" rel="noopener"
 &gt;Micron FY3Q26 official release&lt;/a&gt;, &lt;a class="link" href="https://investors.micron.com/quarterly-results" target="_blank" rel="noopener"
 &gt;Micron quarterly results&lt;/a&gt;, &lt;a class="link" href="https://investors.micron.com/static-files/2354ecda-77a0-4ddd-8462-a631eb491356" target="_blank" rel="noopener"
 &gt;Micron FY3Q26 presentation&lt;/a&gt;, &lt;a class="link" href="https://investors.micron.com/static-files/631b1a32-5537-46ae-8f40-82e42fc79dfe" target="_blank" rel="noopener"
 &gt;Micron prepared remarks&lt;/a&gt;, &lt;a class="link" href="https://www.marketbeat.com/earnings/reports/2026-6-24-micron-technology-inc-stock/" target="_blank" rel="noopener"
 &gt;MarketBeat/Quartr transcript&lt;/a&gt;, &lt;a class="link" href="https://www.thestar.com.my/tech/tech-news/2026/06/25/micron-forecasts-strong-quarterly-results-on-soaring-memory-chip-demand" target="_blank" rel="noopener"
 &gt;Reuters via The Star&lt;/a&gt;, &lt;a class="link" href="https://m.investing.com/news/stock-market-news/micron-forecasts-quarterly-revenue-above-estimates-4758884?ampMode=1" target="_blank" rel="noopener"
 &gt;Reuters via Investing.com&lt;/a&gt;, &lt;a class="link" href="https://www.marketwatch.com/livecoverage/micron-earnings-stock-results-memory-guidance/card/micron-s-spending-is-moving-higher-in-the-face-of-booming-demand-mLjzhU62y5AJfaUStOEC" target="_blank" rel="noopener"
 &gt;MarketWatch capex note&lt;/a&gt;, &lt;a class="link" href="https://www.marketwatch.com/livecoverage/micron-earnings-stock-results-memory-guidance/card/micron-has-no-clue-when-memory-supply-will-finally-catch-up-with-demand-cKcl56PcCiQzIAvLYTnx" target="_blank" rel="noopener"
 &gt;MarketWatch supply note&lt;/a&gt;.&lt;/p&gt;
&lt;p&gt;&lt;em&gt;Disclaimer: For research and information purposes only. Not investment advice.&lt;/em&gt;&lt;/p&gt;</description></item><item><title>Q2 Earnings: Exporters That Can Still Benefit From KRW Weakness</title><link>https://koreainvestinsights.com/post/krw-weakness-2q-earnings-surprise-export-candidates-2026-06-21/</link><pubDate>Sun, 21 Jun 2026 21:35:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/krw-weakness-2q-earnings-surprise-export-candidates-2026-06-21/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This post follows our work on &lt;a class="link" href="https://koreainvestinsights.com/post/korea-market-liquidity-foreign-reallocation-adr-narrow-leadership-2026-06-03/" &gt;Korea market liquidity and narrow leadership&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samnik-crowding-q3-quality-watchlist-eps-up-price-lag-2026-06-11/" &gt;quality names to revisit in Q3&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-semiconductor-top50-two-month-alpha-candidates-samsung-hynix-2026-06-13/" &gt;Korean Semiconductor Top 50 two-month alpha candidates&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/korea-target-price-gap-comfortable-entry-candidates-2026-06-13/" &gt;target-price gap screens&lt;/a&gt;. It belongs in the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;KRW weakness is supportive for Korean exporters into Q2 earnings. But the trade is not “buy all exporters.” The better screen is narrower: high dollar revenue, meaningful won-denominated costs, improving Q2 volume or pricing, and estimates that have not fully caught up.&lt;/p&gt;
&lt;p&gt;Using FRED DEXKOUS daily observations, the 2026 Q2-to-date USD/KRW average through June 12 is 1,493.7. That is 1.9% above the Q1 average of 1,465.6 and 6.7% above the 2025 Q2 average of 1,399.8. The June 1-12 average is even higher at 1,526.2.&lt;/p&gt;
&lt;p&gt;The final ranking from our screen is: Kolmar Korea, PSK Holdings, Wonik QnC, FADU, LIG Nex1, HD Hyundai Heavy Industries and Pearl Abyss. The cleanest mix is Kolmar Korea. Within semis, PSK Holdings and Wonik QnC look more comfortable than the already-crowded mega-cap and substrate names.&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core View&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 FX is not the variant perception by itself. The variant wedge is &lt;strong&gt;FX plus volume or pricing plus under-reflected consensus&lt;/strong&gt;. Only stocks where those three line up should be treated as Q2 FX-alpha candidates.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-the-fx-setup-is-real"&gt;1. The FX Setup Is Real
&lt;/h2&gt;&lt;p&gt;We recalculated the USD/KRW averages directly from FRED DEXKOUS daily observations. FRED is business-day data, so the averages use available observations.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Period&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Observations&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Average USD/KRW&lt;/th&gt;
 &lt;th&gt;Window&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 Q1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;61&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,465.6&lt;/td&gt;
 &lt;td&gt;Jan 2-Mar 31, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2026 Q2 to date&lt;/td&gt;
 &lt;td style="text-align: right"&gt;52&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,493.7&lt;/td&gt;
 &lt;td&gt;Apr 1-Jun 12, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;2025 Q2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;63&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,399.8&lt;/td&gt;
 &lt;td&gt;Apr 1-Jun 30, 2025&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jun 1-12, 2026&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,526.2&lt;/td&gt;
 &lt;td&gt;Jun 1-Jun 12, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The math is straightforward: Q2-to-date is 1.9% above Q1 and 6.7% above the prior-year Q2. For exporters with dollar revenue and won costs, this creates a translation and margin tailwind, although actual P&amp;amp;L sensitivity still depends on hedging, dollar cost, inventory, shipment timing and revenue recognition.&lt;/p&gt;
&lt;h2 id="2-export-data-also-supports-the-setup"&gt;2. Export Data Also Supports The Setup
&lt;/h2&gt;&lt;p&gt;The FX story would not matter if volume were weak. But recent export data also points in the same direction. Korea&amp;rsquo;s May exports rose 53.2% year over year, with semiconductor exports up 169.4%. May ICT exports rose 128.9%, ICT semiconductor exports rose 169.2%, and computers and peripherals rose 259.6%. For June 1-10, total exports rose 85.9%, semiconductors 205.8%, and computer peripherals 259.4%.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Data point&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Growth&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;May total exports&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+53.2% YoY&lt;/td&gt;
 &lt;td&gt;Export demand is not weak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May semiconductor exports&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+169.4% YoY&lt;/td&gt;
 &lt;td&gt;Semis have price, volume and mix support&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May ICT exports&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+128.9% YoY&lt;/td&gt;
 &lt;td&gt;AI, semis and peripherals lead&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May ICT semiconductors&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+169.2% YoY&lt;/td&gt;
 &lt;td&gt;Memory and system demand are strong&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;May computers/peripherals&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+259.6% YoY&lt;/td&gt;
 &lt;td&gt;Server and storage read-through to check&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jun 1-10 total exports&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+85.9% YoY&lt;/td&gt;
 &lt;td&gt;June started strong&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jun 1-10 semiconductors&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+205.8% YoY&lt;/td&gt;
 &lt;td&gt;Semi momentum persisted into late Q2&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Jun 1-10 computer peripherals&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+259.4% YoY&lt;/td&gt;
 &lt;td&gt;AI server/storage linkage remains relevant&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The investor question is not who benefits from FX. It is who benefits from FX without the benefit already being fully reflected in price and consensus.&lt;/p&gt;
&lt;h2 id="3-where-the-structure-is-best"&gt;3. Where The Structure Is Best
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Segment&lt;/th&gt;
 &lt;th&gt;FX effect&lt;/th&gt;
 &lt;th&gt;Earnings-surprise probability&lt;/th&gt;
 &lt;th&gt;What to check&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Semis/HBM/memory&lt;/td&gt;
 &lt;td&gt;Dollar revenue, partial won costs&lt;/td&gt;
 &lt;td&gt;High, but expectations are high&lt;/td&gt;
 &lt;td&gt;Pricing, shipments, HBM mix, inventory marks&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MLCC/substrates/AI parts&lt;/td&gt;
 &lt;td&gt;Dollar revenue and premium mix&lt;/td&gt;
 &lt;td&gt;High if price and volume improve together&lt;/td&gt;
 &lt;td&gt;AI mix, product mix, orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Defense exports&lt;/td&gt;
 &lt;td&gt;Dollar/euro revenue, won costs&lt;/td&gt;
 &lt;td&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;Delivery schedule, revenue recognition, project margin&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Shipbuilding&lt;/td&gt;
 &lt;td&gt;Dollar backlog&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;td&gt;Hedging, long-cycle recognition, steel cost, progress&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Games/software&lt;/td&gt;
 &lt;td&gt;Overseas revenue, won labor cost&lt;/td&gt;
 &lt;td&gt;Potentially high&lt;/td&gt;
 &lt;td&gt;New title, traffic, launch timing, monetization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cosmetics/food exports&lt;/td&gt;
 &lt;td&gt;Export revenue and won costs&lt;/td&gt;
 &lt;td&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;US/Japan/China channels, raw materials, marketing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The best setup has three traits: dollar revenue and won costs; improving Q2 volume or price; and recent consensus upgrades that have not fully moved the share price.&lt;/p&gt;
&lt;h2 id="4-priority-candidates"&gt;4. Priority Candidates
&lt;/h2&gt;&lt;p&gt;Source: Korea Invest Insights internal screen, as of June 19, 2026 for price, flow and consensus. Flow is trailing 20 trading days in KRW 100mn units. OP is in KRW 100mn.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Core logic&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q2 OP&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY26 P/E&lt;/th&gt;
 &lt;th style="text-align: right"&gt;20D return&lt;/th&gt;
 &lt;th&gt;20D flow&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Kolmar Korea&lt;/td&gt;
 &lt;td&gt;Cosmetics exports, won cost base, lower valuation burden&lt;/td&gt;
 &lt;td style="text-align: right"&gt;923&lt;/td&gt;
 &lt;td style="text-align: right"&gt;12.2x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.4%&lt;/td&gt;
 &lt;td&gt;Foreign +482 / Inst. +133&lt;/td&gt;
 &lt;td&gt;Buy candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;PSK Holdings&lt;/td&gt;
 &lt;td&gt;HBM/back-end equipment, high margin, institutional flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;183&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20.3x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.4%&lt;/td&gt;
 &lt;td&gt;Inst. +346&lt;/td&gt;
 &lt;td&gt;Buy candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Wonik QnC&lt;/td&gt;
 &lt;td&gt;Semi consumables, utilization recovery, less crowded&lt;/td&gt;
 &lt;td style="text-align: right"&gt;280&lt;/td&gt;
 &lt;td style="text-align: right"&gt;15.5x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.9%&lt;/td&gt;
 &lt;td&gt;Inst. +240&lt;/td&gt;
 &lt;td&gt;Buy candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;td&gt;SSD/storage re-rating, strong foreign/program flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;93&lt;/td&gt;
 &lt;td style="text-align: right"&gt;113.4x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-3.6%&lt;/td&gt;
 &lt;td&gt;Foreign +4,120 / Inst. -2,866&lt;/td&gt;
 &lt;td&gt;High-beta candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;LIG Nex1&lt;/td&gt;
 &lt;td&gt;Defense exports, strong institutional and real-money flow&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,052&lt;/td&gt;
 &lt;td style="text-align: right"&gt;53.3x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.6%&lt;/td&gt;
 &lt;td&gt;Foreign +340 / Inst. +1,172&lt;/td&gt;
 &lt;td&gt;Valuation gated&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;HD Hyundai Heavy Industries&lt;/td&gt;
 &lt;td&gt;Shipbuilding, defense and nuclear option, KRW tailwind&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9,933&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23.8x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14.4%&lt;/td&gt;
 &lt;td&gt;Foreign -1,440 / Inst. +1,746&lt;/td&gt;
 &lt;td&gt;Pullback candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;Pearl Abyss&lt;/td&gt;
 &lt;td&gt;Overseas revenue, won labor cost and FX leverage&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1,450&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6.8x&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-17.0%&lt;/td&gt;
 &lt;td&gt;Foreign +79 / Inst. -173&lt;/td&gt;
 &lt;td&gt;Event candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="5-name-by-name-judgment"&gt;5. Name-By-Name Judgment
&lt;/h2&gt;&lt;p&gt;Kolmar Korea is the cleanest mix. It has export-linked beauty manufacturing exposure, a won cost base, a reasonable FY26 P/E of 12.2x, and a -4.4% 20-day price move despite positive foreign and institutional flow. The main diligence items are whether overseas demand flows through to margin and whether raw materials or marketing costs offset the FX benefit.&lt;/p&gt;
&lt;p&gt;PSK Holdings and Wonik QnC are the preferred semi names for this specific FX-alpha screen. They are not as crowded as Samsung Electronics, SK Hynix, Samsung Electro-Mechanics, Daeduck Electronics, Simmtech or Korea Circuit. PSK Holdings adds HBM/back-end equipment exposure and institutional buying. Wonik QnC adds semi consumables and utilization recovery with a lower valuation burden.&lt;/p&gt;
&lt;p&gt;FADU has the biggest right-tail risk but also the largest risk. The foreign buying is powerful, but the institution selling is equally important. This is not a clean long-only accumulation signal. It is a high-beta storage re-rating option that needs SSD controller revenue and customer diversification to show up.&lt;/p&gt;
&lt;p&gt;LIG Nex1 has a good defense-export and FX structure, but valuation is the gating item. At 53.3x FY26 P/E, the company needs more than FX. It needs export orders, project margin and evidence that the European/NATO air-defense channel can become a durable pipeline.&lt;/p&gt;
&lt;p&gt;HD Hyundai Heavy Industries benefits from a dollar backlog, but shipbuilding FX effects are slower because of hedging, progress accounting, contract duration, steel cost and labor. Pearl Abyss has clear overseas-revenue and won-cost leverage, but the stock is really a new-title event story, not a pure FX story.&lt;/p&gt;
&lt;h2 id="6-lower-priority-and-excluded-names"&gt;6. Lower-Priority And Excluded Names
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;iFamilySC&lt;/td&gt;
 &lt;td&gt;Low P/E and export-consumer logic, but flow size is small&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;VT&lt;/td&gt;
 &lt;td&gt;Oversold export-consumer candidate, but flow confirmation is needed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Silicon2&lt;/td&gt;
 &lt;td&gt;Export platform structure is attractive, but recent flow is weak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ST Pharm&lt;/td&gt;
 &lt;td&gt;Possible dollar-revenue structure, but biotech flow needs to recover&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;TLB&lt;/td&gt;
 &lt;td&gt;SOCAMM/substrate candidate, but foreign flow is weak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HD Hyundai&lt;/td&gt;
 &lt;td&gt;Undervalued shipbuilding holding company, more NAV/beta than direct exporter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Reason for exclusion or chase discipline&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics / SK Hynix&lt;/td&gt;
 &lt;td&gt;Q2 surprise probability is high, but expectations and price already reflect a lot&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electro-Mechanics&lt;/td&gt;
 &lt;td&gt;Strong thesis, but price and target upside are less comfortable&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daeduck Electronics&lt;/td&gt;
 &lt;td&gt;Good company, but 20D +24.8% and limited target upside&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Simmtech&lt;/td&gt;
 &lt;td&gt;SOCAMM expectation reflected, 20D +29.2%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korea Circuit&lt;/td&gt;
 &lt;td&gt;20D +68.3%, weaker new-entry expected value&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PSK / Eugene Tech / TES&lt;/td&gt;
 &lt;td&gt;Fundamentals may be good, but short-term price moves are already hot&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="7-final-ranking"&gt;7. Final Ranking
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Action&lt;/th&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Kolmar Korea&lt;/td&gt;
 &lt;td&gt;Buy candidate&lt;/td&gt;
 &lt;td&gt;Best balance of FX, exports, valuation, pullback and flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;PSK Holdings&lt;/td&gt;
 &lt;td&gt;Buy candidate&lt;/td&gt;
 &lt;td&gt;HBM back-end exposure, institutional flow and lower chase risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Wonik QnC&lt;/td&gt;
 &lt;td&gt;Buy candidate&lt;/td&gt;
 &lt;td&gt;Semi consumables recovery, lower valuation and less crowded setup&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;td&gt;High-beta candidate&lt;/td&gt;
 &lt;td&gt;Foreign buying and storage re-rating, but high P/E and institution selling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;LIG Nex1&lt;/td&gt;
 &lt;td&gt;Valuation gated&lt;/td&gt;
 &lt;td&gt;Defense-export quality is strong, valuation is the hurdle&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;HD Hyundai Heavy Industries&lt;/td&gt;
 &lt;td&gt;Pullback candidate&lt;/td&gt;
 &lt;td&gt;Good strategic exposure, but FX P&amp;amp;L recognition is slower&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;Pearl Abyss&lt;/td&gt;
 &lt;td&gt;Event candidate&lt;/td&gt;
 &lt;td&gt;FX structure is good, but the real driver is the new-title event&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The portfolio-manager summary: Q2 KRW weakness is supportive for Korean exporters, but generic exporter exposure is too obvious. The better candidates are companies where FX works with volume, price or margin and where consensus has not fully caught up. Kolmar Korea is the cleanest current setup. In semis, PSK Holdings and Wonik QnC are more comfortable than crowded mega-cap or substrate names. FADU has convexity but needs limited sizing. LIG Nex1 needs valuation discipline.&lt;/p&gt;
&lt;h2 id="missing-evidence"&gt;Missing Evidence
&lt;/h2&gt;&lt;p&gt;This is a screen, not a final trade ticket. The next layer should check company-level hedge ratios, exact dollar revenue and dollar cost exposure, the broker notes behind Q2 consensus, trailing five-day detailed investor flow, and actual entry/exit levels.&lt;/p&gt;
&lt;h2 id="sources-and-data"&gt;Sources And Data
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;FRED DEXKOUS daily observations through June 12, 2026. &lt;a class="link" href="https://fred.stlouisfed.org/series/DEXKOUS" target="_blank" rel="noopener"
 &gt;FRED DEXKOUS&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Korea May export data: total exports +53.2%, semiconductors +169.4%. &lt;a class="link" href="https://eiec.kdi.re.kr/policy/materialView.do?num=281941" target="_blank" rel="noopener"
 &gt;KDI Economic Information Center&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Korea May ICT exports: ICT +128.9%, ICT semiconductors +169.2%, computers/peripherals +259.6%. &lt;a class="link" href="https://www.motir.go.kr/kor/article/ATCL3f49a5a8c/171923/view" target="_blank" rel="noopener"
 &gt;MOTIR&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Korea June 1-10 export data: total exports +85.9%, semiconductors +205.8%, computers/peripherals +259.4%. &lt;a class="link" href="https://eiec.kdi.re.kr/policy/materialView.do?num=282517" target="_blank" rel="noopener"
 &gt;KDI Economic Information Center&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;Stock-level price, flow and consensus: Korea Invest Insights internal screen, as of June 19, 2026. No internal filesystem path is exposed in this public post.&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>TechWing: Can HBM Cube Prober Expand Across the Big Three Memory Makers?</title><link>https://koreainvestinsights.com/post/techwing-hbm-cube-prober-big3-conditional-buy-2026-06-21/</link><pubDate>Sun, 21 Jun 2026 21:10:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/techwing-hbm-cube-prober-big3-conditional-buy-2026-06-21/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This note follows &lt;a class="link" href="https://koreainvestinsights.com/post/korea-semiconductor-top50-two-month-alpha-candidates-samsung-hynix-2026-06-13/" &gt;the Korea Semiconductor Top 50 two-month alpha screen&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/jensen-huang-hbm4-three-vendors-qualified-market-missed-2026-06-05/" &gt;Jensen Huang&amp;rsquo;s HBM4 three-vendor qualification comment&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/samsung-electronics-hbm4e-12h-sample-market-watch-hanmi-tc-bonder-2026-06-01/" &gt;Samsung Electronics&amp;rsquo; HBM4E 12-high sample note&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/gigavis-ai-fcbga-yield-equipment-smart-money-flow-2026-06-05/" &gt;GigaVis as an AI FC-BGA yield bottleneck&lt;/a&gt;. Related hubs: &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-equipment-ip-hub/" &gt;Korea Semiconductor Value Chain Hub&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;Korea AI HBM Hub&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;p&gt;TechWing deserves to move from watchlist to &lt;strong&gt;conditional buy candidate&lt;/strong&gt;. The reason is not a vague HBM theme. The evidence now includes Samsung Electronics supply, the first SK hynix Cube Prober order and Micron qualification work.&lt;/p&gt;
&lt;p&gt;This is still not a blind chase. The stock is at KRW 60,200 after a 25.4% gain over the last 20 trading days and an 11.3% pullback from the recent high. Flow is strong, but it is led by foreigners and programs rather than clearly confirmed long-only real money. The practical entry levels are support at KRW 58,000 to 60,000 or a reclaim of KRW 65,300.&lt;/p&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core Judgment&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The question is not whether TechWing is an HBM theme stock. The question is whether &lt;strong&gt;HBM Cube Prober can move from Samsung expectations into repeat SK hynix orders and Micron mass-production orders&lt;/strong&gt;. If that path is confirmed, TechWing can be reclassified from a memory handler company into an HBM inspection platform company.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-key-data"&gt;1. Key Data
&lt;/h2&gt;&lt;p&gt;The price base is June 19, 2026. Flow data uses the Thesis OS local database through June 17, 2026, with pykrx used to reinforce price data. Some detailed real-money sub-investor fields are incomplete.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reference price&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 60,200&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Market cap&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 2.23tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+25.4%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;60D return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.7%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RSI14&lt;/td&gt;
 &lt;td style="text-align: right"&gt;61.0&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Premium to 20D average&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Premium to 50D average&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Drawdown from 20D high&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.3%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D foreign net buy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 79.2bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D institutional net buy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 18.9bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D program net buy&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 98.2bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026 revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 52.4bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026 operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 9.7bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026 OPM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18.5%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026 net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 6.1bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2026 consensus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Revenue KRW 371.8bn, OP KRW 98.5bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2026 P/E&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 31.4x, based on Naver consensus&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The first conclusion is simple. TechWing is not a stock to buy on 2025 earnings. FY2025 revenue was KRW 159.1bn, operating profit was KRW 15.8bn and net income was KRW 9.0bn. That basis cannot explain a KRW 2.23tn market cap. The market is buying the possibility of a 2026 HBM equipment revenue step-up.&lt;/p&gt;
&lt;p&gt;The picture changes if FY2026 consensus is achieved. If revenue reaches KRW 371.8bn and operating profit reaches KRW 98.5bn, the stock trades near 31x earnings. That is not cheap, but it can be defensible if repeat HBM inspection orders are confirmed.&lt;/p&gt;
&lt;h2 id="2-business-essence-from-memory-handler-to-hbm-inspection-platform"&gt;2. Business Essence: From Memory Handler to HBM Inspection Platform
&lt;/h2&gt;&lt;p&gt;TechWing is not a front-end semiconductor equipment company. It does not draw circuits on wafers. It makes back-end test equipment that moves, holds, sorts and temperature-controls chips so that electrical test systems can inspect them properly.&lt;/p&gt;
&lt;p&gt;In plain terms, the ATE is the grader. TechWing&amp;rsquo;s test handlers and Cube Prober are the automation systems that put the test sheet in the right place, under the right conditions, without damaging the device. That sounds simple, but it requires micrometer-level precision, thermal control, contact stability, throughput and low-damage handling.&lt;/p&gt;
&lt;p&gt;The revenue mix shows the structure.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Product group&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2025 revenue and mix&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q1 2026 revenue and mix&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Semiconductor test equipment, Test Handler and related&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 57.3bn, 36.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 20.5bn, 39.2%&lt;/td&gt;
 &lt;td&gt;Equipment revenue is recovering. HBM and memory test equipment are inside this bucket.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C.O.K&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 58.6bn, 36.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 16.7bn, 31.8%&lt;/td&gt;
 &lt;td&gt;A replaceable interface component that changes with device type. It creates recurring revenue after handlers are installed.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Parts and Board&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 32.3bn, 20.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 10.7bn, 20.5%&lt;/td&gt;
 &lt;td&gt;Parts, boards and service-like revenue attached to the installed base.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;OLED/LCD evaluation and measurement&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 11.0bn, 6.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 4.5bn, 8.5%&lt;/td&gt;
 &lt;td&gt;Less strategic than semiconductor test equipment.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;C.O.K matters. Equipment revenue can be lumpy. But once a handler is installed, changes in device size, package, pin layout and test conditions generate recurring C.O.K demand. The 36.8% C.O.K mix in 2025 shows that TechWing is not just a one-off equipment seller. It has an installed-base revenue layer.&lt;/p&gt;
&lt;p&gt;HBM Cube Prober is the new growth layer. HBM stacks multiple DRAM dies into an expensive package. If a bad die is stacked, the damage is large. That makes pre-stack and intermediate die inspection more valuable. Cube Prober sits exactly at that point.&lt;/p&gt;
&lt;h2 id="3-customer-expansion-samsung-sk-hynix-and-micron"&gt;3. Customer Expansion: Samsung, SK hynix and Micron
&lt;/h2&gt;&lt;p&gt;The most important change in the TechWing thesis is customer expansion. This is no longer just a claim that HBM demand is strong. Customer names are appearing.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Customer&lt;/th&gt;
 &lt;th&gt;Current state&lt;/th&gt;
 &lt;th&gt;Evidence&lt;/th&gt;
 &lt;th&gt;Investment meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td&gt;Supply confirmed&lt;/td&gt;
 &lt;td&gt;KRW 9.72bn HBM inspection equipment contract in May 2026, contract period May 18 to August 13. &lt;a class="link" href="https://www.digitaltoday.co.kr/news/articleView.html?idxno=667207" target="_blank" rel="noopener"
 &gt;DigitalToday&lt;/a&gt;&lt;/td&gt;
 &lt;td&gt;Samsung HBM inspection reference secured&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix&lt;/td&gt;
 &lt;td&gt;First commercial Cube Prober order&lt;/td&gt;
 &lt;td&gt;June 2026 report that TechWing secured its first SK hynix HBM Cube Prober order. Value and volume undisclosed. &lt;a class="link" href="https://www.thelec.net/news/articleView.html?idxno=11441" target="_blank" rel="noopener"
 &gt;The Elec&lt;/a&gt;&lt;/td&gt;
 &lt;td&gt;First signal of expansion into the HBM leader&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron&lt;/td&gt;
 &lt;td&gt;Test equipment supply, Cube Prober evaluation&lt;/td&gt;
 &lt;td&gt;Reports of KRW 22.86bn Micron Malaysia and KRW 11.06bn Micron Xi&amp;rsquo;an test equipment contracts, plus Cube Prober qualification units. &lt;a class="link" href="https://www.thelec.net/news/articleView.html?idxno=11441" target="_blank" rel="noopener"
 &gt;The Elec&lt;/a&gt;&lt;/td&gt;
 &lt;td&gt;If converted into mass-production Cube Prober orders, TechWing completes HBM Big Three exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Existing memory customers&lt;/td&gt;
 &lt;td&gt;Memory Test Handler base&lt;/td&gt;
 &lt;td&gt;Company IR lists SK hynix, Micron and Kioxia as memory handler customers&lt;/td&gt;
 &lt;td&gt;Existing handler base may reinforce HBM inspection growth&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SoC and OSAT customers&lt;/td&gt;
 &lt;td&gt;Expansion option&lt;/td&gt;
 &lt;td&gt;Company IR lists Infineon and Amkor as SoC handler customers&lt;/td&gt;
 &lt;td&gt;Long-term option to reduce pure memory capex dependence&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;According to The Elec, TechWing passed SK hynix qualification in March and secured its first commercial order in June. The same report says Cube Prober tests individual HBM dies, logic dies and DRAM core dies, supports 256-way parallelism and operates from minus 40 to 150 degrees Celsius. Actual chip testing is performed through ATE from companies such as Advantest and Teradyne.&lt;/p&gt;
&lt;p&gt;The caveat is important. The Samsung contract value is confirmed. The SK hynix order is confirmed by report, but value and volume are undisclosed. Micron test equipment contracts and qualification-unit delivery are visible, but a Micron mass-production Cube Prober order has not yet been confirmed.&lt;/p&gt;
&lt;p&gt;Therefore the real questions are:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Does the initial SK hynix order become repeat orders?&lt;/li&gt;
&lt;li&gt;Does Micron qualification convert into mass-production orders?&lt;/li&gt;
&lt;li&gt;Does Samsung Electronics place additional HBM inspection equipment orders?&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id="4-technology-moat-customer-line-validation-matters-more-than-standalone-specs"&gt;4. Technology Moat: Customer-Line Validation Matters More Than Standalone Specs
&lt;/h2&gt;&lt;p&gt;TechWing&amp;rsquo;s moat is not one patent. It is the bundle of precision transfer, vision alignment, temperature control, contact stability, ATE integration, customer qualification data and the C.O.K installed base.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Technology area&lt;/th&gt;
 &lt;th&gt;Content&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Memory Test Handler&lt;/td&gt;
 &lt;td&gt;Company IR emphasizes global leadership in memory test handlers&lt;/td&gt;
 &lt;td&gt;Existing cash cow. Customer qualification and installed base are the moat.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HBM Cube Prober&lt;/td&gt;
 &lt;td&gt;Full inspection solution for individual HBM dies&lt;/td&gt;
 &lt;td&gt;As HBM4 and high-stack structures grow, avoiding bad dies becomes more valuable.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Precision transfer, vision and contact&lt;/td&gt;
 &lt;td&gt;Company IR presents 836 registered IPs and micrometer-level precision&lt;/td&gt;
 &lt;td&gt;Essential for handling thin, expensive devices at speed.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Temperature control&lt;/td&gt;
 &lt;td&gt;Company materials cite wide-range thermal control, ATC and cryogenic chiller technology&lt;/td&gt;
 &lt;td&gt;Important for memory, SoC and automotive semiconductor testing.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C.O.K / TIB&lt;/td&gt;
 &lt;td&gt;Device-specific replaceable interface components&lt;/td&gt;
 &lt;td&gt;Creates recurring revenue after the installed base expands.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The reason inspection matters in HBM is clear. A failed die becomes much more expensive once it is stacked. As HBM4 and 12-high to 16-high stacks become more common, Known Good Die and Known Good Stack Die workflows matter more.&lt;/p&gt;
&lt;p&gt;The customer qualification barrier is also real. Semiconductor test equipment is tied to the customer&amp;rsquo;s test recipe, ATE environment, handling conditions and yield data. Replacing a qualified tool means rechecking contact stability, throughput, thermal uniformity, defect classification and field reliability.&lt;/p&gt;
&lt;p&gt;Competition can still arrive. If the market grows, Cohu, SEMES, Hontec and other test handler or inspection equipment companies may strengthen HBM-oriented tools. Large customers may also prefer internalized or dual-source equipment ecosystems. If HBM4 test flows diverge by customer, today&amp;rsquo;s Cube Prober advantage may not translate equally across all fabs.&lt;/p&gt;
&lt;h2 id="5-earnings-and-valuation-the-operating-turnaround-is-real-but-net-income-quality-needs-proof"&gt;5. Earnings and Valuation: The Operating Turnaround Is Real, But Net Income Quality Needs Proof
&lt;/h2&gt;&lt;p&gt;Q1 2026 was strong at the operating level. According to Yonhap, Q1 revenue was KRW 52.4bn and operating profit was KRW 9.7bn. Revenue rose 51.5% YoY and operating profit rose 444.1% YoY. OPM reached 18.5%. &lt;a class="link" href="https://www.yna.co.kr/view/AKR20260415079200527" target="_blank" rel="noopener"
 &gt;Yonhap&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;But net income was negative KRW 6.1bn. The key issue was derivative losses. DigitalToday reported cumulative derivative losses of KRW 14.741bn from currency forwards and futures, equal to 7.48% of equity. The company described a significant portion as mark-to-market losses on open positions, but FX hedge management is still a valuation discount. &lt;a class="link" href="https://www.digitaltoday.co.kr/news/articleView.html?idxno=659318" target="_blank" rel="noopener"
 &gt;DigitalToday&lt;/a&gt;&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2024&lt;/th&gt;
 &lt;th style="text-align: right"&gt;FY2025&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Q1 2026&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 185.5bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 159.1bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 52.4bn&lt;/td&gt;
 &lt;td&gt;2025 slowed, Q1 2026 recovered strongly&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating profit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 23.4bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 15.8bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 9.7bn&lt;/td&gt;
 &lt;td&gt;Operating leverage returned&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;OPM&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 12.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 10.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;18.5%&lt;/td&gt;
 &lt;td&gt;Mix and fixed-cost absorption likely helped&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Net income&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Positive&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About KRW 9.0bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 6.1bn&lt;/td&gt;
 &lt;td&gt;Derivative losses damaged net income quality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2026 consensus&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-&lt;/td&gt;
 &lt;td style="text-align: right"&gt;Revenue KRW 371.8bn, OP KRW 98.5bn&lt;/td&gt;
 &lt;td&gt;The numbers the current stock price requires&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Valuation must be split into several frames.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Basis&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2025 earnings&lt;/td&gt;
 &lt;td&gt;Expensive. FY2025 net income of KRW 9.0bn cannot explain KRW 2.23tn market cap.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q1 2026 annualized&lt;/td&gt;
 &lt;td&gt;Still not cheap. Q1 operating profit annualizes to about KRW 38.8bn.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FY2026 consensus&lt;/td&gt;
 &lt;td&gt;Conditionally defensible. If OP reaches KRW 98.5bn, the P/E is around 31x.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Post-2026 repeat HBM order path&lt;/td&gt;
 &lt;td&gt;Premium can be justified if SK hynix repeats and Micron converts.&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;So the real question is not whether TechWing is a good company. The real question is whether 2026 operating profit can move toward KRW 100bn and whether that profit remains in net income and cash flow.&lt;/p&gt;
&lt;h2 id="6-flow-and-technical-setup"&gt;6. Flow and Technical Setup
&lt;/h2&gt;&lt;p&gt;Flow is strong, but it needs classification.&lt;/p&gt;
&lt;p&gt;Over the last 20 trading days, foreigners bought KRW 79.2bn and programs bought KRW 98.2bn. Institutions bought KRW 18.9bn. The strongest signal came on June 9, when the stock rose 20.97% and foreigners bought KRW 53.2bn.&lt;/p&gt;
&lt;p&gt;But 40D and 60D real-money flow was not as strong. Current flow looks more like a foreign and program-led re-rating trade than a fully confirmed long-only accumulation.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Flow item&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;20D foreign +KRW 79.2bn&lt;/td&gt;
 &lt;td&gt;Strong buying with price confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D program +KRW 98.2bn&lt;/td&gt;
 &lt;td&gt;Could include basket or mechanical buying&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;20D institution +KRW 18.9bn&lt;/td&gt;
 &lt;td&gt;Positive, but not the main driver&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Partial real-money gaps&lt;/td&gt;
 &lt;td&gt;Long-only accumulation is not fully confirmed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Technically the trend is still alive. The stock is above its 20D, 50D, 120D and 200D moving averages. But after the mid-June high it has fallen below the 5D average and is in a short-term pullback.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Price zone&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 58,000 to 60,000&lt;/td&gt;
 &lt;td&gt;First support zone&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 56,000 to 58,000&lt;/td&gt;
 &lt;td&gt;Break would weaken the short-term trend&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 65,300&lt;/td&gt;
 &lt;td&gt;Reclaim line&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KRW 67,900&lt;/td&gt;
 &lt;td&gt;Recent high. Breakout would restart second-stage momentum&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;That argues for confirmation rather than chase. If the stock holds KRW 58,000 to 60,000 on lower turnover and foreign/program selling slows, a first entry setup appears. If the stock reclaims KRW 65,300 with volume, momentum can restart.&lt;/p&gt;
&lt;h2 id="7-quantum-jump-triggers"&gt;7. Quantum-Jump Triggers
&lt;/h2&gt;&lt;p&gt;For TechWing to move from HBM equipment candidate to HBM inspection platform company, five things need to happen.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Trigger&lt;/th&gt;
 &lt;th&gt;Checkpoint&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron Cube Prober qualification&lt;/td&gt;
 &lt;td&gt;Micron mass-production HBM inspection or Cube Prober order disclosure&lt;/td&gt;
 &lt;td&gt;Completes HBM Big Three exposure&lt;/td&gt;
 &lt;td&gt;Qualification delay, different test flow, price pressure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Repeat Samsung and SK hynix orders&lt;/td&gt;
 &lt;td&gt;Repeated KRW 5bn-plus supply contracts, extended delivery, HBM4 capex expansion&lt;/td&gt;
 &lt;td&gt;Converts initial orders into structural revenue&lt;/td&gt;
 &lt;td&gt;Customer capex delay, competitor tools&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C.O.K revenue stability&lt;/td&gt;
 &lt;td&gt;C.O.K stays above 30% and grows in absolute terms&lt;/td&gt;
 &lt;td&gt;Strengthens recurring revenue base&lt;/td&gt;
 &lt;td&gt;Pricing pressure, lower handler utilization&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SoC, automotive and OSAT handler expansion&lt;/td&gt;
 &lt;td&gt;New SoC handler customers, OSAT orders, automotive or power semiconductor revenue&lt;/td&gt;
 &lt;td&gt;Reduces memory capex dependence&lt;/td&gt;
 &lt;td&gt;Existing global customer lock-in&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Manufacturing capacity and delivery capability&lt;/td&gt;
 &lt;td&gt;Backlog, inventory, capacity expansion and field service capability&lt;/td&gt;
 &lt;td&gt;Moves from good product to scalable product&lt;/td&gt;
 &lt;td&gt;Delivery delay, quality issues, working-capital burden&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The first two matter most. If Micron converts and SK hynix repeats, the market can view TechWing as a candidate standard vendor for HBM inspection.&lt;/p&gt;
&lt;h2 id="8-risks-and-watch-items"&gt;8. Risks and Watch Items
&lt;/h2&gt;&lt;p&gt;The risks fall into three buckets.&lt;/p&gt;
&lt;h3 id="technology-risk"&gt;Technology risk
&lt;/h3&gt;&lt;p&gt;Cube Prober qualification is customer-specific. Success at Samsung or SK hynix does not automatically guarantee success at Micron. HBM stack structure, test recipes, ATE configuration and temperature conditions can differ by customer.&lt;/p&gt;
&lt;p&gt;High-parallel contact and thin-die handling are also difficult. As parallelism rises, contact stability, physical damage, thermal uniformity and defect-classification accuracy all become harder at the same time.&lt;/p&gt;
&lt;h3 id="business-risk"&gt;Business risk
&lt;/h3&gt;&lt;p&gt;Customer concentration is high. TechWing depends heavily on large memory makers. Media estimates of Micron revenue mix can be useful, but they are not the same as official customer-level revenue disclosure.&lt;/p&gt;
&lt;p&gt;Order disclosure is also lumpy. Semiconductor equipment revenue depends on customer order timing, manufacturing, delivery, acceptance and revenue recognition. The SK hynix Cube Prober order was reported without disclosed value or volume.&lt;/p&gt;
&lt;h3 id="financial-risk"&gt;Financial risk
&lt;/h3&gt;&lt;p&gt;Derivative losses should not be ignored. If operating profit improves but net income is damaged by FX hedge losses for multiple quarters, the market will reduce the premium.&lt;/p&gt;
&lt;p&gt;Working capital also matters when equipment orders rise quickly. Inventory, receivables, prepayments and outsourced manufacturing cost can rise before cash collection. Revenue growth must convert into operating cash flow.&lt;/p&gt;
&lt;p&gt;Key checkpoints for the next quarter and half-year are:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Checkpoint&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron Cube Prober qualification and order&lt;/td&gt;
 &lt;td&gt;The biggest event for HBM Big Three exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Additional Samsung and SK hynix orders&lt;/td&gt;
 &lt;td&gt;Confirms repeat revenue beyond initial orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 and Q3 product mix&lt;/td&gt;
 &lt;td&gt;Shows whether equipment and C.O.K grow together&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Operating margin durability&lt;/td&gt;
 &lt;td&gt;Tests whether Q1 2026 OPM of 18.5% is structural&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Inventory, receivables and backlog&lt;/td&gt;
 &lt;td&gt;Checks whether growth creates cash-flow burden&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cube Prober standalone revenue disclosure&lt;/td&gt;
 &lt;td&gt;Helps separate HBM equipment contribution&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Financial income and derivative losses&lt;/td&gt;
 &lt;td&gt;Shows whether operating profit becomes net income&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="9-investment-view"&gt;9. Investment View
&lt;/h2&gt;&lt;p&gt;The stance is &lt;strong&gt;Conditional Buy / buy-on-pullback candidate&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;TechWing is no longer just a thematic name. Samsung&amp;rsquo;s HBM inspection contract, the first SK hynix Cube Prober order and Micron qualification work are real customer-expansion signals. Q1 2026 operating profit has recovered, and FY2026 consensus can partly justify the current valuation if it is achieved.&lt;/p&gt;
&lt;p&gt;But the conditions matter. The current stock price already reflects part of the re-rating. Flow is good, but foreign/program-led. Long-only real-money accumulation is not fully confirmed. Net income quality is also not clean because of derivative losses.&lt;/p&gt;
&lt;h3 id="entry"&gt;Entry
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Hold KRW 58,000 to 60,000&lt;/td&gt;
 &lt;td&gt;Lower turnover, slower foreign/program selling, intraday support&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reclaim KRW 65,300&lt;/td&gt;
 &lt;td&gt;Higher turnover, foreign or institutional buying, ideally with follow-up news&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 2026 earnings proof&lt;/td&gt;
 &lt;td&gt;Revenue growth, 15 to 20% OPM and smaller derivative losses&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="catalysts"&gt;Catalysts
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Catalyst&lt;/th&gt;
 &lt;th&gt;What to check&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;June 29, 2026 IR&lt;/td&gt;
 &lt;td&gt;Cube Prober, Memory Test Handler and customer progress updates&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK hynix follow-up orders&lt;/td&gt;
 &lt;td&gt;Whether the first order becomes repeat purchasing&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron qualification&lt;/td&gt;
 &lt;td&gt;Whether evaluation units become mass-production orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung additional supply&lt;/td&gt;
 &lt;td&gt;Whether HBM inspection equipment orders repeat&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q2 2026 earnings&lt;/td&gt;
 &lt;td&gt;Whether operating profit meets or beats KRW 13.3bn consensus&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="invalidation"&gt;Invalidation
&lt;/h3&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Invalidation&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Break below KRW 58,000 without recovery&lt;/td&gt;
 &lt;td&gt;Short-term flow turn weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;No follow-up after SK hynix initial order&lt;/td&gt;
 &lt;td&gt;HBM platform re-rating weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Micron qualification delay or failure&lt;/td&gt;
 &lt;td&gt;Big Three expansion logic weakens&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Large Q2 operating-profit miss&lt;/td&gt;
 &lt;td&gt;FY2026 consensus credibility falls&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Repeated derivative losses&lt;/td&gt;
 &lt;td&gt;Operating profit does not become net income&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Rising short interest plus falling price&lt;/td&gt;
 &lt;td&gt;Foreign/program buying quality needs review&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-comment"&gt;Final Comment
&lt;/h2&gt;&lt;p&gt;TechWing is one of the more interesting names in the Korean HBM equipment group. The positive point is that this is no longer only expectation. Customer and order evidence has started to appear. If Samsung, SK hynix and Micron all remain in the picture, TechWing can be reclassified from a memory handler company into an HBM inspection platform company.&lt;/p&gt;
&lt;p&gt;But the stock already reflects part of that re-rating. A good company story is no longer enough. The next proof must come from numbers: Q2 to Q3 earnings, repeat orders, Micron conversion, C.O.K revenue durability and smaller derivative losses.&lt;/p&gt;
&lt;p&gt;The conclusion is clear. TechWing can move from watchlist to conditional buy candidate. But chasing is not the clean setup. The better setup is support confirmation around KRW 58,000 to 60,000. If the June 29 IR confirms SK hynix follow-up orders, Micron progress and Cube Prober revenue visibility, the case for adding size improves.&lt;/p&gt;</description></item><item><title>KOSPI 60-Day Disparity At +28.6%: Not A Top Call, But A Partial Risk-Reduction Signal</title><link>https://koreainvestinsights.com/post/kospi-disparity60-overheat-framework-partial-trim-2026-06-21/</link><pubDate>Sun, 21 Jun 2026 02:35:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/kospi-disparity60-overheat-framework-partial-trim-2026-06-21/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/kospi-benchmark-hard-to-beat-narrow-market-monte-carlo-2026-06-20/" &gt;How Rare Is It To Beat The Pure KOSPI Benchmark?&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-etf-flow-led-market-volatility-strategy-2026-06-13/" &gt;ETF Flow Is Leading The Korean Market&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-market-liquidity-foreign-reallocation-adr-narrow-leadership-2026-06-03/" &gt;Korea Has Liquidity, But Breadth Has Broken&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/foreign-return-after-24-day-kospi-selling-memory-rebalance-2026-06-12/" &gt;Have Foreign Investors Returned?&lt;/a&gt;. Related hubs: &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/page/korea-daily-market-hub/" &gt;Korea Daily Market Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;As of June 19, 2026, KOSPI closed at &lt;strong&gt;9,052&lt;/strong&gt;, its 60-day moving average was &lt;strong&gt;7,039&lt;/strong&gt;, and its 60-day disparity was &lt;strong&gt;+28.6%&lt;/strong&gt;. That is clear medium-term overspeed.&lt;/li&gt;
&lt;li&gt;But disparity overheat is not automatically a market top. Reproducing 355 trading days from January 2, 2025 to June 19, 2026 shows that 20/60/120-day overheat signals were followed by positive average 5-day and 20-day returns. In a strong trend, overheat can first mean acceleration.&lt;/li&gt;
&lt;li&gt;However, when 60-day disparity rises above +20%, the probability of a -5% drawdown within the next 10 trading days rises from a 16% baseline to &lt;strong&gt;63%&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;That 63% must be handled carefully. The 41 signal days are not 41 independent observations. They compress into &lt;strong&gt;8 episodes&lt;/strong&gt;, and only 6 had completed forward windows.&lt;/li&gt;
&lt;li&gt;The current read is therefore: &lt;strong&gt;partial risk reduction is justified, but this is not a top call.&lt;/strong&gt; Disparity and volatility gates are on; the 20-day moving-average slope is still rising.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core Point&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 KOSPI's +28.6% 60-day disparity does not say &lt;strong&gt;"this is the top"&lt;/strong&gt;. It says &lt;strong&gt;the trend has become fast enough that investors should reduce chase risk, harvest part of the risk budget, and prepare for a pullback window&lt;/strong&gt;.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-what-was-tested"&gt;1. What was tested
&lt;/h2&gt;&lt;p&gt;The analysis independently reproduced a KOSPI disparity framework. The questions were:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Do the current KOSPI disparity numbers tie out?&lt;/li&gt;
&lt;li&gt;Did disparity overheat historically raise correction probability?&lt;/li&gt;
&lt;li&gt;Should the current reading be treated as a sell signal?&lt;/li&gt;
&lt;/ol&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Setting&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Index&lt;/td&gt;
 &lt;td&gt;KOSPI, &lt;code&gt;^KS11&lt;/code&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Period&lt;/td&gt;
 &lt;td&gt;2025-01-02 to 2026-06-19&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Sample&lt;/td&gt;
 &lt;td&gt;355 trading days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price&lt;/td&gt;
 &lt;td&gt;Close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity&lt;/td&gt;
 &lt;td&gt;&lt;code&gt;close / N-day moving average - 1&lt;/code&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Windows&lt;/td&gt;
 &lt;td&gt;20, 60, 120 days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Correction definition&lt;/td&gt;
 &lt;td&gt;A -5% or worse low within the next 10 trading days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Baseline&lt;/td&gt;
 &lt;td&gt;Same correction frequency across all valid days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is a market-risk framework, not a trading instruction.&lt;/p&gt;
&lt;h2 id="2-current-state-the-core-numbers-tie-out"&gt;2. Current state: the core numbers tie out
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Original&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Reproduction&lt;/th&gt;
 &lt;th&gt;Read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI close&lt;/td&gt;
 &lt;td style="text-align: right"&gt;9,052&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;9,052&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Matched&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MA20 / disparity20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;8,322 / +8.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;8,322 / +8.77%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Matched&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MA60 / disparity60&lt;/td&gt;
 &lt;td style="text-align: right"&gt;7,039 / +28.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;7,039 / +28.61%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Matched&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MA120 / disparity120&lt;/td&gt;
 &lt;td style="text-align: right"&gt;6,088 / +48.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;6,088 / +48.69%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Matched&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MA20 5-day slope&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+4.25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+4.25%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Matched&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RSI(14)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;57&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;SMA 57.3 / Wilder 65.9&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Formula difference&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ATR(20)%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3.61%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;Wilder 4.15 / SMA 4.70&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Formula difference&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The key moving-average and disparity numbers tie out almost exactly. The RSI difference is explainable by formula choice: SMA RSI is close to 57, while Wilder RSI is 65.9. Both remain below 70. The conclusion is unchanged: this is not primarily an RSI overbought signal. It is a medium-term disparity signal.&lt;/p&gt;
&lt;h2 id="3-overheat-did-not-predict-tops-well"&gt;3. Overheat did not predict tops well
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Signal&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Signal days&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Avg +5D return&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Avg +20D return&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Negative +20D share&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity20 ≥ 7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;74&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.68%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.07%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;24%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 ≥ 12%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;120&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.26%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.88%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;22%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity120 ≥ 20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;107&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.55%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.11%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;23%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Baseline&lt;/td&gt;
 &lt;td style="text-align: right"&gt;335&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.86%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.83%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;19%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;This is the first uncomfortable point. In this 2025-2026 trend, overheat did not reliably identify the top. Average returns after overheat signals were still positive.&lt;/p&gt;
&lt;p&gt;That makes sense in a strong trend. A large distance above moving averages often means money is still being forced into the leaders.&lt;/p&gt;
&lt;h2 id="4-but-60-day-disparity-did-raise-pullback-odds"&gt;4. But 60-day disparity did raise pullback odds
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Condition&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Signal days&lt;/th&gt;
 &lt;th style="text-align: right"&gt;-5% correction within 10 trading days&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Baseline&lt;/td&gt;
 &lt;td style="text-align: right"&gt;All valid days&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;16%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 ≥ 15%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;89&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;36%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 ≥ 18%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;54&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;48%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 ≥ 20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;41&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;63%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 ≥ 22%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;33&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;67%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 ≥ 25%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;26&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;65%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The signal is useful, but not as a top-calling tool. It is a pullback-risk warning. Above +20% disparity60, the probability of a -5% drawdown within 10 trading days rose about fourfold versus the baseline.&lt;/p&gt;
&lt;h2 id="5-the-statistical-caveat-41-days-are-only-8-episodes"&gt;5. The statistical caveat: 41 days are only 8 episodes
&lt;/h2&gt;&lt;p&gt;The 41 days of +20% disparity are not independent. Grouped into runs, they become 8 episodes:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;#&lt;/th&gt;
 &lt;th&gt;Episode&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Signal days&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Peak disparity60&lt;/th&gt;
 &lt;th&gt;-5% correction within 10D&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;2025-11-03&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+22.3%&lt;/td&gt;
 &lt;td&gt;Yes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;2026-01-28 to 01-30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.6%&lt;/td&gt;
 &lt;td&gt;Yes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;2026-02-03 to 02-04&lt;/td&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+22.8%&lt;/td&gt;
 &lt;td&gt;Yes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;2026-02-12 to 03-03&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+33.6%&lt;/td&gt;
 &lt;td&gt;Yes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;2026-05-04 to 05-18&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+32.1%&lt;/td&gt;
 &lt;td&gt;Yes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;6&lt;/td&gt;
 &lt;td&gt;2026-05-21 to 06-05&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+36.6%&lt;/td&gt;
 &lt;td&gt;Yes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;7&lt;/td&gt;
 &lt;td&gt;2026-06-09&lt;/td&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+21.6%&lt;/td&gt;
 &lt;td&gt;Not complete&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;8&lt;/td&gt;
 &lt;td&gt;2026-06-15 to 06-18&lt;/td&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+29.9%&lt;/td&gt;
 &lt;td&gt;Not complete&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The six completed episodes all produced a -5% correction. That is strong, but the sample is tiny. The right lesson is not to hard-code exact thresholds. The right lesson is to treat extreme 60-day disparity as a risk-budget tightening signal.&lt;/p&gt;
&lt;h2 id="6-which-indicators-mattered"&gt;6. Which indicators mattered?
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Indicator&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Pre-correction average&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Full-sample average&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Difference&lt;/th&gt;
 &lt;th&gt;Signal quality&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+20.35&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+11.33&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.02&lt;/td&gt;
 &lt;td&gt;Strong&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ATR%(Wilder)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.56&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.03&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.53&lt;/td&gt;
 &lt;td&gt;Medium to strong&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MA20 slope&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.74&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.81&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.93&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+5.06&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.50&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+1.56&lt;/td&gt;
 &lt;td&gt;Weak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RSI(Wilder)&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+64.53&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+65.46&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-0.93&lt;/td&gt;
 &lt;td&gt;Little signal&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The ranking is intuitive. A 20-day average follows price quickly, so it gets digested fast. A 60-day average moves more slowly, so extreme 60-day disparity captures medium-term overspeed better. RSI was not the useful axis in this sample.&lt;/p&gt;
&lt;h2 id="7-the-current-three-gate-read"&gt;7. The current three-gate read
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Gate&lt;/th&gt;
 &lt;th&gt;Rule&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Current value&lt;/th&gt;
 &lt;th&gt;Status&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Medium-term overspeed&lt;/td&gt;
 &lt;td&gt;Disparity60 ≥ 18-20%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+28.6%&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;On&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Volatility expansion&lt;/td&gt;
 &lt;td&gt;ATR ≥ 2.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;4% range&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;On&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Trend deterioration&lt;/td&gt;
 &lt;td&gt;MA20 slope slowing or turning down&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+4.25% rising&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Not yet&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Two gates are on. That argues against aggressive chasing.&lt;/p&gt;
&lt;p&gt;But the third gate is not on. The 20-day moving-average slope is still rising. Therefore the current signal is not a broad-market exit signal. It is a partial rebalancing signal.&lt;/p&gt;
&lt;h2 id="8-use-dynamic-levels-not-fixed-index-levels"&gt;8. Use dynamic levels, not fixed index levels
&lt;/h2&gt;&lt;p&gt;One common mistake is to hard-code an index level. If the 60-day moving average is 7,039 and the re-entry zone is roughly 8% above it, investors may anchor on something like KOSPI 7,600.&lt;/p&gt;
&lt;p&gt;But moving averages move. In a strong trend, the 60-day average will keep rising.&lt;/p&gt;
&lt;div class="highlight"&gt;&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;-webkit-text-size-adjust:none;"&gt;&lt;code class="language-text" data-lang="text"&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Static frame: KOSPI around 7,600
&lt;/span&gt;&lt;/span&gt;&lt;span style="display:flex;"&gt;&lt;span&gt;Dynamic frame: the then-current 60-day moving average plus roughly 8%
&lt;/span&gt;&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;The dynamic frame is better because the pullback zone moves up as the trend matures.&lt;/p&gt;
&lt;h2 id="9-why-this-matters-more-in-a-narrow-market"&gt;9. Why this matters more in a narrow market
&lt;/h2&gt;&lt;p&gt;The current KOSPI overheat is not a broad-market overheat. It is heavily influenced by Samsung Electronics and SK Hynix.&lt;/p&gt;
&lt;p&gt;That links directly to the &lt;a class="link" href="https://koreainvestinsights.com/post/kospi-benchmark-hard-to-beat-narrow-market-monte-carlo-2026-06-20/" &gt;pure KOSPI benchmark analysis&lt;/a&gt;. In 2026, KOSPI has behaved less like the average Korean stock and more like a concentrated AI-memory benchmark.&lt;/p&gt;
&lt;p&gt;So the next useful test is not just KOSPI disparity. It is:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Is KOSPI overheated?&lt;/td&gt;
 &lt;td&gt;Index-level risk budget&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Is KOSPI ex Samsung/SK Hynix overheated?&lt;/td&gt;
 &lt;td&gt;Broad-market versus mega-cap effect&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Are memory mega-caps driving the signal?&lt;/td&gt;
 &lt;td&gt;Whether the signal is narrow leadership&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Are second-line stocks also stretched?&lt;/td&gt;
 &lt;td&gt;Whether this is a broad extension&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="10-practical-usage"&gt;10. Practical usage
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Situation&lt;/th&gt;
 &lt;th&gt;Read&lt;/th&gt;
 &lt;th&gt;Action frame&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 12-18%&lt;/td&gt;
 &lt;td&gt;Strong trend&lt;/td&gt;
 &lt;td&gt;Hold and verify trend&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 18-20%&lt;/td&gt;
 &lt;td&gt;Medium-term overspeed&lt;/td&gt;
 &lt;td&gt;Reduce new chase risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 20%+ with high ATR&lt;/td&gt;
 &lt;td&gt;Pullback probability rising&lt;/td&gt;
 &lt;td&gt;Reduce part of risk budget&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 20%+ and MA20 slope slows&lt;/td&gt;
 &lt;td&gt;Distribution risk rising&lt;/td&gt;
 &lt;td&gt;Consider further reduction&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Disparity60 cools to 8-12%&lt;/td&gt;
 &lt;td&gt;Overheat easing&lt;/td&gt;
 &lt;td&gt;Re-underwrite with flow and earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Break below the 60-day average&lt;/td&gt;
 &lt;td&gt;Different regime&lt;/td&gt;
 &lt;td&gt;Stop using simple pullback-buy rules&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The one-line rule: &lt;strong&gt;60-day disparity overheat is not a short signal. It is a rebalancing signal inside a strong trend.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="final-view"&gt;Final view
&lt;/h2&gt;&lt;p&gt;KOSPI&amp;rsquo;s +28.6% 60-day disparity is real overheat. But the reproduced data does not support treating it as a confirmed top.&lt;/p&gt;
&lt;p&gt;Overheat signals were followed by positive average returns in this trend. At the same time, +20% or greater 60-day disparity materially raised the odds of a -5% pullback within 10 trading days.&lt;/p&gt;
&lt;p&gt;The right conclusion is balanced:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;This is not the zone to abandon the bull market. It is the zone to stop chasing, harvest part of the risk budget, and wait for either trend deterioration or a cleaner pullback.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="coverage-health"&gt;Coverage Health
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;As of: June 19, 2026 close.&lt;/li&gt;
&lt;li&gt;Data: KOSPI &lt;code&gt;^KS11&lt;/code&gt;, 2025-01-02 to 2026-06-19, 355 trading days.&lt;/li&gt;
&lt;li&gt;Reproduced: 20/60/120-day disparity, RSI 14, ATR 20%, 5-day slope of the 20-day moving average, and -5% forward-10D correction events.&lt;/li&gt;
&lt;li&gt;Confirmed: current disparity values, post-overheat forward returns, disparity60 threshold correction rates, RSI formula difference.&lt;/li&gt;
&lt;li&gt;Not yet confirmed: 2020-2022 cross-validation, KOSPI ex Samsung Electronics/SK Hynix disparity, and robustness of re-entry rules in a true bear-market transition.&lt;/li&gt;
&lt;li&gt;This is a market-risk framework, not investment advice.&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>How Rare Is It To Beat The Pure KOSPI Benchmark In This Narrow Market?</title><link>https://koreainvestinsights.com/post/kospi-benchmark-hard-to-beat-narrow-market-monte-carlo-2026-06-20/</link><pubDate>Sat, 20 Jun 2026 10:30:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/kospi-benchmark-hard-to-beat-narrow-market-monte-carlo-2026-06-20/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context&lt;br&gt;
This is a follow-up to our narrow-market, foreign-flow and semiconductor relative-alpha work. Related hub: &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Korea looks like a roaring bull market at the index level. As of June 19, 2026, KOSPI traded around the 9,050 level and was up roughly 115% YTD.&lt;/li&gt;
&lt;li&gt;Under the index, the market is extremely narrow. In our proprietary 2,730-stock Korea dataset, the median stock return was &lt;strong&gt;-7.6%&lt;/strong&gt;, only &lt;strong&gt;37%&lt;/strong&gt; of names were up, and only about &lt;strong&gt;30%&lt;/strong&gt; traded above their 200-day moving average.&lt;/li&gt;
&lt;li&gt;Monte Carlo simulations show how hard the pure KOSPI benchmark was to beat. A random equal-weight 20-stock portfolio beat KOSPI only &lt;strong&gt;1.0%&lt;/strong&gt; of the time. A 30-stock version beat it only &lt;strong&gt;0.5%&lt;/strong&gt; of the time.&lt;/li&gt;
&lt;li&gt;Turnover-weighted portfolios did much better, with 10-stock and 30-stock versions beating KOSPI &lt;strong&gt;60.1%&lt;/strong&gt; and &lt;strong&gt;70.8%&lt;/strong&gt; of the time. But that was essentially a concentrated exposure to the most popular mega-cap AI-memory winners.&lt;/li&gt;
&lt;li&gt;The conclusion: in 2026, pure KOSPI is not “the average Korean equity market.” It is closer to a concentrated Samsung Electronics / SK Hynix AI-memory benchmark.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core View&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 This year's KOSPI was not the average path. It was the narrowest path. Beating it required being heavily and persistently exposed to the Samsung Electronics, SK Hynix and high-turnover AI-memory winner cohort.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="the-index-and-the-median-stock-diverged"&gt;The Index And The Median Stock Diverged
&lt;/h2&gt;&lt;p&gt;The KOSPI level around 9,050 on June 19 is visible on external index screens such as Investing.com and Trading Economics. KOSDAQ, by contrast, was near 966.6 and up roughly 11% YTD on public screens.&lt;/p&gt;
&lt;p&gt;Our proprietary breadth dataset shows the internal divergence.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Metric&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Value&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSPI YTD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;around +115%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;KOSDAQ YTD&lt;/td&gt;
 &lt;td style="text-align: right"&gt;around +11%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Median Korean stock&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;-7.6%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Advancing-stock ratio&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;37%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cumulative A-D line&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;-20,391&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Stocks above MA200&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;30%&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Top-decile return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+88% or more&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Top-1% return&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;+537% or more&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The gap between KOSPI and the median Korean stock is roughly 122 percentage points. That is not a normal broad bull market. It is a narrow leadership regime.&lt;/p&gt;
&lt;h2 id="monte-carlo-result"&gt;Monte Carlo Result
&lt;/h2&gt;&lt;p&gt;We simulated random Korean equity portfolios from a 2,730-stock universe, using returns from the start of 2026 to June 19. Each case ran 20,000 trials.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Portfolio type&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Holdings&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Beat KOSPI&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Median portfolio return&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Equal-weight random&lt;/td&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;5.3%&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.6%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Equal-weight random&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;2.8%&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+9.8%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Equal-weight random&lt;/td&gt;
 &lt;td style="text-align: right"&gt;20&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;1.0%&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+14.1%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Equal-weight random&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;0.5%&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.9%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Turnover-weighted&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;60.1%&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+127%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Turnover-weighted&lt;/td&gt;
 &lt;td style="text-align: right"&gt;30&lt;/td&gt;
 &lt;td style="text-align: right"&gt;&lt;strong&gt;70.8%&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+131%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Only &lt;strong&gt;7.9%&lt;/strong&gt; of individual stocks beat the KOSPI return threshold.&lt;/p&gt;
&lt;p&gt;The interpretation is straightforward. Diversification reduced benchmark-relative performance because the benchmark itself was heavily concentrated in a tiny winner cohort. Turnover weighting did better because it leaned into the same crowded winners that drove the index.&lt;/p&gt;
&lt;h2 id="what-it-means"&gt;What It Means
&lt;/h2&gt;&lt;p&gt;An active account lagging pure KOSPI in 2026 was not necessarily a failed account. It may have simply avoided the extreme concentration embedded in the benchmark.&lt;/p&gt;
&lt;p&gt;The right evaluation has three layers:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Question&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Absolute return&lt;/td&gt;
 &lt;td&gt;Did the account make money?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Median-stock comparison&lt;/td&gt;
 &lt;td&gt;Did it beat the typical Korean stock?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Pure KOSPI comparison&lt;/td&gt;
 &lt;td&gt;Did it beat the mega-cap AI-memory benchmark?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Those are different questions. A diversified account could do very well in absolute terms and still lag pure KOSPI.&lt;/p&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;In 2026, beating the pure KOSPI benchmark required more than good stock picking. It required concentration in the narrow mega-cap AI-memory path, early entry and long holding periods. Diversification was not wrong, but it carried a relative-performance cost.&lt;/p&gt;
&lt;p&gt;Coverage: proprietary 2,730-stock Korea dataset, price returns from the start of 2026 to June 19, 2026, and 20,000 Monte Carlo trials per portfolio structure. External index levels checked against &lt;a class="link" href="https://www.investing.com/indices/kospi-historical-data" target="_blank" rel="noopener"
 &gt;Investing.com&lt;/a&gt;, &lt;a class="link" href="https://tradingeconomics.com/south-korea/stock-market" target="_blank" rel="noopener"
 &gt;Trading Economics&lt;/a&gt;, &lt;a class="link" href="https://finance.yahoo.com/quote/%5EKQ11/" target="_blank" rel="noopener"
 &gt;Yahoo Finance KOSDAQ&lt;/a&gt; and &lt;a class="link" href="https://www.marketwatch.com/investing/index/kq11?countrycode=xx" target="_blank" rel="noopener"
 &gt;MarketWatch KOSDAQ&lt;/a&gt;.&lt;/p&gt;</description></item><item><title>Rheinmetall and LIG: Korea’s Medium- and Long-Range Missile Channel Into Europe and NATO</title><link>https://koreainvestinsights.com/post/rheinmetall-lig-europe-nato-air-defense-channel-2026-06-16/</link><pubDate>Tue, 16 Jun 2026 10:05:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/rheinmetall-lig-europe-nato-air-defense-channel-2026-06-16/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This follows &lt;a class="link" href="https://koreainvestinsights.com/post/us-nonsemi-rerating-ai-power-software-korea-translation-2026-05-31/" &gt;the U.S. non-semiconductor rerating translation into Korea&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/spacex-ipo-korea-market-liquidity-ai-space-readthrough-2026-06-05/" &gt;SpaceX IPO and Korea&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/us-investment-special-act-team-korea-nuclear-opportunity-2026-06-13/" &gt;Korea’s U.S. strategic investment law and Team Korea nuclear opportunity&lt;/a&gt;, and &lt;a class="link" href="https://koreainvestinsights.com/post/construction-epc-rerating-ai-power-nuclear-datacenter-reconstruction-2026-06-14/" &gt;the infrastructure EPC rerating note&lt;/a&gt;. The common question is where Korean companies fit into the global security, power and AI-infrastructure rebuild.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;This is not a simple defence MOU. Rheinmetall’s official release states that Rheinmetall and LIG Defense&amp;amp;Aerospace will work together to address &lt;strong&gt;European and NATO customers&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;The short-term aim is a &lt;strong&gt;joint venture with Rheinmetall holding the majority shares&lt;/strong&gt;. Through that structure, LIG’s &lt;strong&gt;medium- and long-range air-defence missile systems&lt;/strong&gt; are to be localized, further developed and marketed in Europe.&lt;/li&gt;
&lt;li&gt;The release also says these systems will be combined with Rheinmetall’s &lt;strong&gt;VSHORAD&lt;/strong&gt; portfolio, while both companies will co-develop new missiles and capabilities in the &lt;strong&gt;SHORAD&lt;/strong&gt; segment.&lt;/li&gt;
&lt;li&gt;This matters because Europe is facing a structural multi-layer air-defence gap after Ukraine: missiles, drones, ammunition stockpiles, local production and NATO interoperability all have to be rebuilt at the same time.&lt;/li&gt;
&lt;li&gt;What is not yet confirmed: contract value, customer country, delivery schedule or order volume. The confirmed event is a &lt;strong&gt;channel architecture&lt;/strong&gt;, not a booked order.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core Point&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 The Rheinmetall-LIG partnership creates a route to package Korean air-defence missiles in European procurement language. It is not yet revenue. It is a structural channel change.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-why-europe-has-to-act-this-way"&gt;1. Why Europe Has To Act This Way
&lt;/h2&gt;&lt;p&gt;Europe’s air-defence problem is not just about higher defence budgets. The character of the threat has changed.&lt;/p&gt;
&lt;p&gt;Russia’s war against Ukraine has shown simultaneous use of ballistic missiles, cruise missiles, glide bombs, drones and aircraft threats. One expensive system cannot economically intercept every layer. Using high-end interceptors against low-cost drones exhausts the defender. At the same time, short-range systems cannot solve medium- and long-range missile threats.&lt;/p&gt;
&lt;p&gt;Europe therefore needs &lt;strong&gt;layered air defence&lt;/strong&gt;, not just one “better missile.”&lt;/p&gt;
&lt;p&gt;NATO describes Integrated Air and Missile Defence, or IAMD, as a continuous mission to protect territory, populations and forces from air and missile threats. NATO also states that Russia’s war against Ukraine has underlined the importance of robust air and missile defence and that Allies have reinforced the eastern flank. (&lt;a class="link" href="https://www.nato.int/en/what-we-do/deterrence-and-defence/nato-integrated-air-and-missile-defence" target="_blank" rel="noopener"
 &gt;NATO IAMD&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The EU is moving in the same direction. The European Defence Readiness 2030 agenda and White Paper emphasize urgent capability gaps, joint procurement, and the need for European industry to deliver at speed and scale. European Air Shield and the European Drone Defence Initiative sit in that framework. (&lt;a class="link" href="https://defence-industry-space.ec.europa.eu/eu-defence-industry/white-paper-european-defence-readiness-2030_en" target="_blank" rel="noopener"
 &gt;European Commission&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;EDIRPA also identifies air and missile defence as an urgent procurement area following Russia’s aggression against Ukraine. (&lt;a class="link" href="https://defence-industry-space.ec.europa.eu/eu-defence-industry/edirpa-addressing-capability-gaps_en" target="_blank" rel="noopener"
 &gt;European Commission EDIRPA&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The conclusion is straightforward.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;European need&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;th&gt;Link to Rheinmetall-LIG&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Layered defence&lt;/td&gt;
 &lt;td&gt;Drones, cruise missiles, ballistic missiles and aircraft require different layers&lt;/td&gt;
 &lt;td&gt;LIG MRAD/LRAD plus Rheinmetall VSHORAD&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Missile stockpiles and production&lt;/td&gt;
 &lt;td&gt;In a war of attrition, interceptors and reloads matter as much as launchers&lt;/td&gt;
 &lt;td&gt;JV and localization matter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Political procurement&lt;/td&gt;
 &lt;td&gt;European customers want local production, a European prime and NATO interoperability&lt;/td&gt;
 &lt;td&gt;Rheinmetall supplies the procurement wrapper&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="2-what-the-official-release-confirms"&gt;2. What The Official Release Confirms
&lt;/h2&gt;&lt;p&gt;Rheinmetall’s June 15, 2026 release is titled: “Rheinmetall and LIG Defense&amp;amp;Aerospace join forces to address European and NATO customers.” (&lt;a class="link" href="https://www.rheinmetall.com/en/media/news-watch/news/2026/06/2026-06-15-rheinmetall-and-lig-defense-aerospace-establish-strategic-partnership" target="_blank" rel="noopener"
 &gt;Rheinmetall&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The confirmed items are:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Official item&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Strategic partnership between Rheinmetall and LIG D&amp;amp;A&lt;/td&gt;
 &lt;td&gt;A joint market-access structure, not only a product pitch&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Short-term aim of a Rheinmetall-majority JV&lt;/td&gt;
 &lt;td&gt;A European procurement and localization vehicle&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Localization, further development and marketing of LIG MRAD/LRAD systems in Europe&lt;/td&gt;
 &lt;td&gt;LIG systems could be adapted to European customer requirements&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Combination with Rheinmetall VSHORAD and co-development in SHORAD&lt;/td&gt;
 &lt;td&gt;A multi-layer portfolio, from very short range to long range&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The important words are &lt;strong&gt;JV&lt;/strong&gt;, &lt;strong&gt;Europe and NATO territory&lt;/strong&gt;, &lt;strong&gt;localize&lt;/strong&gt;, &lt;strong&gt;further develop&lt;/strong&gt; and &lt;strong&gt;market&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;Those words shift the nature of the news. This is not simply “LIG wants to sell missiles to Europe.” It is a plan to create a Rheinmetall-majority European channel through which LIG missile systems can be localized, developed and sold.&lt;/p&gt;
&lt;p&gt;The release also describes LIG D&amp;amp;A as “formerly known as LIG Nex1” and directly mentions L-SAM, MSAM-II and CHIRON. That frames LIG not as a parts supplier, but as a multi-layer air-defence missile-system provider.&lt;/p&gt;
&lt;h2 id="3-why-this-is-a-channel-change"&gt;3. Why This Is A Channel Change
&lt;/h2&gt;&lt;p&gt;Defence exports usually move through several stages: direct export, local assembly, and then local joint development with a trusted prime.&lt;/p&gt;
&lt;p&gt;This agreement is closer to the third stage.&lt;/p&gt;
&lt;p&gt;Rheinmetall has European customer access, local production credibility, procurement trust and integration capability. LIG has medium- and long-range missile technology, but it is an external supplier in European procurement. European customers care about performance, but also about local production, maintenance, NATO interoperability, and political reliability.&lt;/p&gt;
&lt;p&gt;That is why a Rheinmetall-majority JV matters.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;LIG alone&lt;/th&gt;
 &lt;th&gt;Rheinmetall-LIG JV&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean system sold abroad&lt;/td&gt;
 &lt;td&gt;European-procurable air-defence system&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Customer-by-customer negotiation&lt;/td&gt;
 &lt;td&gt;Access to Rheinmetall’s European network&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Performance and price focus&lt;/td&gt;
 &lt;td&gt;Performance, localization, interoperability and political procurement logic&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Export product&lt;/td&gt;
 &lt;td&gt;Localization, co-development, marketing and potential upgrades&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The accurate wording is not “LIG is already inside NATO air defence.” The accurate wording is: &lt;strong&gt;LIG now has an official route to present its air-defence systems in European and NATO procurement language.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="4-the-layered-air-defence-logic"&gt;4. The Layered Air-Defence Logic
&lt;/h2&gt;&lt;p&gt;Air defence is not one layer.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Role&lt;/th&gt;
 &lt;th&gt;Threats&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;VSHORAD&lt;/td&gt;
 &lt;td&gt;Very short-range protection&lt;/td&gt;
 &lt;td&gt;Low-altitude drones, helicopters, close air threats&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SHORAD&lt;/td&gt;
 &lt;td&gt;Short-range defence&lt;/td&gt;
 &lt;td&gt;Drones, some cruise missiles, aircraft, glide-bomb-related threats&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;MRAD&lt;/td&gt;
 &lt;td&gt;Medium-range defence&lt;/td&gt;
 &lt;td&gt;Aircraft, cruise missiles, some ballistic threats&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;LRAD&lt;/td&gt;
 &lt;td&gt;Long-range defence&lt;/td&gt;
 &lt;td&gt;High-altitude aircraft, ballistic missiles, long-range threats&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Rheinmetall is strong in very-short-range and ground-based air defence, sensors, guns, ammunition and integration. LIG can add missile depth through M-SAM, L-SAM and CHIRON-related capabilities.&lt;/p&gt;
&lt;p&gt;That is why the release explicitly combines LIG MRAD/LRAD with Rheinmetall VSHORAD and adds SHORAD co-development.&lt;/p&gt;
&lt;p&gt;For a European customer, the package could be more attractive than a standalone missile product. A single European prime can propose a broader air-defence portfolio that covers several layers and can be connected to NATO’s IAMD architecture.&lt;/p&gt;
&lt;h2 id="5-what-this-means-for-korean-defence"&gt;5. What This Means For Korean Defence
&lt;/h2&gt;&lt;p&gt;Korean defence’s first major rerating wave was platform exports: artillery, tanks, armoured vehicles, ships and aircraft.&lt;/p&gt;
&lt;p&gt;This is different. It is a &lt;strong&gt;missile and air-defence layer&lt;/strong&gt; story.&lt;/p&gt;
&lt;p&gt;Air-defence missiles have different economics:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;War consumes interceptor inventories.&lt;/li&gt;
&lt;li&gt;Reload missiles can create repeated demand.&lt;/li&gt;
&lt;li&gt;Cost-per-kill matters as much as range.&lt;/li&gt;
&lt;li&gt;NATO systems require sensor, C2, IFF and data-link interoperability.&lt;/li&gt;
&lt;li&gt;Local production and long-term maintenance heavily influence procurement.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The Rheinmetall-LIG partnership therefore broadens the K-defense narrative from platform exports to missile-layer access.&lt;/p&gt;
&lt;p&gt;The read-through should not be automatic, however. The direct parties in the release are Rheinmetall and LIG D&amp;amp;A. Possible second-order effects for sensors, C2, radar, electronic warfare or propulsion suppliers still require separate evidence.&lt;/p&gt;
&lt;h2 id="6-what-is-still-unconfirmed"&gt;6. What Is Still Unconfirmed
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Unknown&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Final JV establishment and timing&lt;/td&gt;
 &lt;td&gt;Turns the partnership into an operating entity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;JV economics&lt;/td&gt;
 &lt;td&gt;Determines how value is shared between Rheinmetall and LIG&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;First European or NATO customer&lt;/td&gt;
 &lt;td&gt;Converts channel optionality into real demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Actual localized product scope&lt;/td&gt;
 &lt;td&gt;Which of L-SAM, M-SAM II or CHIRON becomes Europeanized&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Production scope&lt;/td&gt;
 &lt;td&gt;Assembly, core components, maintenance and upgrades have different economics&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NATO C2 and sensor integration&lt;/td&gt;
 &lt;td&gt;Required for IAMD adoption&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Contract value and delivery timing&lt;/td&gt;
 &lt;td&gt;Needed for revenue recognition&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The right interpretation is:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Rheinmetall and LIG D&amp;amp;A have formalized a channel to localize, develop and market medium- and long-range Korean air-defence missile systems for Europe and NATO customers.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;This is one of the stronger Korean defence news items of 2026 because it changes the channel structure, not because it comes with an immediate contract value.&lt;/p&gt;
&lt;p&gt;Europe has to rebuild layered air defence, interceptor production and local defence supply chains. NATO and the EU are pushing air and missile defence, drone defence, joint procurement and European production capacity at the same time. In that environment, Rheinmetall brings the European procurement wrapper and LIG brings medium- and long-range missile technology.&lt;/p&gt;
&lt;p&gt;The key sentence is:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;LIG has gained an official route to become a candidate supplier for the medium- and long-range layers of Europe and NATO’s air-defence architecture.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;It is still a route, not a confirmed order. The next evidence points are the JV, first customer country, product scope, localization scope and contract value.&lt;/p&gt;
&lt;h2 id="sources"&gt;Sources
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;&lt;a class="link" href="https://www.rheinmetall.com/en/media/news-watch/news/2026/06/2026-06-15-rheinmetall-and-lig-defense-aerospace-establish-strategic-partnership" target="_blank" rel="noopener"
 &gt;Rheinmetall, Strategic partnership with LIG Defense&amp;amp;Aerospace, 2026-06-15&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://www.nato.int/en/what-we-do/deterrence-and-defence/nato-integrated-air-and-missile-defence" target="_blank" rel="noopener"
 &gt;NATO Integrated Air and Missile Defence&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://defence-industry-space.ec.europa.eu/eu-defence-industry/white-paper-european-defence-readiness-2030_en" target="_blank" rel="noopener"
 &gt;European Commission, White Paper for European Defence: Readiness 2030&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a class="link" href="https://defence-industry-space.ec.europa.eu/eu-defence-industry/edirpa-addressing-capability-gaps_en" target="_blank" rel="noopener"
 &gt;European Commission, EDIRPA: addressing capability gaps&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>Is Korea’s Construction Rerating Coming? The Answer Is Nuclear, Power and AI Infrastructure EPC</title><link>https://koreainvestinsights.com/post/construction-epc-rerating-ai-power-nuclear-datacenter-reconstruction-2026-06-14/</link><pubDate>Sun, 14 Jun 2026 22:30:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/construction-epc-rerating-ai-power-nuclear-datacenter-reconstruction-2026-06-14/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Connection
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/us-investment-special-act-team-korea-nuclear-opportunity-2026-06-13/" &gt;Korea’s $350bn U.S. strategic investment law and Team Korea nuclear opportunity&lt;/a&gt; and &lt;a class="link" href="https://koreainvestinsights.com/post/hyundai-ec-woojin-us-nuclear-expansion-bargaining-power-2026-06-13/" &gt;Hyundai E&amp;amp;C and Woojin in U.S. nuclear expansion&lt;/a&gt;. Also read &lt;a class="link" href="https://koreainvestinsights.com/post/korea-etf-flow-led-market-volatility-strategy-2026-06-13/" &gt;ETF flows are driving Korea’s volatility&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/ai-datacenter-capex-5p3t-korea-power-substrate-storage-bottleneck-2026-06-05/" &gt;AI data center capex and Korea’s bottlenecks&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/us-nonsemi-rerating-ai-power-software-korea-translation-2026-05-31/" &gt;U.S. non-semiconductor rerating translated into Korea&lt;/a&gt;, the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-daily-market-hub/" &gt;Korea Daily Market Hub&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;The conclusion is &lt;strong&gt;not a broad construction-sector rerating&lt;/strong&gt;. The names that can rerate are those being reclassified from housing contractors into &lt;strong&gt;nuclear, SMR, AI data center power, energy EPC and reconstruction EPC&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;The old frame was &lt;code&gt;construction = domestic housing + PF risk + low-margin EPC + low ROE&lt;/code&gt;. The new frame needs to become &lt;code&gt;construction = nuclear/SMR + AI data centers + power infrastructure + energy EPC + reconstruction + development/operation income&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;The clearest axes are &lt;strong&gt;Hyundai E&amp;amp;C = U.S. nuclear/SMR&lt;/strong&gt;, &lt;strong&gt;Samsung C&amp;amp;T = data centers, SMR and renewable development&lt;/strong&gt;, &lt;strong&gt;DL E&amp;amp;C = undervalued reconstruction and SMR optionality&lt;/strong&gt;, and &lt;strong&gt;Samsung E&amp;amp;A = Middle East and energy EPC optionality&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;This is not a sector-wide buy signal. &lt;strong&gt;Hyundai E&amp;amp;C is the cleanest direct exposure but has chase risk&lt;/strong&gt;, &lt;strong&gt;DL E&amp;amp;C has the best asymmetry&lt;/strong&gt;, &lt;strong&gt;Samsung C&amp;amp;T is the highest-quality platform but already reflects much of the story&lt;/strong&gt;, and &lt;strong&gt;Samsung E&amp;amp;A needs flow confirmation&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;Ranking: &lt;strong&gt;Hyundai E&amp;amp;C pullback buy&lt;/strong&gt;, &lt;strong&gt;DL E&amp;amp;C conditional buy&lt;/strong&gt;, &lt;strong&gt;Samsung C&amp;amp;T hold/pullback only&lt;/strong&gt;, &lt;strong&gt;Samsung E&amp;amp;A wait&lt;/strong&gt;, and &lt;strong&gt;GS E&amp;amp;C/Daewoo E&amp;amp;C as trading candidates&lt;/strong&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core View&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 A construction rerating can happen. But the multiple goes to &lt;strong&gt;power, nuclear, data center, energy and reconstruction EPC operators&lt;/strong&gt;, not to generic housing contractors.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-why-revisit-construction-now"&gt;1. Why revisit construction now?
&lt;/h2&gt;&lt;p&gt;Korean construction stocks have been assigned low multiples for a long time. The reasons were clear: domestic housing cyclicality, PF risk, low-margin overseas EPC, and weak ROE.&lt;/p&gt;
&lt;p&gt;That old market frame was:&lt;/p&gt;
&lt;p&gt;&lt;code&gt;construction = domestic housing + PF risk + low-margin EPC + low ROE&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;Under that frame, rerating is hard. Housing is cyclical, PF losses rarely end in one quarter, and overseas EPC can destroy margin when costs move against the contractor.&lt;/p&gt;
&lt;p&gt;But the 2026 question is different:&lt;/p&gt;
&lt;p&gt;&lt;code&gt;construction = nuclear/SMR + data centers + power infrastructure + energy EPC + reconstruction EPC&lt;/code&gt;&lt;/p&gt;
&lt;p&gt;That changes the debate. The new drivers are not domestic apartments. They are &lt;strong&gt;electricity for AI data centers&lt;/strong&gt;, &lt;strong&gt;the U.S. nuclear renaissance&lt;/strong&gt;, &lt;strong&gt;Korea-U.S. policy capital&lt;/strong&gt;, &lt;strong&gt;Ukraine and Middle East reconstruction&lt;/strong&gt;, and &lt;strong&gt;energy infrastructure such as renewables, ESS, gas and ammonia&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;The important point is selectivity. This is not a blanket buy call on all builders. It is a &lt;strong&gt;reclassification trade&lt;/strong&gt;. The market can price a housing-heavy contractor very differently from an EPC operator tied to nuclear, power, data centers and energy infrastructure.&lt;/p&gt;
&lt;h2 id="2-what-must-happen-for-rerating-p--q--c"&gt;2. What must happen for rerating: P × Q × C
&lt;/h2&gt;&lt;p&gt;The rerating needs price, quantity and cost to move together.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Factor&lt;/th&gt;
 &lt;th&gt;Rerating condition&lt;/th&gt;
 &lt;th&gt;What to verify&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;P, price&lt;/td&gt;
 &lt;td&gt;The market pays an infrastructure EPC multiple, not a housing multiple&lt;/td&gt;
 &lt;td&gt;Is the company being reclassified as nuclear, data center or energy infrastructure EPC?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Q, quantity&lt;/td&gt;
 &lt;td&gt;U.S. nuclear, SMR, data center, reconstruction and energy EPC orders increase&lt;/td&gt;
 &lt;td&gt;Do MOUs convert into FEED, EPC, O&amp;amp;M, equipment, development equity or recurring work?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;C, cost&lt;/td&gt;
 &lt;td&gt;Domestic housing losses, PF provisions and overseas cost overruns stabilize&lt;/td&gt;
 &lt;td&gt;Does backlog convert into profit?&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Construction stocks can move on order headlines. But true rerating happens after the order: backlog needs to expand, the mix needs to improve, margin needs to hold, and housing losses need to stop offsetting the new story.&lt;/p&gt;
&lt;p&gt;So the key question is:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Which Korean construction names can be read as AI power, nuclear and infrastructure EPC rather than housing stocks?&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="3-catalyst-1-korea-us-strategic-investment-and-us-nuclear"&gt;3. Catalyst 1: Korea-U.S. strategic investment and U.S. nuclear
&lt;/h2&gt;&lt;p&gt;The Korea-U.S. strategic investment law is the first catalyst.&lt;/p&gt;
&lt;p&gt;LawTimes describes the law as the legal framework for implementing the &lt;strong&gt;$350bn&lt;/strong&gt; strategic investment package agreed under the Korea-U.S. MOU: &lt;strong&gt;$200bn&lt;/strong&gt; of U.S.-bound strategic direct investment and &lt;strong&gt;$150bn&lt;/strong&gt; of shipbuilding cooperation investment. (&lt;a class="link" href="https://www.lawtimes.co.kr/news/articleView.html?idxno=217827" target="_blank" rel="noopener"
 &gt;LawTimes&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Korea’s Ministry of Trade, Industry and Resources said the implementing decree establishes a 20-year Korea-U.S. Strategic Investment Corporation, a statutory capital base of KRW 2tn, and a structure that can delegate work to KEXIM, KDB, K-SURE, KIC, KOBC and KIND. The ministry said the law and decree are expected to take effect on &lt;strong&gt;June 18, 2026&lt;/strong&gt;. (&lt;a class="link" href="https://www.motir.go.kr/kor/article/ATCL3f49a5a8c/171908/view" target="_blank" rel="noopener"
 &gt;MOTIR&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Two caveats matter.&lt;/p&gt;
&lt;p&gt;First, this is not a guaranteed order book for Korean companies. MOTIR also notes that actual projects will be determined after commercial-rationality reviews, operating committee deliberation, National Assembly reporting and U.S. negotiations.&lt;/p&gt;
&lt;p&gt;Second, energy, AI and supply-chain projects are natural candidates. AI data centers require electricity, and the U.S. is pushing nuclear permitting and nuclear supply-chain rebuilding at the same time.&lt;/p&gt;
&lt;p&gt;The U.S. Department of Energy has framed the policy goal as expanding U.S. nuclear capacity from around 100GW today to 400GW by 2050, while discussing TerraPower, Holtec, TVA, Palisades, uprates and AI-linked energy demand. (&lt;a class="link" href="https://www.energy.gov/ne/articles/one-year-after-executive-orders-us-nuclear-energy-renaissance-full-swing" target="_blank" rel="noopener"
 &gt;U.S. DOE&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Hyundai E&amp;amp;C has the most direct connection. The company announced a FEED contract with Fermi America for four large nuclear reactors inside an integrated energy and AI campus. The project combines AP1000 nuclear 4GW, SMR 2GW, combined-cycle gas 4GW, solar and BESS 1GW, for an 11GW independent power infrastructure tied to hyperscale AI data centers. Hyundai E&amp;amp;C said the parties are discussing an EPC contract targeted for the first half of next year. (&lt;a class="link" href="https://www.hdec.kr/kr/newsroom/news_view.aspx?NewsListType=news_list&amp;amp;NewsSeq=1370&amp;amp;NewsType=BRAND" target="_blank" rel="noopener"
 &gt;Hyundai E&amp;amp;C&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;That is the type of project that can change the market’s language. It is not housing. It is AI campus power, nuclear, SMR, gas, solar, batteries and data centers in one package.&lt;/p&gt;
&lt;h2 id="4-catalyst-2-ai-data-center-power-bottlenecks"&gt;4. Catalyst 2: AI data center power bottlenecks
&lt;/h2&gt;&lt;p&gt;The second catalyst is the data center power bottleneck.&lt;/p&gt;
&lt;p&gt;IEA estimates global data center electricity consumption at about &lt;strong&gt;415TWh in 2024&lt;/strong&gt; and projects it to reach about &lt;strong&gt;945TWh by 2030&lt;/strong&gt; in its Base Case. The same report notes that data centers can be operational in two to three years, while the broader energy system requires longer planning, permitting and construction lead times. (&lt;a class="link" href="https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai" target="_blank" rel="noopener"
 &gt;IEA&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;That is why EPC matters.&lt;/p&gt;
&lt;p&gt;AI data centers are not built with GPUs alone. They need land, grid access, substations, transmission, cooling, backup power, gas generation, renewables, batteries, nuclear, mechanical and electrical systems, piping and project management.&lt;/p&gt;
&lt;p&gt;But the market does not simply buy old-style construction. In the U.S., the rerating has concentrated in power grid, substation, electrical/mechanical, cooling, piping and modular construction names, not in generic EPC.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;U.S. peer&lt;/th&gt;
 &lt;th&gt;Rerating driver&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Quanta Services&lt;/td&gt;
 &lt;td&gt;Grid, transmission, substations, renewables EPC&lt;/td&gt;
 &lt;td&gt;Direct AI power-grid bottleneck&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EMCOR&lt;/td&gt;
 &lt;td&gt;Electrical/mechanical systems, data centers, industrial infrastructure&lt;/td&gt;
 &lt;td&gt;Data center systems and project execution&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Comfort Systems&lt;/td&gt;
 &lt;td&gt;Mechanical systems, HVAC, cooling, piping&lt;/td&gt;
 &lt;td&gt;Data center cooling and power-density bottleneck&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Fluor&lt;/td&gt;
 &lt;td&gt;Traditional EPC, energy/infrastructure&lt;/td&gt;
 &lt;td&gt;Can rerate, but premium is more limited&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The message is simple.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The market is buying bottleneck execution, not construction as a label.&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;In Korea, that frame points to Samsung C&amp;amp;T, Hyundai E&amp;amp;C, Samsung E&amp;amp;A, DL E&amp;amp;C and HanmiGlobal. The multiple depends on whether each name solves a real data center, power, nuclear or energy infrastructure bottleneck.&lt;/p&gt;
&lt;h2 id="5-catalyst-3-ukraine-and-middle-east-reconstruction-are-options"&gt;5. Catalyst 3: Ukraine and Middle East reconstruction are options
&lt;/h2&gt;&lt;p&gt;Reconstruction matters, but it requires more caution.&lt;/p&gt;
&lt;p&gt;The World Bank, the European Commission, the UN and Ukraine’s government estimate Ukraine’s reconstruction and recovery needs at almost &lt;strong&gt;$588bn&lt;/strong&gt; over the next decade as of end-2025. Housing, transport and energy are among the most affected sectors, and energy-sector damaged or destroyed assets increased by around 21% versus the prior assessment. (&lt;a class="link" href="https://www.worldbank.org/en/news/press-release/2026/02/23/updated-ukraine-recovery-and-reconstruction-needs-assessment-released" target="_blank" rel="noopener"
 &gt;World Bank&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;That is a large TAM, but it is not yet revenue for listed companies.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stage&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;th&gt;What investors need&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reconstruction need&lt;/td&gt;
 &lt;td&gt;TAM&lt;/td&gt;
 &lt;td&gt;Not revenue yet&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Multilateral/government financing&lt;/td&gt;
 &lt;td&gt;Payment capacity&lt;/td&gt;
 &lt;td&gt;MDB, ECA, guarantees and insurance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Project package&lt;/td&gt;
 &lt;td&gt;Real opportunity&lt;/td&gt;
 &lt;td&gt;Energy, roads, ports, housing or industrial sites&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Contract type&lt;/td&gt;
 &lt;td&gt;Margin and risk&lt;/td&gt;
 &lt;td&gt;EPC, EPCM, PM, equipment, O&amp;amp;M&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue recognition&lt;/td&gt;
 &lt;td&gt;Earnings&lt;/td&gt;
 &lt;td&gt;Cost inflation, FX, security and sanctions risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The same applies to the Middle East and Iran. Samsung E&amp;amp;A, DL E&amp;amp;C, Hyundai E&amp;amp;C and Daewoo E&amp;amp;C can benefit from energy EPC or reconstruction headlines. But without sanctions clarity, financing, guarantees and order structure, this remains an option.&lt;/p&gt;
&lt;p&gt;So the reconstruction axis should be framed this way:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Reconstruction can move stock prices. But the core is still AI power, nuclear and energy EPC.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="6-korean-candidate-matrix"&gt;6. Korean candidate matrix
&lt;/h2&gt;&lt;p&gt;The table below uses Thesis OS local DB as of the June 12, 2026 close. The flow column looks at recent 10-day foreign plus real-money flow, giving more weight to quality institutional money than to headline institutional totals.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Price&lt;/th&gt;
 &lt;th style="text-align: right"&gt;20D&lt;/th&gt;
 &lt;th style="text-align: right"&gt;60D&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Target upside&lt;/th&gt;
 &lt;th style="text-align: right"&gt;10D foreign + real money&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 157,500&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-2.0%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+40.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 160.5bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Top priority, but no chase&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung C&amp;amp;T&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 432,000&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+0.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+55.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+12.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 10.6bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Best quality, less comfortable price&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;DL E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 73,900&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+57.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+69.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 42.9bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Best asymmetry&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung E&amp;amp;A&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 47,600&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-10.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+51.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+39.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 49.5bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Good theme, wait for flow&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;GS E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 28,800&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-13.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+28.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+69.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 47.6bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Housing beta plus reconstruction option&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 21,850&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-24.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+82.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+58.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 78.4bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Event-driven, high volatility&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HanmiGlobal&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 19,340&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-21.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+89.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 1.5bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Data center/reconstruction PM option, weak liquidity&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Target upside alone is not enough. A large gap often exists because the market still discounts housing risk, PF risk, margin volatility or weak liquidity.&lt;/p&gt;
&lt;h2 id="7-hyundai-ec-the-clearest-nuclear-and-smr-leader"&gt;7. Hyundai E&amp;amp;C: the clearest nuclear and SMR leader
&lt;/h2&gt;&lt;p&gt;Hyundai E&amp;amp;C is the most direct name in this theme.&lt;/p&gt;
&lt;p&gt;Its classification is &lt;strong&gt;Idiosyncratic Alpha&lt;/strong&gt;. It has the Fermi large nuclear FEED contract, Holtec SMR-300 cooperation, and a U.S. nuclear execution story.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Strength&lt;/th&gt;
 &lt;th&gt;Why it matters&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Direct exposure&lt;/td&gt;
 &lt;td&gt;The Fermi America FEED contract is official&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AI power linkage&lt;/td&gt;
 &lt;td&gt;Fermi is an AI data center plus independent power infrastructure project&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nuclear EPC scarcity&lt;/td&gt;
 &lt;td&gt;Large nuclear EPC and PMO execution capability is scarce&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Policy capital fit&lt;/td&gt;
 &lt;td&gt;The project aligns with U.S.-bound energy and AI investment themes&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flow&lt;/td&gt;
 &lt;td&gt;10D foreign + real money is +KRW 160.5bn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;But a good company and a good entry point are different things. Hyundai E&amp;amp;C has already moved as the nuclear/SMR leader. After a sharp short-term rally, chasing is less efficient.&lt;/p&gt;
&lt;p&gt;View: &lt;strong&gt;Wait / Pullback Buy&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;The key is whether FEED converts to EPC. FEED is important, but EPC contract scope, risk sharing and margin determine earnings. If Hyundai E&amp;amp;C is recognized as the execution bottleneck, not merely a low-margin contractor, the multiple can move higher.&lt;/p&gt;
&lt;h2 id="8-dl-ec-the-cheapest-reconstruction-and-smr-option"&gt;8. DL E&amp;amp;C: the cheapest reconstruction and SMR option
&lt;/h2&gt;&lt;p&gt;DL E&amp;amp;C has the best asymmetry among the candidates.&lt;/p&gt;
&lt;p&gt;Its classification is &lt;strong&gt;Idiosyncratic Alpha&lt;/strong&gt;. It is less direct than Hyundai E&amp;amp;C, but valuation is more comfortable. Based on the local dataset used here, DL E&amp;amp;C trades around 8x forward PER and 0.6x PBR, while target upside is +69.8%. The 20D return is -11.2%, while the 60D return is +57.2%, which means the larger trend is alive but the near-term heat has cooled.&lt;/p&gt;
&lt;p&gt;DL E&amp;amp;C has two axes.&lt;/p&gt;
&lt;p&gt;First, reconstruction optionality. If Ukraine or Middle East reconstruction turns into funded projects, discounted construction and plant EPC names can rerate quickly. But this remains an option until contracts, financing and margins are visible.&lt;/p&gt;
&lt;p&gt;Second, SMR optionality. X-energy has highlighted Korean partners including DL E&amp;amp;C and Doosan Enerbility, and DL E&amp;amp;C is tied to SMR standardization design and collaboration. It is less direct than Hyundai E&amp;amp;C’s Fermi exposure, but lower valuation increases sensitivity to positive news.&lt;/p&gt;
&lt;p&gt;View: &lt;strong&gt;Conditional Buy&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;DL E&amp;amp;C is less direct, but the price is more forgiving. If the trade moves from “buy the leader” to “rerate undervalued second-line EPC,” DL E&amp;amp;C can have better expected return.&lt;/p&gt;
&lt;h2 id="9-samsung-ct-highest-quality-less-comfortable-entry"&gt;9. Samsung C&amp;amp;T: highest quality, less comfortable entry
&lt;/h2&gt;&lt;p&gt;Samsung C&amp;amp;T is a different type of stock.&lt;/p&gt;
&lt;p&gt;Its classification is &lt;strong&gt;Quality Compounder&lt;/strong&gt;. It is not just a builder. It is closer to a Samsung-group NAV, high-tech EPC, data center, renewable, ESS, IPP and SMR-option platform.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Axis&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Data centers&lt;/td&gt;
 &lt;td&gt;North American development and power infrastructure optionality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Renewables/ESS&lt;/td&gt;
 &lt;td&gt;Solar, storage, power sales and development&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SMR/nuclear&lt;/td&gt;
 &lt;td&gt;Energy-transition projects and Samsung group optionality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;High-tech EPC&lt;/td&gt;
 &lt;td&gt;Semiconductor, battery and bio manufacturing EPC record&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NAV&lt;/td&gt;
 &lt;td&gt;Samsung group stake value and shareholder-return optionality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The issue is price. The local dataset shows Samsung C&amp;amp;T is up +55.1% over 60D, with target upside narrowed to +12.4%. Recent 10D foreign + real-money flow is -KRW 10.6bn, suggesting consolidation rather than fresh accumulation.&lt;/p&gt;
&lt;p&gt;View: &lt;strong&gt;Hold / Pullback only&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;The long-term quality is strong. But for new entry, Hyundai E&amp;amp;C and DL E&amp;amp;C are more compelling. Samsung C&amp;amp;T becomes more interesting after a pullback or after an official data-center development catalyst.&lt;/p&gt;
&lt;h2 id="10-samsung-ea-good-energy-epc-theme-but-flow-has-not-turned"&gt;10. Samsung E&amp;amp;A: good energy EPC theme, but flow has not turned
&lt;/h2&gt;&lt;p&gt;Samsung E&amp;amp;A matters for the U.S. peer read-through.&lt;/p&gt;
&lt;p&gt;If Quanta, EMCOR and Comfort Systems can rerate because they solve power, mechanical and data center bottlenecks, Samsung E&amp;amp;A needs to be read as more than chemical EPC. The rerating path requires energy transition, hydrogen/ammonia, gas, data-center power systems and Samsung-group high-tech infrastructure.&lt;/p&gt;
&lt;p&gt;The research input is constructive. Samsung E&amp;amp;A’s 1Q26 revenue was KRW 2.2674tn, operating profit KRW 188.2bn and net profit KRW 163.3bn. New orders were KRW 4.6tn and backlog around KRW 20.6tn. That backlog base is not weak.&lt;/p&gt;
&lt;p&gt;But stock selection also needs flow.&lt;/p&gt;
&lt;p&gt;The local dataset shows Samsung E&amp;amp;A is up +51.8% over 60D and down -10.2% over 20D. Target upside is +39.5%, but 10D foreign + real-money flow is -KRW 49.5bn.&lt;/p&gt;
&lt;p&gt;The theme is good, but the money is not yet back.&lt;/p&gt;
&lt;p&gt;View: &lt;strong&gt;Wait&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;It needs to reclaim KRW 48,000 and show foreign-selling exhaustion. Energy EPC headlines alone may not be enough without orders and flow.&lt;/p&gt;
&lt;h2 id="11-gs-ec-daewoo-ec-and-hanmiglobal"&gt;11. GS E&amp;amp;C, Daewoo E&amp;amp;C and HanmiGlobal
&lt;/h2&gt;&lt;p&gt;These are secondary candidates, not core positions for this framework.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;GS E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Housing beta plus reconstruction option. Upside looks large, but housing and PF discounts remain.&lt;/td&gt;
 &lt;td&gt;Trading candidate&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Czech nuclear, Team Korea and reconstruction beta. Less direct to U.S. nuclear than Hyundai E&amp;amp;C.&lt;/td&gt;
 &lt;td&gt;Event trading&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HanmiGlobal&lt;/td&gt;
 &lt;td&gt;Data center and reconstruction PM optionality, but liquidity and flow are weaker.&lt;/td&gt;
 &lt;td&gt;Watch&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Daewoo E&amp;amp;C is often mentioned in nuclear baskets, but if the core question is U.S. nuclear, AI power and infrastructure EPC, it is less direct than Hyundai E&amp;amp;C. HanmiGlobal has an interesting PM angle, but liquidity and flow make it a watchlist name rather than a core position.&lt;/p&gt;
&lt;h2 id="12-entry-strategy"&gt;12. Entry strategy
&lt;/h2&gt;&lt;p&gt;This is a pullback and flow-confirmation market, not a blanket buy market.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Company&lt;/th&gt;
 &lt;th&gt;Action&lt;/th&gt;
 &lt;th&gt;Entry&lt;/th&gt;
 &lt;th&gt;Invalidation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Hyundai E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Pullback Buy&lt;/td&gt;
 &lt;td&gt;Support around KRW 150,000-155,000, or KRW 165,000 breakout with continued foreign/real-money flow&lt;/td&gt;
 &lt;td&gt;Fermi/Holtec EPC delay, or U.S. strategic investment first project not going toward nuclear&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;DL E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Conditional Buy&lt;/td&gt;
 &lt;td&gt;KRW 72,000-75,000 support plus sustained foreign/real-money flow&lt;/td&gt;
 &lt;td&gt;Housing cost risk returns, or overseas orders fail to convert into numbers&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Samsung C&amp;amp;T&lt;/td&gt;
 &lt;td&gt;Hold / Pullback only&lt;/td&gt;
 &lt;td&gt;Recovery after a 20D moving-average break, or official data-center development catalyst&lt;/td&gt;
 &lt;td&gt;Samsung Electronics stake-value pass-through weakens, or data center/SMR options fail to become contracts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;Samsung E&amp;amp;A&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;Reclaim KRW 48,000 plus foreign selling slows&lt;/td&gt;
 &lt;td&gt;Oil price shock, Middle East order delays, order-margin deterioration&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;GS E&amp;amp;C/Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Trading&lt;/td&gt;
 &lt;td&gt;News flow plus flow confirmation&lt;/td&gt;
 &lt;td&gt;Housing risk returns, weak absorption of retail flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="13-what-proves-or-kills-the-thesis"&gt;13. What proves or kills the thesis?
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Proof point&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;FEED converts to EPC&lt;/td&gt;
 &lt;td&gt;Most important proof for direct names like Hyundai E&amp;amp;C&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Backlog quality improves&lt;/td&gt;
 &lt;td&gt;Low-margin contracting versus high-value infrastructure EPC&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Data-center development becomes official&lt;/td&gt;
 &lt;td&gt;Key for Samsung C&amp;amp;T, HanmiGlobal and power-infra chains&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign + real-money flow persists&lt;/td&gt;
 &lt;td&gt;Confirms institutional sponsorship beyond theme trading&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Housing losses stabilize&lt;/td&gt;
 &lt;td&gt;Prevents old discounts from overpowering new rerating&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Risk&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Only MOUs, no EPC&lt;/td&gt;
 &lt;td&gt;Policy headlines do not become revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Fixed-price contracts and cost inflation&lt;/td&gt;
 &lt;td&gt;Orders hurt margin instead of helping earnings&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PF and housing losses return&lt;/td&gt;
 &lt;td&gt;Old construction discount overwhelms infrastructure story&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Reconstruction financing is delayed&lt;/td&gt;
 &lt;td&gt;Ukraine/Middle East optionality stays as news flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Foreign and real-money flow exits&lt;/td&gt;
 &lt;td&gt;Theme exists, but money does not follow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final view
&lt;/h2&gt;&lt;p&gt;A construction rerating is possible. But the market is not buying the entire construction sector.&lt;/p&gt;
&lt;p&gt;The question is how each company is reclassified.&lt;/p&gt;
&lt;p&gt;Hyundai E&amp;amp;C can be reclassified as U.S. nuclear, SMR and AI power infrastructure EPC. It is the most direct name, but the entry should be disciplined.&lt;/p&gt;
&lt;p&gt;DL E&amp;amp;C has the best asymmetry. It is less direct than Hyundai E&amp;amp;C, but its valuation and target gap are more comfortable.&lt;/p&gt;
&lt;p&gt;Samsung C&amp;amp;T is the highest-quality platform, but much of the story has already been priced.&lt;/p&gt;
&lt;p&gt;Samsung E&amp;amp;A has a good Middle East and energy EPC narrative, but flow needs to turn first.&lt;/p&gt;
&lt;p&gt;Fund-manager summary:&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;The leader is Hyundai E&amp;amp;C, the asymmetry is DL E&amp;amp;C, the quality platform is Samsung C&amp;amp;T, and the event option is Samsung E&amp;amp;A. Do not buy construction broadly. Buy only names that can be explained as AI power, nuclear, energy or reconstruction EPC.&lt;/strong&gt;&lt;/p&gt;
&lt;h2 id="evidence-and-limits"&gt;Evidence and Limits
&lt;/h2&gt;&lt;p&gt;&lt;strong&gt;Evidence&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Korea-U.S. strategic investment law and decree: LawTimes, MOTIR&lt;/li&gt;
&lt;li&gt;U.S. nuclear policy: U.S. DOE&lt;/li&gt;
&lt;li&gt;Hyundai E&amp;amp;C Fermi America FEED: Hyundai E&amp;amp;C newsroom&lt;/li&gt;
&lt;li&gt;Data center electricity demand: IEA Energy and AI&lt;/li&gt;
&lt;li&gt;Ukraine reconstruction needs: World Bank RDNA5&lt;/li&gt;
&lt;li&gt;Price, flow and target-price data: Thesis OS local DB, June 12, 2026 close&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;strong&gt;Limits&lt;/strong&gt;&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;As of June 14, 2026, the first concrete project under Korea’s strategic investment law has not been finalized.&lt;/li&gt;
&lt;li&gt;Fermi FEED is important, but final EPC contract and margin structure remain unconfirmed.&lt;/li&gt;
&lt;li&gt;Ukraine and Middle East reconstruction are large TAMs, but financing, insurance, sanctions and contract structure remain uncertain.&lt;/li&gt;
&lt;li&gt;Stock prices and consensus estimates can change after the June 12 close used in the dataset.&lt;/li&gt;
&lt;/ul&gt;</description></item><item><title>A Large Target-Price Gap Is Not Enough: KB Financial and Hana Financial Screen Best</title><link>https://koreainvestinsights.com/post/korea-target-price-gap-comfortable-entry-candidates-2026-06-13/</link><pubDate>Sat, 13 Jun 2026 23:50:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/korea-target-price-gap-comfortable-entry-candidates-2026-06-13/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This note follows &lt;a class="link" href="https://koreainvestinsights.com/post/samnik-crowding-q3-quality-watchlist-eps-up-price-lag-2026-06-11/" &gt;quality names to revisit after Samsung/SK crowding&lt;/a&gt; and the &lt;a class="link" href="https://koreainvestinsights.com/post/real-money-flow-framework-korea-institution-quality-2026-06-03/" &gt;Real Money flow framework&lt;/a&gt;. Related reads: &lt;a class="link" href="https://koreainvestinsights.com/post/korea-market-liquidity-foreign-reallocation-adr-narrow-leadership-2026-06-03/" &gt;Korea market liquidity and narrow leadership&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-etf-flow-led-market-volatility-strategy-2026-06-13/" &gt;ETF flow is leading the Korean market&lt;/a&gt;, and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;A large target-price gap is not the same as a comfortable entry. It is only a list of names that deserve a second look.&lt;/li&gt;
&lt;li&gt;The screen uses Korea Top 100 stocks as of the 2026-06-12 close, Naver consensus target-price history from 2026-04-12 to 2026-06-12, foreign plus real-money flow, and 20/50-day trend position.&lt;/li&gt;
&lt;li&gt;The only comfortable names in this specific framework are &lt;strong&gt;KB Financial&lt;/strong&gt; and &lt;strong&gt;Hana Financial&lt;/strong&gt;. They have remaining upside to latest consensus targets, positive target revisions, live flow, and price support above the 20-day and 50-day moving averages.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Samsung Biologics&lt;/strong&gt; is a watchlist name. The target gap is large and some foreign demand is visible, but target direction and trend are not yet supportive.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Hanwha Aerospace, Samyang Foods, and Alteogen&lt;/strong&gt; may still be good companies, but they are not comfortable entries now. Their target gaps are large, while current price trend and flow are weak.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core Point&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 A target-price gap is not proof that a stock is cheap. The cleaner setup requires &lt;strong&gt;gap, upward target revision, smart-money support, and 20/50-day trend confirmation&lt;/strong&gt; at the same time.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-method"&gt;1. Method
&lt;/h2&gt;&lt;p&gt;The question is simple: among large Korean stocks with meaningful target-price upside, which names can be held with less entry stress?&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Input&lt;/th&gt;
 &lt;th&gt;Definition&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Universe&lt;/td&gt;
 &lt;td&gt;Korea market-cap Top 100&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price date&lt;/td&gt;
 &lt;td&gt;2026-06-12 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Target window&lt;/td&gt;
 &lt;td&gt;2026-04-12 to 2026-06-12&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Target data&lt;/td&gt;
 &lt;td&gt;Naver consensus target-price time series&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flow&lt;/td&gt;
 &lt;td&gt;Foreign plus real-money flow over 5/10/20 trading days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Trend&lt;/td&gt;
 &lt;td&gt;20-day and 50-day moving average position&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The target data is a consensus time series, not broker-by-broker original target reports. This makes it useful for screening, but it cannot by itself prove a thesis.&lt;/p&gt;
&lt;h2 id="2-why-the-raw-gap-misleads"&gt;2. Why The Raw Gap Misleads
&lt;/h2&gt;&lt;p&gt;Several large-cap names show big target-price gaps: Alteogen, Hanwha Aerospace, Samsung Biologics and Samyang Foods all have latest-target upside above 60%.&lt;/p&gt;
&lt;p&gt;But a large gap can mean three different things:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Reason&lt;/th&gt;
 &lt;th&gt;Interpretation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;The stock sold off sharply&lt;/td&gt;
 &lt;td&gt;Could be opportunity, could be thesis damage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;The target price is slow to fall&lt;/td&gt;
 &lt;td&gt;Potential value trap&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Targets are rising while the stock lags&lt;/td&gt;
 &lt;td&gt;Best candidate type&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;That is why the second filter matters.&lt;/p&gt;
&lt;h2 id="3-the-six-name-readout"&gt;3. The Six-Name Readout
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Latest target gap&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2M target change&lt;/th&gt;
 &lt;th style="text-align: right"&gt;10D smart money&lt;/th&gt;
 &lt;th style="text-align: right"&gt;20D price&lt;/th&gt;
 &lt;th&gt;Trend&lt;/th&gt;
 &lt;th&gt;Read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;KB Financial&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+26.3%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+3.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 138.9bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.1%&lt;/td&gt;
 &lt;td&gt;Above 20D/50D&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Comfortable&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Hana Financial&lt;/strong&gt;&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+29.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+6.5%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 44.4bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.1%&lt;/td&gt;
 &lt;td&gt;Above 20D/50D&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Possible, weaker than KB&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Biologics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+60.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+KRW 53.4bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-8.2%&lt;/td&gt;
 &lt;td&gt;Below 20D/50D&lt;/td&gt;
 &lt;td&gt;Watch&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Hanwha Aerospace&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+69.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+13.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 16.6bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-16.2%&lt;/td&gt;
 &lt;td&gt;Below 20D/50D&lt;/td&gt;
 &lt;td&gt;Wait for proof&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Samyang Foods&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+62.9%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+2.8%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 4.7bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-11.6%&lt;/td&gt;
 &lt;td&gt;Below 20D/50D&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alteogen&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+68.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-1.4%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 16.8bn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-4.8%&lt;/td&gt;
 &lt;td&gt;Below 20D/50D&lt;/td&gt;
 &lt;td&gt;Not a comfort entry&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The message is clear: financials have smaller headline gaps, but the full setup is cleaner. Growth, biotech and defense names have bigger gaps, but the market is not yet confirming them.&lt;/p&gt;
&lt;h2 id="4-candidate-notes"&gt;4. Candidate Notes
&lt;/h2&gt;&lt;h3 id="kb-financial"&gt;KB Financial
&lt;/h3&gt;&lt;p&gt;KB Financial has the cleanest setup. Latest target upside is +26.3%, the target has risen +3.8% over two months, 10-day smart money is positive, and the stock is above both the 20-day and 50-day averages.&lt;/p&gt;
&lt;p&gt;This is not the largest upside in the screen. It is the best combination of credible upside, flow and trend.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actionability:&lt;/strong&gt; top comfort-entry candidate.&lt;br&gt;
&lt;strong&gt;First rejection:&lt;/strong&gt; loss of 20-day support and foreign selling.&lt;br&gt;
&lt;strong&gt;What would make it investable:&lt;/strong&gt; support in the 155,000 to 162,000 won area with continued foreign and institutional buying.&lt;/p&gt;
&lt;h3 id="hana-financial"&gt;Hana Financial
&lt;/h3&gt;&lt;p&gt;Hana is the next candidate. It has a larger target gap than KB and a stronger two-month target revision, but the price trend is less convincing. It is above the 20-day and 50-day averages, yet 20-day performance is slightly negative.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actionability:&lt;/strong&gt; possible, but below KB.&lt;br&gt;
&lt;strong&gt;First rejection:&lt;/strong&gt; failure around 118,000 to 121,000 won and weaker real-money flow.&lt;br&gt;
&lt;strong&gt;What would make it investable:&lt;/strong&gt; stabilization above 121,000 won and better relative strength within financials.&lt;/p&gt;
&lt;h3 id="samsung-biologics"&gt;Samsung Biologics
&lt;/h3&gt;&lt;p&gt;Samsung Biologics has a large +60.9% latest-target gap, but the target has fallen over two months and the stock is below both key moving averages. It is a watchlist name, not a comfort entry.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actionability:&lt;/strong&gt; watchlist.&lt;br&gt;
&lt;strong&gt;First rejection:&lt;/strong&gt; further target-price cuts or failure to reclaim the 20-day average.&lt;br&gt;
&lt;strong&gt;What would make it investable:&lt;/strong&gt; target cuts stop, the stock regains the 20-day average, and foreign buying persists.&lt;/p&gt;
&lt;h3 id="hanwha-aerospace"&gt;Hanwha Aerospace
&lt;/h3&gt;&lt;p&gt;Hanwha Aerospace has one of the strongest target-revision profiles: latest-target gap +69.9%, two-month target change +13.1%. But smart money is negative and the stock is down sharply over 20 days.&lt;/p&gt;
&lt;p&gt;This is a good company with a strong structural defense story, but not yet a comfortable entry.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actionability:&lt;/strong&gt; wait for proof.&lt;br&gt;
&lt;strong&gt;First rejection:&lt;/strong&gt; continued smart-money selling and failure to recover the 20-day average.&lt;br&gt;
&lt;strong&gt;What would make it investable:&lt;/strong&gt; renewed foreign and institutional buying plus evidence of order or earnings upgrades.&lt;/p&gt;
&lt;h3 id="samyang-foods"&gt;Samyang Foods
&lt;/h3&gt;&lt;p&gt;Samyang remains a high-quality export growth story, but price and flow have not turned. The target gap is large, but smart money is negative and the stock is below the 20-day and 50-day averages.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actionability:&lt;/strong&gt; wait.&lt;br&gt;
&lt;strong&gt;What would make it investable:&lt;/strong&gt; reclaim the 20-day average, real-money inflow returns, and export/margin expectations hold.&lt;/p&gt;
&lt;h3 id="alteogen"&gt;Alteogen
&lt;/h3&gt;&lt;p&gt;Alteogen has a large target gap, but target direction, smart-money flow and trend are all weak. It can still be an event-driven biotech trade, but it does not belong in a comfort-entry basket.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Actionability:&lt;/strong&gt; pass for this screen.&lt;br&gt;
&lt;strong&gt;What would make it investable:&lt;/strong&gt; a clear clinical or licensing catalyst, renewed flow, and recovery above the 20-day average.&lt;/p&gt;
&lt;h2 id="final-basket"&gt;Final Basket
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Bucket&lt;/th&gt;
 &lt;th&gt;Stocks&lt;/th&gt;
 &lt;th&gt;Read&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Comfortable&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;KB Financial, Hana Financial&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Target revision, flow and trend align&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Watch&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Samsung Biologics&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Big gap, but target direction and trend are weak&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Hanwha Aerospace, Samyang Foods&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Good companies, but entry still needs proof&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Excluded for comfort entry&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Alteogen&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Large gap, weak confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The conclusion is intentionally conservative. The best current setup is not the stock with the largest target gap. It is the stock where the target gap, revision direction, flow and chart agree.&lt;/p&gt;
&lt;h2 id="evidence-ledger"&gt;Evidence Ledger
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Basis&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Price date&lt;/td&gt;
 &lt;td&gt;2026-06-12 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Target data&lt;/td&gt;
 &lt;td&gt;Naver consensus target-price time series, 2026-04-12 to 2026-06-12&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Universe&lt;/td&gt;
 &lt;td&gt;Korea market-cap Top 100&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Target coverage&lt;/td&gt;
 &lt;td&gt;95 covered names&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Flow&lt;/td&gt;
 &lt;td&gt;Foreign plus real-money flow over recent 5/10/20 trading days&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Caveat&lt;/td&gt;
 &lt;td&gt;Consensus time-series screen, not original broker-by-broker target reports&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;</description></item><item><title>Hyundai E&amp;C and Woojin: Where Is the Asymmetric Bargaining Power in U.S. Nuclear Expansion?</title><link>https://koreainvestinsights.com/post/hyundai-ec-woojin-us-nuclear-expansion-bargaining-power-2026-06-13/</link><pubDate>Sat, 13 Jun 2026 23:10:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/hyundai-ec-woojin-us-nuclear-expansion-bargaining-power-2026-06-13/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/us-investment-special-act-team-korea-nuclear-opportunity-2026-06-13/" &gt;Korea&amp;rsquo;s $350bn U.S. investment law: what is the nuclear opportunity for Team Korea?&lt;/a&gt;. Related reads: &lt;a class="link" href="https://koreainvestinsights.com/post/korea-etf-flow-led-market-volatility-strategy-2026-06-13/" &gt;ETF flow is leading the Korean market&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/hd-hyundai-heavy-industries-smr-terrapower-natrium-option-2026-05-27/" &gt;HD Hyundai Heavy Industries&amp;rsquo; SMR option&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/us-nonsemi-rerating-ai-power-software-korea-translation-2026-05-31/" &gt;U.S. non-semiconductor rerating into Korea&lt;/a&gt;, and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis Hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;U.S. nuclear momentum is real. DOE has reiterated the goal of expanding U.S. nuclear capacity from about 100GW today toward 400GW by 2050, while the White House has pushed NRC reform and faster reactor decisions.&lt;/li&gt;
&lt;li&gt;Do not buy &amp;ldquo;Korea nuclear&amp;rdquo; as one undifferentiated basket. For U.S. nuclear expansion, business directness looks like &lt;strong&gt;Hyundai E&amp;amp;C &amp;gt; Doosan Enerbility &amp;gt; Woojin &amp;gt; Daewoo E&amp;amp;C&lt;/strong&gt;. But stock attractiveness depends on price, flow, and bargaining power.&lt;/li&gt;
&lt;li&gt;Hyundai E&amp;amp;C is the most direct Korean equity here. It is tied to Fermi&amp;rsquo;s four-unit AP1000 FEED path and Holtec&amp;rsquo;s SMR-300/Palisades/10GW fleet narrative. Its bargaining power is not reactor IP. It is the U.S. shortage of proven nuclear EPC and project-management execution.&lt;/li&gt;
&lt;li&gt;Woojin is smaller and more asymmetric. Its base case is Korea-native nuclear instrumentation, ICI replacement demand, and RSPT/RTD references. Direct U.S. AP1000 adoption is not publicly confirmed. Treat that as an option, not the base case.&lt;/li&gt;
&lt;li&gt;Entry discipline: &lt;strong&gt;Hyundai E&amp;amp;C is Pullback Buy&lt;/strong&gt;, &lt;strong&gt;Woojin is small Conditional Buy&lt;/strong&gt;, &lt;strong&gt;Doosan Enerbility is Wait&lt;/strong&gt;, and &lt;strong&gt;Daewoo E&amp;amp;C is Avoid for this specific thesis&lt;/strong&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core Point&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Hyundai E&amp;C sits on the &lt;strong&gt;execution bottleneck&lt;/strong&gt;. Woojin sits on the &lt;strong&gt;instrumentation bottleneck&lt;/strong&gt;. Both have bargaining power, but Hyundai needs contract-scope proof and Woojin needs evidence of U.S. AP1000 or SMR adoption.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-why-this-follow-up-matters"&gt;1. Why This Follow-Up Matters
&lt;/h2&gt;&lt;p&gt;The previous note framed Korea&amp;rsquo;s $350bn U.S. investment plan as policy capital, not an automatic order book for Korean companies. That distinction matters even more for nuclear.&lt;/p&gt;
&lt;p&gt;The structure is closer to:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Layer&lt;/th&gt;
 &lt;th&gt;Meaning&lt;/th&gt;
 &lt;th&gt;Investment implication&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S.-selected strategic projects&lt;/td&gt;
 &lt;td&gt;Project selection is largely U.S.-driven&lt;/td&gt;
 &lt;td&gt;Korean vendors are not automatically guaranteed&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean policy capital&lt;/td&gt;
 &lt;td&gt;Financing, commercial reasonableness and risk sharing matter&lt;/td&gt;
 &lt;td&gt;Capital alone does not create listed-company revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Korean vendor participation&lt;/td&gt;
 &lt;td&gt;Best-effort participation&lt;/td&gt;
 &lt;td&gt;Alpha depends on procurement, EPC, O&amp;amp;M and instrumentation language&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;$200bn strategic-investment track&lt;/td&gt;
 &lt;td&gt;Principal/interest recovery and profit sharing matter&lt;/td&gt;
 &lt;td&gt;Project economics matter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;$150bn shipbuilding track&lt;/td&gt;
 &lt;td&gt;Separate from nuclear&lt;/td&gt;
 &lt;td&gt;Do not mix shipbuilding finance with nuclear revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The nuclear value chain therefore needs to be broken into nodes:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Value-chain node&lt;/th&gt;
 &lt;th&gt;Korean bargaining power&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reactor design IP&lt;/td&gt;
 &lt;td&gt;Low&lt;/td&gt;
 &lt;td&gt;AP1000 is Westinghouse, SMR-300 is Holtec, Natrium is TerraPower&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Generic EPC/labor&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;td&gt;U.S. local labor and regulation matter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Nuclear EPC/PMO execution&lt;/td&gt;
 &lt;td&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;Few firms have large nuclear build experience&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Primary equipment and heavy components&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;td&gt;Supply-chain bottlenecks and QA matter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Instrumentation and I&amp;amp;C&lt;/td&gt;
 &lt;td&gt;High, but reactor-specific&lt;/td&gt;
 &lt;td&gt;Certification and installed references matter&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;O&amp;amp;M and maintenance&lt;/td&gt;
 &lt;td&gt;Medium-high&lt;/td&gt;
 &lt;td&gt;Nuclear creates long recurring service demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Hyundai E&amp;amp;C and Woojin sit in different nodes. Hyundai E&amp;amp;C is the EPC/PMO execution name. Woojin is the instrumentation and recurring-replacement option.&lt;/p&gt;
&lt;h2 id="2-us-nuclear-momentum-is-real"&gt;2. U.S. Nuclear Momentum Is Real
&lt;/h2&gt;&lt;p&gt;DOE&amp;rsquo;s May 2026 update described U.S. nuclear momentum across Gen III+ SMRs, TerraPower Natrium, Palisades restart, uprates and AI-related power demand. The policy objective is to expand U.S. nuclear capacity from roughly &lt;strong&gt;100GW to 400GW by 2050&lt;/strong&gt;. (&lt;a class="link" href="https://www.energy.gov/ne/articles/one-year-after-executive-orders-us-nuclear-energy-renaissance-full-swing" target="_blank" rel="noopener"
 &gt;U.S. DOE&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The White House NRC reform order points to faster review clocks, including final decisions on new reactor construction and operation applications within no more than &lt;strong&gt;18 months&lt;/strong&gt;. (&lt;a class="link" href="https://www.whitehouse.gov/presidential-actions/2025/05/ordering-the-reform-of-the-nuclear-regulatory-commission/" target="_blank" rel="noopener"
 &gt;White House&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Project evidence is also visible:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Project&lt;/th&gt;
 &lt;th&gt;Confirmed datapoint&lt;/th&gt;
 &lt;th&gt;Korea read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fermi Project Matador&lt;/td&gt;
 &lt;td&gt;AP1000 Units 1-4 COLA submitted; NRC review underway&lt;/td&gt;
 &lt;td&gt;Direct Hyundai E&amp;amp;C FEED channel&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Holtec Palisades SMR-300&lt;/td&gt;
 &lt;td&gt;Two SMR-300 units targeted for 2030 commercial operation&lt;/td&gt;
 &lt;td&gt;Hyundai&amp;rsquo;s SMR-300 construction channel&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Holtec 10GW SMR fleet&lt;/td&gt;
 &lt;td&gt;Holtec and Hyundai E&amp;amp;C target a 10GW North American fleet&lt;/td&gt;
 &lt;td&gt;Potential repeat-build platform&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;TerraPower Natrium&lt;/td&gt;
 &lt;td&gt;NRC construction permit and construction start&lt;/td&gt;
 &lt;td&gt;Longer-term equipment-chain option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NuScale and X-energy&lt;/td&gt;
 &lt;td&gt;SMR design and supply-chain competition continues&lt;/td&gt;
 &lt;td&gt;Doosan and heavy-equipment chain&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Fermi America describes Project Matador as a large energy campus for AI, data centers and advanced manufacturing, with four AP1000 units and a strategic Hyundai E&amp;amp;C partnership. (&lt;a class="link" href="https://fermiamerica.com/fermi-america-deepens-strategic-partnership-with-hyundai-ec-to-lead-the-return-of-large-scale-nuclear-construction-in-the-u-s-with-project-matador/" target="_blank" rel="noopener"
 &gt;Fermi America&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The NRC page confirms that Fermi submitted a combined license application for AP1000 Units 1-4, that two of three parts have been submitted, and that NRC review is underway. (&lt;a class="link" href="https://www.nrc.gov/reactors/new-reactors/large-lwr/col/fermi-energy-intel-campus" target="_blank" rel="noopener"
 &gt;NRC&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;World Nuclear News reported Holtec&amp;rsquo;s Palisades SMR-300 plan and the expanded construction agreement with Hyundai E&amp;amp;C, with a 10GW North American SMR-300 fleet target. (&lt;a class="link" href="https://world-nuclear-news.org/articles/holtec-and-hyundai-ec-target-10gw-fleet-of-smrs-in-us" target="_blank" rel="noopener"
 &gt;World Nuclear News&lt;/a&gt;)&lt;/p&gt;
&lt;h2 id="3-hyundai-ec-the-most-direct-korean-execution-name"&gt;3. Hyundai E&amp;amp;C: The Most Direct Korean Execution Name
&lt;/h2&gt;&lt;p&gt;Hyundai E&amp;amp;C is not a reactor-IP owner. It does not own AP1000, SMR-300 or Natrium. Its value is different: it is one of the few Korean companies with credible nuclear construction, site execution and large-project PMO credentials.&lt;/p&gt;
&lt;h3 id="fermi-ap1000"&gt;Fermi AP1000
&lt;/h3&gt;&lt;p&gt;Fermi gives Hyundai E&amp;amp;C the cleanest direct U.S. nuclear exposure. The company is tied to FEED work for four AP1000 units in Texas. That does not yet equal a full EPC revenue contract, but it is a real channel.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Hyundai E&amp;amp;C position&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Reactor IP&lt;/td&gt;
 &lt;td&gt;No&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;AP1000 selection&lt;/td&gt;
 &lt;td&gt;Westinghouse-centered&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FEED and project design&lt;/td&gt;
 &lt;td&gt;Direct channel&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;EPC, BOP and PMO&lt;/td&gt;
 &lt;td&gt;Upside option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Local labor and regulation&lt;/td&gt;
 &lt;td&gt;Must be shared with U.S. partners&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Revenue certainty&lt;/td&gt;
 &lt;td&gt;Needs post-FEED EPC confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h3 id="holtec-smr-300"&gt;Holtec SMR-300
&lt;/h3&gt;&lt;p&gt;Holtec may be the more asymmetric channel. Palisades is the first visible project, but the 10GW fleet ambition matters more.&lt;/p&gt;
&lt;p&gt;Simple math:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Calculation&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Result&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Target fleet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;10GW&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SMR-300 unit&lt;/td&gt;
 &lt;td style="text-align: right"&gt;300MW&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Implied units&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 33.3&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Palisades two units&lt;/td&gt;
 &lt;td style="text-align: right"&gt;600MW&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Palisades share of fleet&lt;/td&gt;
 &lt;td style="text-align: right"&gt;About 6.0%&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;If Palisades works, Hyundai E&amp;amp;C is not just a one-project beneficiary. It could become a repeat-build execution partner.&lt;/p&gt;
&lt;h3 id="where-hyundais-bargaining-power-comes-from"&gt;Where Hyundai&amp;rsquo;s Bargaining Power Comes From
&lt;/h3&gt;&lt;p&gt;U.S. nuclear expansion is not only a technology problem. It is also a build problem. The market needs firms that can manage nuclear QA, civil construction, schedule discipline, BOP integration, documentation and local supply-chain complexity.&lt;/p&gt;
&lt;p&gt;Hyundai E&amp;amp;C&amp;rsquo;s experience with Barakah, Shin Hanul and large global infrastructure does not guarantee U.S. economics. But it gives the company a credible seat at the table.&lt;/p&gt;
&lt;p&gt;The main risk is contract structure:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Contract type&lt;/th&gt;
 &lt;th&gt;Read-through&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Fixed-price EPC&lt;/td&gt;
 &lt;td&gt;Large headline order, but cost-overrun risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Cost-plus&lt;/td&gt;
 &lt;td&gt;Lower margin, lower loss risk&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Target-price plus incentive&lt;/td&gt;
 &lt;td&gt;Execution skill can become profit&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Alliance / JV / IPD&lt;/td&gt;
 &lt;td&gt;Best for repeat fleet delivery&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;FEED only&lt;/td&gt;
 &lt;td&gt;Direct but limited revenue&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For Hyundai, the key evidence is FEED-to-EPC conversion, scope, risk sharing and whether Holtec fleet economics become repeatable.&lt;/p&gt;
&lt;h2 id="4-woojin-the-smaller-instrumentation-option"&gt;4. Woojin: The Smaller Instrumentation Option
&lt;/h2&gt;&lt;p&gt;Woojin is a very different setup. It is not an EPC company. It is a nuclear instrumentation and sensor company.&lt;/p&gt;
&lt;p&gt;The base case is:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Korea-native nuclear instrumentation references&lt;/li&gt;
&lt;li&gt;ICI recurring replacement demand&lt;/li&gt;
&lt;li&gt;RSPT, RTD and other safety-related measurement products&lt;/li&gt;
&lt;li&gt;U.S. SMR/AP1000 adoption as an unconfirmed option&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;Woojin&amp;rsquo;s 1Q26 numbers improved sharply. Edaily Marketin reported 1Q26 revenue of &lt;strong&gt;KRW 37.2bn&lt;/strong&gt;, up &lt;strong&gt;36.9% YoY&lt;/strong&gt;, and operating profit of &lt;strong&gt;KRW 5.0bn&lt;/strong&gt;, turning positive. Operating margin was about &lt;strong&gt;13.4%&lt;/strong&gt;. (&lt;a class="link" href="https://marketin.edaily.co.kr/News/ReadE?newsId=05366086645449248" target="_blank" rel="noopener"
 &gt;Edaily Marketin&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Bloter reported that Woojin&amp;rsquo;s nuclear-equipment revenue rose to &lt;strong&gt;KRW 12.2bn&lt;/strong&gt;, up &lt;strong&gt;60.8% YoY&lt;/strong&gt;, while nuclear operating profit rose to &lt;strong&gt;KRW 3.0bn&lt;/strong&gt; from KRW 0.3bn. The temperature-sensor and instrumentation segment also rose &lt;strong&gt;43.5% YoY&lt;/strong&gt;. (&lt;a class="link" href="https://v.daum.net/v/U9LZp0716x?f=p" target="_blank" rel="noopener"
 &gt;Bloter via Daum&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;Woojin&amp;rsquo;s product map matters:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Product&lt;/th&gt;
 &lt;th&gt;Role&lt;/th&gt;
 &lt;th&gt;Investment angle&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;ICI Assembly&lt;/td&gt;
 &lt;td&gt;Measures neutron flux and core power distribution&lt;/td&gt;
 &lt;td&gt;3-4.5 year replacement cycle&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;RSPT&lt;/td&gt;
 &lt;td&gt;Measures control rod position&lt;/td&gt;
 &lt;td&gt;Shin Hanul 3/4 contract reference&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;HJTC&lt;/td&gt;
 &lt;td&gt;Reactor water-level thermocouple&lt;/td&gt;
 &lt;td&gt;Long replacement cycle&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;PI RTD / fast-response RTD&lt;/td&gt;
 &lt;td&gt;Coolant temperature sensing&lt;/td&gt;
 &lt;td&gt;Safety/control instrumentation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Industrial temperature sensors&lt;/td&gt;
 &lt;td&gt;Non-nuclear sensor growth&lt;/td&gt;
 &lt;td&gt;Secondary earnings support&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Woojin signed a &lt;strong&gt;KRW 6.4bn&lt;/strong&gt; RSPT contract with Doosan Enerbility for Shin Hanul 3/4. (&lt;a class="link" href="https://marketin.edaily.co.kr/News/ReadE?newsId=02788006639119504" target="_blank" rel="noopener"
 &gt;Edaily Marketin&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;It also disclosed a &lt;strong&gt;KRW 8.27bn&lt;/strong&gt; ICI contract for standard nuclear-plant overhaul demand, equal to about &lt;strong&gt;5.88%&lt;/strong&gt; of recent revenue. (&lt;a class="link" href="https://v.daum.net/v/20250627140734831" target="_blank" rel="noopener"
 &gt;Hankyung via Daum&lt;/a&gt;)&lt;/p&gt;
&lt;p&gt;The important caveat: there is no public confirmation that Woojin is directly adopted in U.S. AP1000 projects. Westinghouse has its own qualified supply chain, and U.S. nuclear instrumentation has high QA, environmental and seismic qualification barriers. Woojin&amp;rsquo;s Korean APR1400 references do not automatically transfer to AP1000.&lt;/p&gt;
&lt;p&gt;So the correct framing is:&lt;/p&gt;

 &lt;blockquote&gt;
 &lt;p&gt;Woojin is a Korea-native nuclear instrumentation moat with a U.S. optionality layer, not a confirmed AP1000 supplier.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="5-hyundai-ec-vs-woojin"&gt;5. Hyundai E&amp;amp;C vs Woojin
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Hyundai E&amp;amp;C&lt;/th&gt;
 &lt;th&gt;Woojin&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Core exposure&lt;/td&gt;
 &lt;td&gt;U.S. nuclear EPC/PMO execution&lt;/td&gt;
 &lt;td&gt;Nuclear instrumentation and ICI replacement&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Directness&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Asymmetry&lt;/td&gt;
 &lt;td&gt;Medium&lt;/td&gt;
 &lt;td&gt;High&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Confirmed U.S. channel&lt;/td&gt;
 &lt;td&gt;Fermi AP1000 FEED, Holtec SMR-300&lt;/td&gt;
 &lt;td&gt;No direct AP1000 confirmation&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Bargaining power&lt;/td&gt;
 &lt;td&gt;Build experience, schedule, QA, PMO&lt;/td&gt;
 &lt;td&gt;Instrument references, certification, replacement demand&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Key risk&lt;/td&gt;
 &lt;td&gt;Contract structure and cost overrun&lt;/td&gt;
 &lt;td&gt;AP1000 adoption unverified, contract size small&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Good news&lt;/td&gt;
 &lt;td&gt;FEED-to-EPC, Holtec fleet scope&lt;/td&gt;
 &lt;td&gt;AP1000/SMR order, ICI/RSPT follow-on&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;Hyundai E&amp;amp;C has more direct business exposure. Woojin has more small-cap convexity.&lt;/p&gt;
&lt;h2 id="6-why-doosan-and-daewoo-are-lower-priority-here"&gt;6. Why Doosan and Daewoo Are Lower Priority Here
&lt;/h2&gt;&lt;p&gt;Doosan Enerbility is strategically important. It sits in large equipment, NuScale, TerraPower and X-energy supply-chain conversations. But as a stock, the optionality is already large, valuation is demanding, and recent foreign flow is still weak. The correct posture is &lt;strong&gt;Wait&lt;/strong&gt;, not chase.&lt;/p&gt;
&lt;p&gt;Daewoo E&amp;amp;C has Czech nuclear and Team Korea relevance, but its direct U.S. nuclear exposure is weaker than Hyundai E&amp;amp;C&amp;rsquo;s. It may trade as nuclear beta, but it is not the clean expression of this U.S. thesis.&lt;/p&gt;
&lt;h2 id="7-entry-screen"&gt;7. Entry Screen
&lt;/h2&gt;&lt;p&gt;Internal price and flow data as of the June 12, 2026 close:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Exposure&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Price / flow&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Hyundai E&amp;amp;C&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Fermi AP1000 FEED, Holtec SMR-300&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 157,500, 5D +19.4%, 1M foreign +KRW 209.7bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Priority. Pullback Buy&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Woojin&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;SMR/instrumentation option, ICI/RSPT&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 18,070, 1M -22.1%, RSI 31.5&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Small Conditional Buy&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Doosan Enerbility&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;NuScale, TerraPower, X-energy equipment&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 93,100, 1M foreign -KRW 591.4bn&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Wait&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;&lt;strong&gt;Daewoo E&amp;amp;C&lt;/strong&gt;&lt;/td&gt;
 &lt;td&gt;Czech / Team Korea, less U.S.-direct&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 21,850, 1M -24.5%&lt;/td&gt;
 &lt;td&gt;&lt;strong&gt;Avoid for this thesis&lt;/strong&gt;&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Name&lt;/th&gt;
 &lt;th&gt;Action&lt;/th&gt;
 &lt;th&gt;Entry&lt;/th&gt;
 &lt;th&gt;Invalidation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Hyundai E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Pullback Buy&lt;/td&gt;
 &lt;td&gt;Support near KRW 150,000 or breakout above KRW 160,000 with foreign re-buying&lt;/td&gt;
 &lt;td&gt;Fermi/Holtec EPC delay, foreign buying fades&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Woojin&lt;/td&gt;
 &lt;td&gt;Small conditional buy&lt;/td&gt;
 &lt;td&gt;KRW 17,500-18,200 support plus institutional/foreign buying&lt;/td&gt;
 &lt;td&gt;Break below KRW 17,000, no follow-on SMR/instrument orders&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Doosan Enerbility&lt;/td&gt;
 &lt;td&gt;Wait&lt;/td&gt;
 &lt;td&gt;KRW 96,000 recovery and foreign selling slowdown&lt;/td&gt;
 &lt;td&gt;Continued foreign selling, no FY26 earnings upgrade&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Daewoo E&amp;amp;C&lt;/td&gt;
 &lt;td&gt;Avoid&lt;/td&gt;
 &lt;td&gt;Needs direct U.S. nuclear contract&lt;/td&gt;
 &lt;td&gt;Retail absorption fails, nuclear momentum fades&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;If choosing one name, it is Hyundai E&amp;amp;C. It is the cleanest Korean equity for U.S. nuclear development exposure. But after a 5-day 19.4% move, buying the pullback is better than chasing.&lt;/p&gt;
&lt;p&gt;If taking a small asymmetric position, Woojin is the candidate. The price has cooled, but the U.S. AP1000 angle is unverified. Treat it as a sub-1% pilot option, not a core position.&lt;/p&gt;
&lt;p&gt;Doosan is the right strategic company but currently a rich option. Daewoo is not the clean U.S. nuclear expression.&lt;/p&gt;
&lt;h2 id="evidence-ledger"&gt;Evidence Ledger
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Item&lt;/th&gt;
 &lt;th&gt;Source&lt;/th&gt;
 &lt;th&gt;Confirmation&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;U.S. nuclear expansion&lt;/td&gt;
 &lt;td&gt;U.S. DOE&lt;/td&gt;
 &lt;td&gt;Capacity goals, SMRs, TerraPower, Palisades, AI-power link&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;NRC reform&lt;/td&gt;
 &lt;td&gt;White House&lt;/td&gt;
 &lt;td&gt;Faster reactor decision timelines&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Fermi AP1000 Units 1-4&lt;/td&gt;
 &lt;td&gt;Fermi America, NRC&lt;/td&gt;
 &lt;td&gt;Project Matador, AP1000 COLA and NRC review&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Holtec SMR-300&lt;/td&gt;
 &lt;td&gt;World Nuclear News, Holtec&lt;/td&gt;
 &lt;td&gt;Palisades, 2030 target, 10GW fleet, DOE support&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Woojin 1Q26&lt;/td&gt;
 &lt;td&gt;Edaily, Bloter&lt;/td&gt;
 &lt;td&gt;Revenue, operating profit and nuclear segment improvement&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Woojin orders&lt;/td&gt;
 &lt;td&gt;Edaily, Hankyung/Daum, Newsis&lt;/td&gt;
 &lt;td&gt;Shin Hanul RSPT, ICI overhaul contracts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price and flow&lt;/td&gt;
 &lt;td&gt;Internal price/flow DB&lt;/td&gt;
 &lt;td&gt;June 12, 2026 close&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;</description></item><item><title>Korean Semiconductor Top 50: Which Names Can Beat Samsung and SK Hynix Over the Next Two Months?</title><link>https://koreainvestinsights.com/post/korea-semiconductor-top50-two-month-alpha-candidates-samsung-hynix-2026-06-13/</link><pubDate>Sat, 13 Jun 2026 19:30:00 +0900</pubDate><guid>https://koreainvestinsights.com/post/korea-semiconductor-top50-two-month-alpha-candidates-samsung-hynix-2026-06-13/</guid><description>
 &lt;blockquote&gt;
 &lt;p&gt;Context
This is a follow-up to &lt;a class="link" href="https://koreainvestinsights.com/post/korea-semiconductor-etf-exposure-marketcap-gap-strategy-2026-06-13/" &gt;Samsung and SK Hynix Are 90.8% of Korean Semis&lt;/a&gt;. It also connects to &lt;a class="link" href="https://koreainvestinsights.com/post/kr-theme-etf-rebalance-flow-semicap-cap-trim-2026-06-12/" &gt;Korea theme ETF rebalance flow&lt;/a&gt;, &lt;a class="link" href="https://koreainvestinsights.com/post/korea-etf-flow-led-market-volatility-strategy-2026-06-13/" &gt;ETF flow is leading the Korean market&lt;/a&gt;, the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-semiconductor-hbm-kospi-hub/" &gt;HBM hub&lt;/a&gt;, the &lt;a class="link" href="https://koreainvestinsights.com/page/korea-ai-pcb-substrate-hub/" &gt;AI PCB and substrate hub&lt;/a&gt;, and the &lt;a class="link" href="https://koreainvestinsights.com/page/exclusive-analysis-hub/" &gt;Exclusive Analysis hub&lt;/a&gt;.&lt;/p&gt;

 &lt;/blockquote&gt;
&lt;h2 id="tldr"&gt;TL;DR
&lt;/h2&gt;&lt;ul&gt;
&lt;li&gt;Samsung Electronics and SK Hynix remain the engine of Korean semiconductors. But for &lt;strong&gt;relative alpha over the next two months&lt;/strong&gt;, the better candidates may be second-line equipment, substrate, inspection and AI storage names where ETF ownership is still lower and flow is just beginning to build.&lt;/li&gt;
&lt;li&gt;A scan of 1,137 Naver ETFs, Korean semiconductor Top 50 constituents, and local price/flow data highlights &lt;strong&gt;Korea Circuit, HPSP, TES, VM and FADU&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Korea Circuit&lt;/strong&gt; is the cleanest balance of low ETF exposure, institutional accumulation and substrate catalysts. &lt;strong&gt;HPSP&lt;/strong&gt; has the best flow quality. &lt;strong&gt;TES and VM&lt;/strong&gt; have strong fundamental triggers but have already moved very fast. &lt;strong&gt;FADU&lt;/strong&gt; is the high-risk convex option.&lt;/li&gt;
&lt;/ul&gt;
&lt;div class="thesis-callout"&gt;
 &lt;div class="thesis-callout__label"&gt;Core Point&lt;/div&gt;
 &lt;div class="thesis-callout__body"&gt;
 Samsung and SK Hynix are the market engine. But the next two months of relative alpha are more likely to come from &lt;strong&gt;under-owned second-line equipment, substrate, inspection and AI storage names&lt;/strong&gt; than from the megacap memory names everyone already owns.
 &lt;/div&gt;
&lt;/div&gt;
&lt;h2 id="1-method"&gt;1. Method
&lt;/h2&gt;&lt;p&gt;This is not a generic semiconductor watchlist. It is a screen for Korean semiconductor Top 50 stocks that could outperform Samsung Electronics and SK Hynix over a two-month horizon.&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Input&lt;/th&gt;
 &lt;th&gt;Coverage&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Universe&lt;/td&gt;
 &lt;td&gt;Korean semiconductor Top 50&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ETF data&lt;/td&gt;
 &lt;td&gt;1,137 Naver ETFs and 75,001 constituent rows&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Price and flow&lt;/td&gt;
 &lt;td&gt;Thesis OS local database, as of June 12, 2026&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;Windows&lt;/td&gt;
 &lt;td&gt;2-week, 1-month and 3-month returns plus foreign/institutional flow&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;ETF exposure&lt;/td&gt;
 &lt;td&gt;Estimated ownership from ETF NAV and constituent weights&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;The ETF exposure figure is a proxy, not official PCF data. It is useful for comparing ownership pressure, but it does not directly measure creation/redemption flow.&lt;/p&gt;
&lt;h2 id="2-baseline-samsung-and-sk-hynix"&gt;2. Baseline: Samsung and SK Hynix
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th style="text-align: right"&gt;2W&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1M&lt;/th&gt;
 &lt;th style="text-align: right"&gt;3M&lt;/th&gt;
 &lt;th style="text-align: right"&gt;Estimated ETF exposure&lt;/th&gt;
 &lt;th style="text-align: right"&gt;ETF count&lt;/th&gt;
 &lt;th style="text-align: right"&gt;1M foreign + institution&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td&gt;Samsung Electronics&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+7.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+15.6%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+69.7%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 51.5tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;229&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 15.2tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td&gt;SK Hynix&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-6.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+17.2%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;+125.1%&lt;/td&gt;
 &lt;td style="text-align: right"&gt;KRW 52.1tn&lt;/td&gt;
 &lt;td style="text-align: right"&gt;214&lt;/td&gt;
 &lt;td style="text-align: right"&gt;-KRW 15.3tn&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;These two stocks remain core exposure. But they already sit inside more than 200 ETFs each, with estimated ETF ownership above KRW 50tn. That makes them the market beta. The alpha question is different: where can incremental flow have more price impact?&lt;/p&gt;
&lt;h2 id="3-compressed-candidate-list"&gt;3. Compressed Candidate List
&lt;/h2&gt;&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;Segment&lt;/th&gt;
 &lt;th&gt;Main evidence&lt;/th&gt;
 &lt;th&gt;View&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Korea Circuit&lt;/td&gt;
 &lt;td&gt;Substrate&lt;/td&gt;
 &lt;td&gt;ETF exposure only KRW 85.6bn, 1M foreign+institution +KRW 66.6bn, distributed institutional buying&lt;/td&gt;
 &lt;td&gt;Best balance&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;HPSP&lt;/td&gt;
 &lt;td&gt;Equipment&lt;/td&gt;
 &lt;td&gt;1M foreign+institution +KRW 197.2bn and 2W +KRW 196.5bn&lt;/td&gt;
 &lt;td&gt;Best flow quality&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;TES&lt;/td&gt;
 &lt;td&gt;Equipment&lt;/td&gt;
 &lt;td&gt;1M return +67.8%, SK Hynix equipment contract&lt;/td&gt;
 &lt;td&gt;Strong but hot&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;VM&lt;/td&gt;
 &lt;td&gt;Equipment&lt;/td&gt;
 &lt;td&gt;Low ETF exposure, strong 2026E growth, institutional accumulation&lt;/td&gt;
 &lt;td&gt;Under-owned equipment beta&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;td&gt;Fabless / AI storage&lt;/td&gt;
 &lt;td&gt;2W pullback, heavy foreign buying, SSD controller contracts&lt;/td&gt;
 &lt;td&gt;High-risk convexity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="4-stock-notes"&gt;4. Stock Notes
&lt;/h2&gt;&lt;h3 id="korea-circuit"&gt;Korea Circuit
&lt;/h3&gt;&lt;p&gt;Korea Circuit is the most balanced candidate. It is not yet crowded in ETF ownership, but the price and institutional flow have already started to move. The thesis is not just &amp;ldquo;substrates.&amp;rdquo; It is &lt;strong&gt;SOCAMM plus FC-BGA plus Broadcom AI switch exposure&lt;/strong&gt;. Prime Economy reported that Meritz expected FC-BGA utilization to rise from 55% in 2025 to 70% in 2026 and 85% in 2027, and forecast SOCAMM revenue growth of 301% in 2027. &lt;a class="link" href="https://www.newsprime.co.kr/news/article.html?no=727904" target="_blank" rel="noopener"
 &gt;Prime Economy&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The risk is that the move has already begun: 1M return +36.9% and 3M return +152.6%. It is a watch / buy-on-pullback candidate.&lt;/p&gt;
&lt;h3 id="hpsp"&gt;HPSP
&lt;/h3&gt;&lt;p&gt;HPSP has the cleanest flow. Foreign investors and institutions are both buying, and the 2-week flow almost matches the 1-month flow, which means the accumulation has not faded. The business quality is also clear: 2026E OPM is 53.4%.&lt;/p&gt;
&lt;p&gt;The issue is price. The stock is up 40.5% over two weeks and trades at 50.9x 2026E PER. It needs a pullback or a stronger 2Q confirmation before aggressive entry.&lt;/p&gt;
&lt;h3 id="tes"&gt;TES
&lt;/h3&gt;&lt;p&gt;TES has the clearest near-term catalyst. DataTooza reported a KRW 21.2bn semiconductor equipment contract with SK Hynix, equal to 6.04% of recent sales, and noted that announced supply contracts in 2026 year-to-date were up 81.1% year over year. &lt;a class="link" href="https://www.datatooza.com/article/20260605111420887552ef3d2537_80" target="_blank" rel="noopener"
 &gt;DataTooza&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;This is why TES screens well on fundamentals. But the stock is already up 67.8% in one month. It is a buy-on-pullback candidate, not a chase candidate.&lt;/p&gt;
&lt;h3 id="vm"&gt;VM
&lt;/h3&gt;&lt;p&gt;VM is an institutional accumulation story rather than a foreign-flow story. Foreign investors have sold, but institutions have bought aggressively. The 2026E setup is attractive: revenue +118.6%, operating profit +281.0%, OPM 29.8%, and PER around 30.0x.&lt;/p&gt;
&lt;p&gt;Chosun Biz reported that Daol Investment &amp;amp; Securities viewed VM as undervalued among Korean semiconductor equipment names and set a KRW 140,000 target price. &lt;a class="link" href="https://biz.chosun.com/stock/stock_general/2026/06/12/TCCNX36LGVEW7AASGFWR4WEVVE/" target="_blank" rel="noopener"
 &gt;Chosun Biz&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The risk is short-term overheating. It needs a pullback and evidence that foreign selling is fading.&lt;/p&gt;
&lt;h3 id="fadu"&gt;FADU
&lt;/h3&gt;&lt;p&gt;FADU is different from the others. It is not a substrate or equipment name. It is an AI storage and SSD controller option.&lt;/p&gt;
&lt;p&gt;The catalyst density is high. The Bell reported that FADU signed a KRW 50bn enterprise SSD controller contract with an overseas NAND maker, equal to 54.1% of the prior year&amp;rsquo;s revenue, and that announced orders had exceeded KRW 216.3bn after the contract. &lt;a class="link" href="https://m.thebell.co.kr/m/newsview.asp?newskey=202605070926576760107185" target="_blank" rel="noopener"
 &gt;The Bell&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;The flow is powerful but not clean: foreign investors bought KRW 460.4bn over one month while institutions sold KRW 296.2bn. FADU is therefore an event option, not a core position.&lt;/p&gt;
&lt;h2 id="5-ranking"&gt;5. Ranking
&lt;/h2&gt;&lt;p&gt;For two-month relative alpha:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;Korea Circuit&lt;/td&gt;
 &lt;td&gt;Best mix of ETF under-ownership, institutional flow and substrate catalysts&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;HPSP&lt;/td&gt;
 &lt;td&gt;Best flow quality, but needs a better entry&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;td&gt;Pullback plus foreign accumulation, high-risk convexity&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;TES&lt;/td&gt;
 &lt;td&gt;Strongest beta, but too hot near term&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;VM&lt;/td&gt;
 &lt;td&gt;Institutional accumulation and low ETF exposure&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;p&gt;For fundamental catalyst strength:&lt;/p&gt;
&lt;table&gt;
 &lt;thead&gt;
 &lt;tr&gt;
 &lt;th style="text-align: right"&gt;Rank&lt;/th&gt;
 &lt;th&gt;Stock&lt;/th&gt;
 &lt;th&gt;Why&lt;/th&gt;
 &lt;/tr&gt;
 &lt;/thead&gt;
 &lt;tbody&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;1&lt;/td&gt;
 &lt;td&gt;TES&lt;/td&gt;
 &lt;td&gt;Contract evidence and SK Hynix equipment linkage&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;2&lt;/td&gt;
 &lt;td&gt;VM&lt;/td&gt;
 &lt;td&gt;Large 2026E earnings step-up and new customer/process option&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;3&lt;/td&gt;
 &lt;td&gt;Korea Circuit&lt;/td&gt;
 &lt;td&gt;SOCAMM + FC-BGA dual catalyst&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;4&lt;/td&gt;
 &lt;td&gt;HPSP&lt;/td&gt;
 &lt;td&gt;High quality but needs fresh earnings proof at a high multiple&lt;/td&gt;
 &lt;/tr&gt;
 &lt;tr&gt;
 &lt;td style="text-align: right"&gt;5&lt;/td&gt;
 &lt;td&gt;FADU&lt;/td&gt;
 &lt;td&gt;Strong contracts, but high valuation and institutional selling&lt;/td&gt;
 &lt;/tr&gt;
 &lt;/tbody&gt;
&lt;/table&gt;
&lt;h2 id="final-view"&gt;Final View
&lt;/h2&gt;&lt;p&gt;Samsung Electronics and SK Hynix remain the center of Korean semiconductors. But if the question is which names can beat them over the next two months, the answer likely sits in second-line names where ETF ownership is still lower and flow is improving.&lt;/p&gt;
&lt;p&gt;The cleanest single name is &lt;strong&gt;Korea Circuit&lt;/strong&gt;. The highest-quality flow is &lt;strong&gt;HPSP&lt;/strong&gt;. The strongest fundamental catalysts are &lt;strong&gt;TES and VM&lt;/strong&gt;. The most convex option is &lt;strong&gt;FADU&lt;/strong&gt;.&lt;/p&gt;
&lt;p&gt;The common rule is simple: &lt;strong&gt;do not chase vertical moves; wait for pullbacks, check flow persistence, and demand evidence that the theme is becoming revenue.&lt;/strong&gt;&lt;/p&gt;</description></item></channel></rss>