US-China Summit on Korea Invest Insights

The May 14-15 Beijing US-China Summit — Managed Truce, H200 Refusal, and What It Means for Korean Semiconductors

Sun, 17 May 2026 23:50:00 +0900

📚 US-China Summit Series Pre-Summit Guide: 10 Agendas for Korean Investors / Summit Result: No Grand Bargain / CloudMatrix Expansion Scenario / NVIDIA Earnings and Korea’s Second-Order Supply Chain / AI-RAN and the Korean Supply Chain

The May 14-15 Beijing summit was not a grand bargain. It was a managed truce. Washington and Beijing agreed to avoid immediate escalation, but the core issues — semiconductor export controls, rare earths, Taiwan, and tariff structure — were not solved. The more important signal came outside the room. The U.S. cleared H200 exports, but China has not let purchases proceed. Huawei’s CloudMatrix 384, meanwhile, shows a different Chinese response: weaker chips offset by more chips, more power, and more optical links. Decoupling is not ending. It is moving into a more expensive equilibrium.

Key Takeaways

The summit was a managed truce. It reduced near-term escalation risk but did not resolve the technology-control structure.
The H200 episode matters more than the communique. The U.S. has cleared sales to roughly 10 Chinese firms, but reports indicate no H200 chips have shipped because China has not approved purchases.
CloudMatrix 384 is a system-level workaround. Huawei is not beating NVIDIA at chip efficiency. It is using many more chips, much more power, and a larger optical fabric to reach useful system-level performance.
ITIF’s model complicates the “decoupling is bad for Korea” narrative. In a full semiconductor decoupling scenario, ITIF estimates U.S. firms could lose USD 77B in China-related sales in the first year, while South Korean firms could gain roughly USD 21B.
Korea’s exposure is time-dependent. The short term favors HBM, memory, substrates, MLCC and optical links. The medium term is mixed. The long term brings standard fragmentation, Chinese substitution, and dual-supply-chain costs.
The cleanest second-order alpha is optical interconnect. The U.S. stack needs more CPO and rack-scale optics; the Chinese stack needs even more optical links because it compensates with scale.

1. What the Summit Actually Signaled

The summit should be read as a pause, not a settlement. There were preliminary understandings, purchase intentions, and stability language. But there was no binding package that changes the core technology conflict.

That distinction matters. Tariffs can be negotiated. Semiconductor controls are strategic. If export controls remain structurally intact, then the AI infrastructure map remains split into two partially separate systems.

The political incentives explain the outcome. The U.S. wanted market stability and visible business wins. China wanted time and did not want to look as if it was accepting U.S. terms. Both sides wanted to avoid a visible breakdown; neither side had a strong reason to make a structural concession.

Xi Jinping’s reference to the Thucydides Trap should also be read in that frame. On the surface it is a call to avoid conflict. Underneath it sits a time-horizon argument. If the U.S. restricts exports, China accelerates domestic substitution. If U.S. firms lose China revenue, their R&D capacity can weaken. The conflict is not disappearing; the cost distribution is changing.

2. H200 Refusal Shows the Weakening of the Export-License Carrot

The H200 episode is the clearest signal. The U.S. created a conditional framework for H200 exports to China in January 2026, with oversight and a fee structure. By mid-May, the U.S. had reportedly cleared around 10 Chinese firms to buy H200s, including major internet platforms.

But no chips have shipped. Reporting indicates China has not allowed purchases to proceed. Trump reportedly said China wants to develop its own chips.

This is not just an administrative delay. It means export permission is no longer automatically a carrot. If Beijing refuses the product for strategic reasons, U.S. permission has no commercial value.

For NVIDIA, China H200 revenue should be treated as an option, not as the base case. That makes the next NVIDIA earnings call important. If management confirms that China data-center compute revenue is immaterial or excluded from guidance, the market will need to lower the China reopening option value.

For Korea, the meaning is more nuanced. If Chinese customers do not buy H200s, the immediate HBM linkage is weaker. But if non-China hyperscalers continue to expand AI capex, Korean HBM demand remains strong. The H200 story is less about a direct shipment boost and more about the shape of decoupling.

3. CloudMatrix 384: Losing at the Chip Level, Fighting at the System Level

Huawei’s CloudMatrix 384 shows how China is responding. According to public reports based on SemiAnalysis work, the system uses 384 Ascend chips and can reach roughly 300 PFLOPs of dense BF16 compute, compared with about 180 PFLOPs for NVIDIA’s GB200 NVL72.

But the comparison is not a clean victory. CloudMatrix reportedly consumes roughly 559kW, compared with about 145kW for GB200 NVL72. It is a brute-force design: more chips, more racks, more power, and more optical transceivers.

This tells us two things.

First, China has not achieved chip-level parity. NVIDIA’s efficiency, software ecosystem, CUDA moat, and training stack remain powerful.

Second, China may not need parity for every use case. For domestic policy markets, state cloud, sovereign AI, and “good enough” inference, a less efficient but domestically controlled system can be acceptable.

The most interesting investment implication is optical interconnect. A brute-force cluster needs a much larger communication fabric. CloudMatrix reportedly relies on thousands of high-speed optical transceivers. If weaker chips require more scale-out and scale-up connections, then optical components become a shared bottleneck.

The U.S. stack also moves in the same direction. Rubin, NVL systems, Spectrum-X, and CPO all increase rack-scale optical demand. The U.S. uses optics to improve efficiency; China uses optics to compensate for chip gaps. Different reasons, same direction.

4. ITIF’s Decoupling Model: Korea Is a Short-Term Absorber

ITIF’s November 2025 report gives a useful quantitative frame. In a hypothetical full semiconductor decoupling scenario, U.S. semiconductor firms could lose about USD 77B of China-related sales in the initial year. ITIF estimates that South Korean firms could gain about USD 21B of those lost sales, more than the EU, Taiwan, Japan, mainland China, or other regions.

The memory line is especially important. In ITIF’s table, a large share of the U.S. memory-related loss is absorbed by South Korea. This is why “decoupling is bad for Korea” is too simple.

In the short run, Korea can benefit. The country is not the direct AI accelerator rival to NVIDIA. It supplies memory, HBM, substrates, MLCC, testing, packaging, and other bottleneck components. If U.S. firms lose parts of China while non-China AI capex rises, Korean component suppliers can still benefit.

But the long-run view is less comfortable. The longer decoupling persists, the more China substitutes. Standards fragment. Korean firms may need U.S.-aligned and China-aligned product paths, compliance structures, inventory systems, and qualification cycles. That raises cost and can pressure margins.

The right answer is time-segmented:

6-12 months: favorable for Korean memory and AI components.
1-3 years: mixed, as China substitution begins to pressure pricing and standards.
3-5 years: structurally riskier, because dual supply chains and fragmented standards raise costs.

5. What It Means for Korean Stocks

The first-order beneficiaries remain SK hynix and Samsung Electronics. SK hynix has the cleanest HBM exposure. Samsung Electronics has HBM4, DDR5, eSSD, foundry optionality, and a broader AI-infrastructure platform story. But both are already heavily discovered. For new money, chasing after a large move is less attractive than buying on macro or earnings pullbacks.

The cleaner second-order alpha may sit in the less crowded supply chain.

Optical interconnect is the most interesting. OE Solutions and related optical names should not be framed only as NVIDIA suppliers. They are exposure to a world where both the U.S. and Chinese AI stacks need more high-speed connections.

Samsung Electro-Mechanics is another example. AI servers need MLCCs and FC-BGA substrates regardless of whether the rack is U.S.-standard or China-standard. The company is not the GPU winner, but it is a component bottleneck.

Hana Micron belongs in the structural-improvement bucket. If its Vina pricing structure and overseas margins continue to improve, it is not simply a memory-cycle beta. It is a post-processing margin-upgrade story.

6. Four Market Misreadings

The first misreading is that decoupling is a single big-bang event. It is not. It has been moving layer by layer since the Huawei restrictions: telecom equipment, semiconductor tools, AI chips, AI systems, and now optical links, standards, and software stacks.

The second misreading is that decoupling is uniformly bad for Korea. It is bad over a long horizon if standards fragment and China substitutes. But ITIF’s model shows that in the short run South Korean firms can absorb a meaningful share of U.S. firms’ lost sales.

The third misreading is that China’s rejection of H200 proves self-sufficiency. It does not. China still lags at the chip level. What it shows is a willingness to accept inefficiency in exchange for domestic control.

The fourth misreading is that the summit means decoupling is easing. The summit reduced escalation risk; it did not dismantle the strategic conflict.

7. The Next Catalyst: September and October

The next major window is a potential Xi visit in September and the October anniversary of the prior truce framework. The base case is truce extension: both sides want stability. A tougher scenario is a renewed escalation around export controls, rare earths, or Taiwan language. A true grand bargain remains lower probability.

For investors, the playbook is simple. If the truce is extended, existing Korean semi exposure can remain. If escalation returns, near-term drawdowns can create better entries in memory and AI component names, while optical-interconnect names may become more strategically relevant. If a grand bargain triggers a sharp relief rally, that may be a chance to trim overheated first-order winners.

8. Bottom Line

The Beijing summit did not end the U.S.-China technology conflict. It confirmed that the conflict is being managed.

The H200 episode and CloudMatrix 384 are the real signals. The U.S. can approve exports, but China can refuse them. China may not have the best chip, but it can build a system that is good enough for parts of its domestic market. That is a more expensive equilibrium, not a peaceful one.

For Korea, the answer is not simply bullish or bearish. Decoupling is positive today, mixed tomorrow, and risky later. SK hynix and Samsung Electronics remain the core short-term beneficiaries, but the more interesting medium-term angle may be optical interconnect, AI passive components, substrates, and structural OSAT improvement.

The alpha is not in saying “decoupling is good” or “decoupling is bad.” The alpha is in separating the time horizon.

Sources and Evidence Notes

Fact

ITIF estimates that in a full decoupling scenario U.S. semiconductor firms could lose about USD 77B of China-related sales in the first year, while South Korean firms could gain about USD 21B.
Reuters-cited reporting and Tom’s Hardware coverage indicate the U.S. cleared H200 sales to roughly 10 Chinese firms but no chips had shipped as of mid-May.
Tom’s Hardware and SemiAnalysis-based reporting describe Huawei CloudMatrix 384 as a 384-chip system with roughly 300 PFLOPs BF16 and about 559kW system power.

Inference

H200 non-shipment is interpreted here as a weakening of the export-license carrot.
CloudMatrix 384 is interpreted as a system-level workaround rather than a chip-level victory.
Korea’s exposure is segmented into short-term benefit, medium-term mixed impact, and long-term fragmentation risk.

Main references: ITIF Decoupling Risks report, Reuters H200 licensing report via Investing.com, Tom’s Hardware on China blocking H200 purchases, Tom’s Hardware on CloudMatrix 384, SemiAnalysis CloudMatrix note.

This post is for research and commentary only and is not investment advice. Summit details, H200 shipment status, and CloudMatrix specifications are based on public reporting as of May 17, 2026 KST. ITIF’s numbers are model outputs and may not match actual decoupling outcomes. Scenario analysis and stock implications are the author’s interpretation and may be wrong.

US-China Summit Result — No 'Grand Bargain.' Hormuz Agreement, H200 Licenses, $30B Tariff-Easing Talks Did Land — but KOSPI 7,900 Already Priced It

Fri, 15 May 2026 16:00:00 +0900

📚 US-China summit series. Previous: Pre-Analysis — 10 Agendas and Scenario Strategy Follow-up: The May 14-15 Beijing US-China Summit — Managed Truce, H200 Refusal, and What It Means for Korean Semiconductors

🔗 Related: KOSPI May 13 V-Reversal — All-Time High Through ₩3.76tn Foreign Selling · Samsung Electronics Citi TP ₩460,000 — Memory-Cycle Frame Reset · Why Korea Part 4 — $6.7B Inflows, Korea-Discount Dissolving or Value Trap?

The summit concluded. The pre-analysis said “the most-likely outcome is a ‘soybeans + Boeing + trade-board’-level symbolic agreement” and “this is already in the price.” Both came true. Hormuz reopening, H200 licenses, and $30B tariff discussions did land, but there was no joint communiqué, no rare-earth pause extension, no semiconductor-control easing. And KOSPI is already +19% in May — semiconductors +39%. The question now is not “should I buy more?” but “which sectors haven’t moved yet?”

TL;DR

Outcome. 135-minute summit. Hormuz reopening + Iran-no-nuclear agreement, NVIDIA H200 China-sales licenses (10 firms, up to 75K units each), $30B non-sensitive tariff-easing discussions. But no joint communiqué, no rare-earth-pause extension, no semiconductor-equipment-control easing.
Vs. pre-analysis. A blend of “Scenario A (expectations met, 50%)” + “Scenario D (Hormuz easing)” + “Scenario E (FX-stability language).” “Scenario B (tech-for-rare-earths grand bargain)” did not materialize.
Market reaction. KOSPI 7,912 (+0.87%); China onshore down 1.5–2.5% (expectations exhausted). Good news arrived, but was already in the price.
Core read. Chasing is inefficient. KOSPI May +19%, semis +39% is extreme concentration. Only 2 of 26 sectors beat KOSPI (semis + autos) — first time since 2005. The asymmetric structure favors finding the least-priced-in beneficiary.

1. How did the pre-analysis perform — scenario validation

1.1 Four pre-analysis scenarios

Comparing the pre-analysis scenarios against actual outcome:

Scenario	Pre-prob	Actual	Match
A. Expectations met (soybeans / Boeing / committee + Iran signal)	50%	✅ closest	Match
B. Tech-for-rare-earths grand bargain	20%	❌ did not happen	Miss
C. Downside (Taiwan tension + control tightening)	25%	△ Taiwan warning but no escalation	Partial
D. Hormuz easing	15–20%	✅ Hormuz reopening agreement	Match
E. FX-stability comment	25–35%	△ direct language unconfirmed	Partial

The pre-analysis core message — “betting before the event is gambling; positioning after is investing” — was validated. Chinese equities actually fell on summit day.

1.2 What was confirmed — exactly what landed

Outcome	Detail	Nature
Hormuz reopening agreement	Both sides agreed on keeping the strait open and Iran no-nuclear	Most substantive outcome
H200 sales licenses	10 Chinese firms (Alibaba, Tencent, ByteDance, etc.) cleared to purchase NVIDIA H200	License granted; zero actual shipments
$30B tariff-easing discussions	Non-sensitive items focus. Items not yet defined	Discussion stage, not agreement
‘Strategic stability’ framing	Both sides agree on constructive relations	Phrase only, non-binding
Xi’s Taiwan warning	“Conflict possible” warning; US says “no policy change”	Tension preserved, not escalated

1.3 What did NOT land — this matters more

Unresolved	Why it matters
Joint communiqué	No binding agreement means no enforcement mechanism
Rare-earth pause extension	November time bomb intact
Semiconductor equipment export-control easing	Korean memory’s “geopolitical scarcity premium” preserved
Specific tariff-reduction items	$30B remains directionally undefined
Boeing / agriculture purchase scale	Risk of Trump election-cycle inflation

2. Why a “small deal” — structural reasons

2.1 Neither side can concede the core

What Trump wants: visible election-cycle wins (soybeans, beef, Boeing)
What Xi wants: semiconductor-control easing, Taiwan security

Neither can give the other their core ask:
→ US can't ease semi/AI controls (national security)
→ China can't concede Taiwan (regime legitimacy)

Result: partial face-saving trades + big picture deferred to follow-up sessions

2.2 What the absence of a joint communiqué means

No communiqué means no binding agreement was created. Hormuz, H200, tariffs — all administrative-action level. Reversible based on political conditions.

The pre-analysis warned that “2017 Trump China visit also produced large MOUs but many were non-binding.” Same pattern repeats here.

2.3 CSIS assessment

Scott Kennedy at CSIS: “a largely superficial truce that tilts in China’s favor." US preserved security cards, but on economic gestures, the direction moved toward what China wanted.

3. Sector-by-sector impact on Korea

3.1 Semiconductors — good news, but already priced in

The H200 licensing logic:
→ NVIDIA H200 can now be sold to Chinese firms
→ H200 contains HBM
→ HBM is SK hynix's market — 64% share, #1
→ Therefore additional demand for Korean memory

But:
→ Semis already up +39% in May
→ Only 2 of 26 KOSPI sectors beat KOSPI (first since 2005)
→ Foreign net sell on May 14: ₩705.2bn

Key issues:
→ "License ≠ actual shipment." Zero H200 actually delivered
→ Beijing restraining NVIDIA inflow to protect domestic AI chips
→ Real HBM-incremental-demand realization lags by quarters

Read: hold-the-semiconductor-weight makes sense; chasing is inefficient. The Samsung Citi piece framed that breaking the “memory cycle” framing requires both sustained price uplift and HBM4E customer qualifications. H200 licensing is one piece of that. Track actual H200 delivery, NVIDIA guidance, and HBM order-backlog change.

3.2 Rare earths — what didn’t happen matters more

The pre-analysis flagged the “November time bomb” of the rare-earth pause expiration. The expected extension was not announced. Two readings:

Read 1 (short-term negative):
→ Rare-earth theme-stocks' near-term catalyst is exhausted
→ The "summit will resolve it" expectation broke
→ Short-term profit-taking pressure

Read 2 (mid-term positive):
→ China continues holding rare earths as a negotiating card
→ Uncertainty persists through November expiration
→ Non-China supply-chain build-out necessity is reinforced
→ Firms with actual supply-chain build capability earn a premium

Read: short-term theme stocks (UnionMatel etc.) in profit-taking. Mid-term, non-China permanent-magnet / rare-earth supply chains remain structurally relevant.

3.3 Hormuz / energy — the most substantive outcome, the least-priced-in beneficiary

The Hormuz reopening is the summit’s most substantive deliverable. Signal that China will engage diplomatically on Iran pressure. The end-of-war scenario probability rose.

What Hormuz opening does for Korea:

1. Oil-price downward pressure
 → Energy-cost relief for Korea (high Middle-East crude dependency)

2. Naphtha price decline
 → NCC-based petrochemical cost burden relief
 → Margin recovery potential for Lotte Chemical, LG Chem

3. KRW strength pressure
 → Lower energy imports → current-account improvement → KRW stability
 → Better foreign flow → constructive for KOSPI as a whole

4. Aviation / chemical cost easing
 → Korean Air, Asiana fuel-cost reduction

Why “least-priced-in”: the market interpreted the summit through the “semiconductor beneficiary” lens first. But semis already moved +39%; petrochemicals are still near the cycle bottom. The Hormuz second-order effect (naphtha price decline → chemical-margin normalization) hasn’t been absorbed by the market yet.

3.4 Sector summary

Sector	Summit impact	Already priced?	Read
Semiconductors	H200 licensing → HBM demand expectation	Already +39%	Chasing inefficient; hold weight
Rare-earth themes	Pause extension absent → catalyst exhausted	Partial	Short-term profit-taking
Petrochemicals	Hormuz → naphtha cost relief	Not yet priced	Watch list
Autos	Neutral. Supply-chain stability +; China EV tariffs preserved	Already +29%	Chasing inefficient
Aviation / chemicals	Oil decline → cost relief	Partial	Watch
Defense	Hormuz end-of-war signal → short-term momentum weakens	Not priced	Short-term pullback → potential opportunity
Cosmetics / tourism	China opening expectation → indirect positive	Not priced	Watch

4. “KOSPI 7,900 already priced the good news” — asymmetric setup

4.1 Extreme concentration

Since start of May:
KOSPI overall: +18.9%
Semiconductors: +38.6%
Autos: +29.1%
Other 24 sectors: under-performed KOSPI

→ Only 2 of 26 sectors beat KOSPI
→ Extreme concentration not seen since 2005

What this means: KOSPI +19% is not “Korean economy strength” but “Samsung Electronics + SK hynix strength.” Most else didn’t move. Such concentration can persist, but historically it’s often a rotation precursor.

4.2 Foreign-flow warning

May 14 foreign net selling: ₩705.2bn. Foreign selling into a rising tape — is it “expectation-exhaustion selling” or “overall position-reduction”? Worth tracking. The May 13 V-reversal piece noted ₩3.76tn of foreign net selling absorbed by domestic flows — same pattern. The durability of “foreigners out, domestic capital catching” needs verification.

May 15 foreign flow is the key checkpoint. Continued selling → “catalyst-exhaustion correction” possible. Net-buying flip → trend preservation.

4.3 What China onshore showed

China onshore equities fell 1.5–2.5% on summit day. The classic “expectation rally → fade on outcome” pattern. Korea could see similar.

5. The least-priced-in spot — Hormuz second-order effect

5.1 Why petrochemicals are on the watch list

Logic chain:

Hormuz reopening agreement (confirmed)
→ End-of-war scenario probability rises
→ Oil-price downward pressure
→ Naphtha price decline
→ NCC-based petrochemical cost burden relief
→ Ethylene-naphtha spread normalization
→ Earnings leverage for trough-level chemicals

Conditions for this logic to work:
→ Dubai oil settles below $70
→ Naphtha-ethylene spread recovers above $250/tonne
→ Verifiable within 4–8 weeks

Why the market hasn’t priced it: the market absorbed the summit through the “semiconductor news” lens first. Hormuz’s chemical-margin effect was covered mainly in English-language press; Korean sell-side estimate revisions haven’t followed yet.

5.2 Cautions

The Hormuz agreement is declarative. Actual military blockade lifting is separate, and if Iran end-of-war fails, it can reverse immediately. Also, Chinese petrochemical oversupply is structural, capping spread recovery.

So the read is not “buy chemicals now” but “monitor Dubai oil and spreads over 4–8 weeks, conditional approach”.

6. Validation of the pre-analysis core messages

6.1 What was right

Pre-analysis message	Outcome
“Most-likely outcome is a ‘soybeans + Boeing + committee’ level”	✅ Was a small deal
“Betting before the event is gambling”	✅ China onshore fell on summit day
“KOSPI 7,500 already priced the good outcome”	✅ Additional upside limited
“Korea is not a counterparty but the maximum stakeholder”	✅ Semis / energy / rare earths all directly affect Korea
“November time bomb (rare earths + truce simultaneous expiry)”	✅ Rare-earth pause extension absent — risk intact

6.2 What needs revision

Pre-analysis message	Revision
“H200 China sales = summit outcome”	→ H200 licensing started Dec 2025. This summit was a “reaffirmation” rather than new licensing
Scenario D (Hormuz) probability 15–20%	→ Hormuz agreement actually landed. Probability was too low

7. Calendar — what to watch next

Window	Event	Why
5/15	Foreign flow check	Continued selling = catalyst-exhaustion correction; flip = trend preservation
5/15	Day-2 joint statement text	Additional agreement detail?
5/21–22	Pearl Abyss Meritz NDR	Institutional re-rating window (separate issue)
June	Dubai oil + naphtha spread trajectory	Hormuz agreement’s real-world effect
July	Samsung Electronics + SK hynix 2Q earnings	H200 incremental demand showing up in results?
August	Half-year reporting season	Hormuz effect appearing in chemical / energy earnings
November	Busan truce expiry + rare-earth pause expiry	The real time bomb. Extension or not = H2’s largest macro variable

8. Bottom line

The US-China summit produced a “managed truce,” not a “grand bargain.” Hormuz agreement, H200 licenses, $30B tariff discussions landed — but no joint communiqué, no rare-earth pause extension, no semiconductor-control easing. Closest to the pre-analysis Scenario A (expectations met).

The problem is KOSPI is already +19% with semis +39% and autos +29%. Good news arrived but is already in the price. Only 2 of 26 sectors beating KOSPI is the most extreme concentration since 2005.

This is not a chase setup. It’s a “find the least-priced-in beneficiary” setup. Semis have moved; rare-earth theme stocks have exhausted the near-term catalyst. The least-priced-in spot is the Hormuz second-order effect — naphtha-cost decline driving petrochemical margin normalization. And the real time bomb hasn’t gone off yet — the November simultaneous expiry of rare-earth pause + trade truce.

FAQ

Q: Why did Chinese onshore equities fall after a “successful” summit? A: The classic “rally on expectation, sell on outcome.” Pre-summit buying anticipated a “grand bargain”; the actual “small deal” outcome triggered profit-taking. Korea faces the same risk.

Q: Is H200 licensing materially positive for Korean memory? A: Limited near-term. License ≠ actual shipment. Beijing has reasons to restrain NVIDIA inflows in favor of domestic AI chips, and real HBM incremental demand lags by quarters. Plus semis are already +39% in May.

Q: Why isn’t the absence of rare-earth pause extension a bigger problem? A: November expiration is 6 months out — additional negotiation opportunities exist, and unilateral Chinese tightening would invite retaliation. But short-term theme stocks (UnionMatel etc.) have exhausted the summit-hope catalyst — profit-taking pressure.

Q: Why is petrochemicals “the least-priced-in spot”? A: The market absorbed the summit through the “semiconductor lens” first. The naphtha-cost-decline → chemical-margin normalization story is lagged-effect and Korean sell-side estimates haven’t been revised yet. But it’s conditional — requires Dubai oil settling below $70 and naphtha-ethylene spread recovering above $250/tonne over 4–8 weeks.

Q: What is the “November time bomb”? A: Busan trade truce (47% tariff regime) and rare-earth export-control pause both expire in November. Without resolution, tariffs reset to 57% and rare earths can be re-weaponized. The best-case May summit deliverable would have been a “roadmap to November” framework — that didn’t land.

Q: KOSPI +19%, semis +39% — buy more? A: Only 2 of 26 sectors beating KOSPI is extreme (first since 2005). Historically, this kind of concentration tends to be a rotation precursor. The disciplined read is to watch under-priced sectors (petrochemicals, aviation, parts of cosmetics / tourism) rather than chase.

Q: Foreigners sold ₩705bn on a rising KOSPI — why? A: Domestic flows (retail + institutional) absorbed it. Same pattern appeared May 13 (₩3.76tn foreign net selling absorbed by domestic flows). The durability of this “foreigners out, domestic capital catching” structure is the key variable — May 15 foreign-flow direction is the next checkpoint.

This article is for research and informational purposes only and does not constitute investment advice. Summit outcome details are based on Yonhap, Reuters, China MOFA, White House releases, CNBC, CNN reporting. H200 licensing (10 firms, up to 75K units each) is per Reuters reporting; zero actual shipments to date. $30B tariff easing is in discussion stage, not agreement. The Hormuz agreement is declarative; actual military-blockade lifting is separate. Day-2 may produce additional disclosures. KOSPI / sector returns reference Kiwoom Securities citations. Petrochemical margin recovery is conditional on oil and naphtha-spread evolution, not confirmed. Analysis can be wrong. Data cut: May 15, 2026 KST.

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.

US-China Summit (May 14-15, Beijing) — A Korean Investor's Guide to 10 Agendas and Scenario Strategy. The Real Question Is Not 'Should I Buy China' — It's 'Where Does Korea Sit'

Sun, 10 May 2026 23:30:00 +0900

🔗 Related reading: Why Korea Part 4 — $6.7B ETF Inflows + KOSPI Valuation Paradox · Korea AI PCB Ecosystem: 10 Companies · Korea Optical / CPO Value Chain — 7 Companies · KOSDAQ Series Part 1 · KOSDAQ Series Part 2

🔄 Post-event analysis (2026-05-15): Summit Result — Hormuz Agreement, H200 Licenses, $30B Tariff Talks Landed, but KOSPI 7,900 Already Priced It

Trump and Xi meet in Beijing on May 14-15. Markets read it as “tariff easing → China-stock bounce.” That framing misses the actual investment substance. Korea and Japan are not at the negotiating table but are the largest stakeholders in the result. If semiconductor export controls hold, Korea’s memory oligopoly tightens. If the rare-earth pause is extended, Korean battery and auto cost structures stabilize. If Hormuz reopens, energy import costs fall. Conversely, if controls loosen, China’s domestic memory self-sufficiency accelerates. If Taiwan tensions escalate, Korea’s geopolitical discount widens. The question is not “should I buy China.” It’s “where does Korea sit in this set of outcomes.”

TL;DR

May 14–15 Beijing summit. First US presidential visit to China since 2017 (8 years). Treasury Secretary Bessent visits Japan (FX, rare earths, energy) and Korea (FX, economic agenda) before the summit.
10 agendas running simultaneously. Trade-truce extension, Boeing / agricultural purchases, semiconductor export controls, rare earths, AI dialogue channel, KRW / JPY / CNY FX, Iran / Hormuz, US-Korea shipbuilding, Taiwan, industrial-overcapacity probe. Not a single event — a 10-variable simultaneous equation.
Korea is not at the table but is the maximum stakeholder. Semiconductors (China-fab operations), rare earths (battery / auto BoM), energy (Middle East dependency), shipbuilding (the $150bn US-Korea KUSPI investment), KRW (foreign-flow stability), Taiwan (semiconductor supply chain) — all are dependent variables of this summit.
Markets have priced “good outcome” substantially. KOSPI 7,500 all-time high. Asymmetric setup — good news already discounted, bad news still has downside. Highest-probability outcome is a “soybeans + Boeing + trade-board” symbolic agreement (~50%).
The rational strategy is not “bet ahead of the summit” but “react after." Pre-event positioning is gambling. Post-communiqué positioning is investing.

1. The full agenda map — 10 variables, simultaneous

1.1 What the US wants vs. what China wants

What the US wants (5 B's):
1. Boeing — 500-aircraft order
2. Beef — resumption of US beef exports
3. Beans — annual 25M tons of soybean purchases
4. Board of Trade — formal trade-committee mechanism
5. Board of Investment — Chinese investment in the US
+ Chinese alignment on Iran nuclear pressure
+ Normalization of rare-earth / critical-mineral supply

What China wants (3 T's):
1. Taiwan — language change ("does not support" → "opposes" independence)
2. Tariffs — additional reductions
3. Technology — semiconductor / AI export-control easing, lifting of 1,000+ entity-list firms
+ AI dialogue channel
+ Chinese EV access to the US market

1.2 The Korea-relevant agendas

Agenda	Korea impact
Semiconductor export controls	Korean fab operations in China; Korean memory demand
Rare earths	Battery / auto BoM costs; EV supply chain
Iran / Hormuz	Energy import costs; KRW; shipping / shipbuilding
Tariffs	Korea’s tariff on US exports (~20%)
US-Korea shipbuilding	$150bn investment package, defense supply chain
KRW / JPY	Foreign-flow stability, export competitiveness
Taiwan	Semiconductor supply-chain diversification, geopolitical discount
Industrial-overcapacity probe	Korea is also under US Section 301 investigation

2. Agenda-by-agenda investment read — Korean-investor frame

2.1 Trade-truce extension — most likely outcome

October 2025 Busan-summit reset US tariffs on China from 57% → 47%; the truce expires November 2026. Reuters reporting suggests China prefers a 1-year extension; the US prefers 6 months.

Investment read: short-term positive. Relief rally possible in China / HK names. But not structural alpha — tail-risk reduction, not new growth catalysts.

2.2 Semiconductor export controls — the most consequential agenda

China seeks easing of advanced semi / AI restrictions; the US seeks continued controls. The April Commerce Department directive halted equipment shipments to Hua Hong (China’s #2 foundry); the April-tabled MATCH Act envisions allied joint export controls.

Why this matters for Korea:

If controls hold (favorable for Korea):
→ Chinese advanced-memory self-sufficiency stays delayed
→ Korean memory (Samsung / SK hynix) oligopoly tightens
→ HBM / high-value memory: US hyperscalers continue Korean sourcing

If controls loosen (mixed for Korea):
→ Short-term: Chinese AI demand expands → Korean memory volumes higher
→ Mid-term: Chinese self-sufficiency accelerates → 2027+ oversupply risk
→ Long-term: Korea / Taiwan / Japan "geopolitical scarcity premium" compresses

Core view: both sides prefer “management” over “resolution” of the semiconductor dispute. Outcome most likely lands in the middle — neither full lift nor full extension. That middle is the most favorable scenario for Korean memory — controls hold (China self-sufficiency stays slow) while AI demand keeps expanding.

2.3 Rare earths — November time bomb

Post-Busan, China issued general licenses on rare-earth export controls, pausing them through November 2026. Whether the pause extends is the central question of this summit.

The numbers tell the severity:

Yttrium (aerospace-critical mineral) price: +6,900% over 12 months
US-bound yttrium exports: -75% over 12 months
China controls 85–90% of global rare-earth processing

Korea and the US already finalized a separate critical-minerals framework — confirmed by the US Treasury in April.

Investment read: pause extension produces short-term relief. But “China unfreezes = problem solved” is wrong — corporates have already begun pricing China-dependence as an internal cost, and non-China supply-chain investment continues regardless.

2.4 Iran / Strait of Hormuz — the under-discussed top-tier variable

The summit was postponed twice because of the Iran war. Hormuz transit fell from 130–150 vessels/day pre-conflict to 4–5/day. One-fifth of the world’s oil and LNG flows through the strait.

Trump pressed China to join an international operation to reopen Hormuz. China is also Iran’s largest crude-oil customer — Hormuz closure hits China directly. China’s recent invitation to the Iranian foreign minister is being read as pre-summit coordination.

Why this matters for Korea: Korea is energy-import-dependent. Hormuz reopens → energy costs fall → KRW strengthens → current account improves → Korea-discount compresses. Hormuz stays closed → energy-cost burden → KRW weakens → import-price inflation.

There is also the HMM-affiliated vessel (Namu) Hormuz fire / explosion incident, which Trump attributed to Iran while urging Korea to participate in allied operations. Korea is investigating the cause.

2.5 US-Korea shipbuilding — already in motion

May 8 saw the signing of the US-Korea Shipbuilding Cooperation MOU (KUSPI). Part of Korea’s $150bn US shipbuilding investment commitment, which itself sits inside a larger $350bn (~₩470tn) total US-Korea investment package.

Investment read: this MOU is more directly bullish for Korean shipbuilders than the summit itself. Shipbuilding gets a re-rating axis from “LNG cycle play” to “US maritime manufacturing rebuild + security supply chain."

But US-onshore shipyard investment carries labor / productivity / union / regulatory risk. Policy premium tends to rerate before earnings show — chasing here is inefficient.

2.6 KRW / JPY / CNY

In April US-Korea bilateral discussion, the Treasury and Korea’s MOEF agreed that “excessive volatility in the KRW is undesirable.” Japan reportedly spent up to $32bn defending JPY. CNY also at a 3-year high.

Investment read: KRW / JPY stabilization improves foreign-flow setup for Korean and Japanese equities. Korea’s $350bn US investment commitment makes KRW weakness a political-and-financial risk linkage. Even without a formal pact, the repeated “no excessive volatility” language compresses upper-bound FX pressure.

2.7 Taiwan — the most dangerous variable

Xi has elevated Taiwan to the summit’s top agenda — a contrast with Busan, where Taiwan was deliberately deferred. China is asking the US to shift its Taiwan-independence language from “does not support” to “opposes.”

Investment read: any drift in Taiwan language moves TSMC, Japanese equipment, Korean semis, defense, and FX simultaneously. Trump’s transactional posture has incentive to use Taiwan as a deal card — agricultural / Boeing purchases in exchange for reduced Taiwan arms sales is plausible. Bipartisan Congressional pro-Taiwan sentiment, however, makes substantive policy change low-probability.

2.8 Remaining agendas

AI dialogue channel: not deregulation — “managed competition” formalization. The assumption that this leads to export-control easing is risky.
Boeing bulk order: 500 737 MAX + dozens of widebodies. The cleaner play is engines (GE Aerospace) and parts (RTX), not Boeing itself.
Industrial-overcapacity probe: Korea is also subject to US Section 301 investigation. Tariff easing isn’t automatically positive.

3. Scenario-based investment strategy

3.1 Four scenarios

Scenario	Probability	Core content	Market reaction
A. Expected (base)	50%	Beans + Boeing + trade board + Iran positive signal + semis / Taiwan “managed”	KOSPI mild rally then profit-taking
B. Upside surprise	20%	Above + Hormuz reopening schedule + extra tariff cuts + rare-earth extension	KOSPI tests 8,000, KRW strengthens
C. Downside risk	25%	Iran progress muted + Taiwan tensions + control tightening signal	KOSPI breaks below 7,000, foreign selling
D. Extreme	5%	Summit collapse or extreme Taiwan rhetoric	Sharp selloff, geopolitical-risk spike

3.2 Why Scenario A is most probable

Corporate executives and analysts expect “smaller deliverables like trade-truce extensions over a major breakthrough.” The 2017 Trump China visit produced large MOUs — many of which were non-binding or multi-year frameworks. The pattern likely repeats.

Korea positioning under Scenario A:

Sector	Stance	Rationale
Korea AI semiconductors	Buy on pullback	Controls hold → Korea oligopoly tightens
Korea substrate / SUMs	Watch → selective buy	AI infra diffusion, earnings verification needed
US-Korea shipbuilding	Watch / selective buy	KUSPI follow-on projects to confirm
China-related names	Trade only, not core hold	Structural-hold rationale weak
Boeing-related	Watch	Engines / parts > Boeing itself

3.3 Scenario C creates opportunity

If the summit ends hollow or Taiwan tensions escalate, KOSPI can selloff. But Korea / Japan AI infrastructure core names become buying opportunities on dislocation. The “controls hold → Korea memory oligopoly tightens” logic is strongest in Scenario C.

Watch list for Scenario C:

Samsung Electronics / SK hynix: scaled accumulation on selloff
Defense: Korea / Japan defense re-rating on Taiwan tension
Energy / gold: geopolitical-risk hedges

4. KOSPI 7,500 already prices in a good outcome

4.1 The asymmetry

KOSPI at all-time-high 7,500 means the market has already priced “summit success → recovery extension” substantially.

Already priced:
- Trade-truce extension expectation
- Iran de-escalation signal expectation
- Strong semiconductor earnings

Not yet priced:
- Taiwan-language drift
- November simultaneous expiry (rare-earth pause + truce)
- Unexpected semiconductor-control shifts
- Recurrence of single-day ₩7tn foreign net selling

Asymmetry: good news limited upside, bad news meaningful downside. Buying ahead of the summit is taking the unfavorable side of this asymmetry.

Why Korea Part 4 framed the larger 2026 question — whether the $6.7bn foreign inflow + 8× PER paradox proves “Korea-discount dissolution” or sets up a value trap. This summit is one variable in that earnings test.

4.2 The 2-name concentration

A large fraction of KOSPI 7,500 is Samsung Electronics (+59%) and SK hynix (+105%). Estimates of “ex-semis KOSPI” land around 4,100. May foreign flow concentrated ₩6tn into the EE sector.

The implication: KOSPI as a whole is not strong; semiconductors are strong. Summit outcome can drive rotation into non-semi sectors, or — conversely — bend the entire index lower if semis break.

5. The real investment angle — not “buy China”

5.1 Surface read vs. real alpha

Market's first-order reading (simple):
"US-China summit → tariff easing → China stocks bounce"
→ China internet, HK, consumer names

Where real alpha lives (complex):
"Structural asymmetry the summit creates"
→ Korea / Japan AI infrastructure core names
→ US-Korea shipbuilding, power, critical-mineral repositioning
→ FX-stabilization beneficiaries
→ Non-China supply-chain investment

China-related names can rally on summit hope. But they fail as structural core holdings — tariffs, tech controls, and Taiwan unresolved means re-rating duration is short.

The higher-quality alpha is in Korea / Japan core companies inside the US-Korea / US-Japan economic-security supply chain. These names have structural demand independent of the summit, see incremental upside if the summit goes well (lower geopolitical discount), and gain premium if the summit goes badly (non-China supply-chain premium expands).

5.2 Korea AI infrastructure — un-broken across all scenarios

SK hynix and Samsung Electronics sit in the most favorable position for “managed resolution” of the US-China semiconductor dispute.

If controls hold:
→ Chinese self-sufficiency stays slow → Korea oligopoly tightens

If controls partially relax:
→ Chinese AI demand expands → Korean memory volume rises

If controls substantially loosen (low probability):
→ Short-term demand spike, long-term Chinese self-sufficiency risk
→ HBM technology gap still holds for years

Substrate and SUMs names — Samsung Electro-Mechanics, Daeduck Electronics, Simmtech — operate on the same logic. The bottleneck-diffusion from GPU/HBM into substrate / packaging proceeds regardless of US-China dynamics. Same logic for optical / CPO peers.

5.3 The November time bomb — H2’s actual risk

More important than this summit is November.

Simultaneously expiring in November:
1. Busan trade truce (47% tariff regime)
2. Rare-earth export-control pause

If both expire together:
→ Tariffs can reset back to 57%
→ Rare earths can be re-weaponized
→ The May summit's best deliverable is a "roadmap to November" framework
→ Without that, H2 uncertainty rises sharply

6. Pragmatic positioning — pre- and post-summit actions

6.1 Pre-summit (now)

No new large-scale buying. KOSPI 7,500 already prices in a good chunk of “summit success.” Asymmetry is unfavorable.
Hold existing positions. Samsung Electronics, SK hynix, Samsung Electro-Mechanics — these structural Korea AI infrastructure positions hold value across scenarios.
Prepare a watch list. Pre-set buy candidates and entry prices by scenario.

6.2 Post-summit — scenario-specific responses

Outcome	Read	Action
Beans + Boeing + trade board (A)	Mild rally then profit-taking	Buy Korea AI infra on pullback; trade China only short-term
Hormuz reopening + tariff cut (B)	KOSPI tests 8,000	Hold large-caps; relief-buy energy / aviation
Taiwan tension + control tightening (C)	Selloff	Scale into Korea AI infra; watch defense
Collapse (D)	Sharp selloff	Cash defense; selective buys after recovery

6.3 Communiqué variables to watch

Semiconductor export-control language: “managed” / “easing” / “tightening”?
Taiwan formulation: “does not support” preserved, or shifted to “opposes”?
Rare-earth pause: extended past November / conditional / unclear?
Trade truce: 6-month, 12-month, or unspecified extension?
Iran / Hormuz: concrete reopening schedule, or signal only?

7. Bottom line

The US-China summit is not a “should I buy China” event. It’s a 10-variable simultaneous equation, and Korea — though not seated at the table — is the maximum stakeholder.

The market has already priced a good outcome into KOSPI 7,500. Good news produces limited upside; bad news produces meaningful downside. The most-likely outcome (“beans + Boeing + trade-board” symbolic agreement) is largely already in the price.

The real risks worth monitoring are Taiwan-language drift and the simultaneous November expiry of the rare-earth pause and the trade truce. The real investment angle is not a China-stock bounce but Korea / Japan AI infrastructure, shipbuilding, power, and critical-mineral supply-chain core names. These keep working regardless of summit outcome — and dislocation creates entry, rather than risk, for them.

Pre-event betting is gambling. Post-event positioning is investing.

FAQ

Q: Should Korean investors buy ahead of the summit? A: KOSPI 7,500 has already priced in a good chunk of “summit success → recovery extension.” Good news has limited upside; bad news has meaningful downside. The asymmetry favors waiting for the communiqué over pre-event positioning.

Q: Should one buy Chinese stocks? A: Short-term relief rally is plausible. But poor structural-hold candidates — tariffs, tech controls, Taiwan all unresolved means re-rating duration is short. Higher-quality alpha is in Korea / Japan AI infrastructure, shipbuilding, and critical-mineral supply-chain names.

Q: Is semiconductor-control easing good for Korean memory? A: It cuts both ways. Short-term, Chinese AI demand expansion lifts Korean memory volumes. Mid-term, Chinese self-sufficiency accelerates → 2027+ oversupply risk. The best scenario is “managed resolution” — controls hold (China self-sufficiency stays slow), AI demand keeps expanding.

Q: What is Hormuz’s impact on Korea? A: Hormuz reopens → energy costs fall → KRW strengthens → current account improves → Korea discount compresses. The HMM Hormuz incident also linked Korean shipping / shipbuilding / energy directly to the resolution path.

Q: What expires in November? A: The Busan trade truce (47% tariff regime) and the rare-earth export-control pause. Simultaneous expiry can reset tariffs to 57% and re-weaponize rare earths. The May summit’s best deliverable is a “roadmap to November” framework.

Q: Why is Taiwan language so important? A: A shift from “does not support” to “opposes” Taiwan independence moves TSMC, Japanese equipment, Korean semis, defense, and FX simultaneously. Trump’s transactional posture has an incentive to use Taiwan as a deal card, but bipartisan Congressional pro-Taiwan sentiment makes substantive policy change low-probability.

Q: What if the summit collapses? A: Scenario D (~5%). On KOSPI selloff, Korea AI infrastructure names (Samsung Electronics, SK hynix, Samsung Electro-Mechanics) become scaled-buy candidates. Defense re-rates on geopolitical tension. Cash defense + selective post-recovery buys is the rational sequence.

This article is for research and informational purposes only and does not constitute investment advice. Sources include Reuters, Yonhap, US Treasury / Commerce announcements, Korea’s Ministry of Economy and Finance, CSIS analyses. Scenario probabilities are analytical estimates and may diverge from actual outcomes. KOSPI / company prices reflect May 8–9, 2026 levels and will move thereafter. Summit outcome can only be confirmed after May 14–15. Analysis can be wrong. Data cut: May 10, 2026 KST.