Foreign Flows on Korea Invest Insights

Is FADU Korea's Sandisk Beta? Foreign Flows and the AI Storage Bottleneck

Sun, 31 May 2026 13:20:00 +0900

Context This note follows SK Hynix vs Micron and the AI Infrastructure Multiple Map. Those notes focused on HBM and mega-cap memory. This one moves down the stack into AI storage, NAND and eSSD controllers. Related hubs: AI HBM Hub and Korea Semiconductor Value Chain Hub.

TL;DR

FADU is starting to behave like a Korean high-beta proxy for Sandisk, but the relationship is still early. The strongest evidence is not daily price correlation; it is foreign flow. Since May 2026, foreigners net bought about KRW 445.1bn of FADU while domestic institutions net sold about KRW 215.2bn. Over the same period, FADU rose +38.1% and Sandisk rose +35.0%.

The price beta is weaker than the SK Hynix-Micron pair. Sandisk’s prior-U.S.-session return versus FADU’s next Korea-session return correlation is +0.16 since May and +0.42 over the latest 10-14 trading days. Micron-to-Hynix is +0.57 and +0.55. Translation: foreign flow says yes, domestic institutions say no, and the broader market pair is still forming.

The conclusion: FADU is a candidate Korean beta for AI storage/NAND, but valuation is already demanding. Structurally, FADU is cleaner. Numerically, Jeju Semiconductor looks cheaper. Globally, Sandisk screens as the cleaner conditional buy candidate for FY2027, while Western Digital already prices in much of the HDD bottleneck.

Data Basis and Coverage

The analysis covers 2026-01-01 to 2026-05-29. FADU and SK Hynix prices and investor flows use Research OS local DB tables prices_daily and investor_flow_daily. Sandisk and Micron prices use Yahoo Finance/yfinance.

Coverage limits matter. Some May FADU individual flow and detailed institution-subtype fields are blank in the local DB, so this note judges FADU mainly through foreign and institution totals. U.S. names do not have Korean-style investor-flow data.

1. Is FADU Following Sandisk?

On price level, yes. FADU rose from KRW 21,250 to KRW 112,100 year-to-date, up about +427.5%. Sandisk also rallied sharply. Both are being repriced around AI storage and NAND scarcity.

But daily return linkage is still modest.

Item	Sandisk → FADU	Micron → SK Hynix	Interpretation
YTD price-level correlation	+0.91	high	Same broad uptrend
Prior U.S. day → next Korea day return	+0.12	+0.46	FADU pair still weak
Since May	+0.16	+0.57	Hynix-Micron is much more structural
Latest 10-14 sessions	+0.42	+0.55	FADU beta has recently strengthened

The key distinction is that “both went up” is not the same as “FADU mechanically tracks Sandisk.” FADU is not yet a mature pair trade. But the recent correlation jump suggests investors are beginning to map the relationship.

2. The Real Signal Is Foreign Flow

Since May, FADU’s flow profile has been unusually clear.

Period	FADU Return	Sandisk Return	FADU Foreign Flow	FADU Institution Flow	Foreign Positive-Day Ratio
2026 YTD	+427.5%	strong rally	+KRW 528.2bn	-KRW 173.2bn	46.4%
Since May	+38.1%	+35.0%	+KRW 445.1bn	-KRW 215.2bn	83.3%
Latest 10-14 sessions	+11.7%	+9.5%	+KRW 346.4bn	-KRW 194.1bn	85.7%

This is not a domestic-institution-led move. It is closer to a foreign-discovery trade in a Korean high-beta proxy for AI storage/NAND.

3. Why It Is Weaker Than SK Hynix-Micron

SK Hynix and Micron are both large-cap global memory stocks tied to the same HBM/DRAM cycle. Their earnings line items, investor base and product overlap are more directly comparable.

FADU and Sandisk are different. Sandisk is a global NAND and data-center SSD company. FADU is closer to an eSSD controller, firmware and enterprise SSD solution fabless company. Both benefit from the AI storage cycle, but the economics and customer-validation risks are not identical.

The clean framing is:

Sandisk = global AI storage/NAND thermometer FADU = foreign-discovered Korean high-beta proxy Key validation = continued foreign buying + breakout above KRW 118,500 + less institution selling

4. FADU and Jeju Through the “Next SEMCO” Lens

The Samsung Electro-Mechanics re-rating showed how a seemingly small component can be reclassified as an AI infrastructure bottleneck. Applying the same frame to memory/storage gives five tests: bottleneck, P up, Q up, customer validation and multiple reclassification.

Company	Bottleneck	Strength	Weakness	View
FADU	AI eSSD controller / firmware / SSD solution	Cleanest structural AI-storage bottleneck in Korea	Valuation already rich	Watch / Wait
Jeju Semiconductor	LPDDR4X, MCP, legacy low-power memory	Looks much cheaper on current numbers	Needs proof that 1Q26 repeats	Tactical Watchlist
Sandisk	NAND / data-center SSD	Global AI storage/NAND thermometer, cheaper FY2027 setup	NAND peak-out risk	Conditional Buy candidate
Western Digital	HDD capacity / nearline storage	Strong HDD cash-flow bottleneck	Much of the premium reflected	Low appeal for fresh entry

5. FADU: Strong Structure, Difficult Price

FADU’s structural thesis is strong. AI data centers are not just a GPU/HBM story. KV-cache, checkpoints, training/inference pipelines, enterprise SSDs, controllers and firmware become parallel bottlenecks. FADU sits near the eSSD controller and firmware qualification layer.

Public reporting says FADU posted Q1 2026 revenue of KRW 59.5bn and operating profit of KRW 7.7bn, returning to profitability, with disclosed new orders reaching KRW 166.3bn by April. (StorageNewsletter)

The issue is price. The input research cites MarketScreener estimates of KRW 323.3bn 2026E revenue and KRW 584.8bn 2027E revenue, with 2027E P/E around 38x. That is not cheap for a company that still needs to convert orders into revenue and margin.

Upgrade Conditions

Watch Item	Buy Upgrade Trigger
Price	Break and hold above KRW 118,500 with turnover
Flow	Continued foreign buying and less institution selling
Earnings	Q2 revenue KRW 70-80bn, OPM above 15%, ideally 20%
Orders	Cumulative new orders above KRW 400bn
Products	Gen6 / CXL / low-latency SSD design wins

Invalidation: foreign buying stops despite Sandisk strength, FADU breaks below KRW 105,900 with foreign selling, orders fail to convert, OPM falls below 10%, or controller mix declines.

6. Jeju: Cheaper Numbers, Higher Repeatability Risk

As discussed in the Jeju Semiconductor 1Q26 note, Jeju is not an HBM company. It benefits from shortages in LPDDR4X, MCP and low-power memory as Samsung, SK Hynix and Micron allocate more capacity toward HBM and server DRAM.

Jeju’s Q1 2026 operating margin was 37.2% on KRW 180.4bn revenue and KRW 67.1bn operating profit. TrendForce expects 2Q26 LPDDR4X ASPs to rise +70-75% QoQ and LPDDR5X +78-83%. (TrendForce)

On simple annualized Q1 numbers, Jeju looks cheaper than FADU. But the question is repeatability. If Q1 was peak margin, it is a trap. If Q2 revenue stays above KRW 180bn and OPM remains above 30%, it becomes a tactical undervaluation candidate.

7. Global Read-Through: Sandisk Screens Best, WDC Less Clean

Sandisk is the global AI storage/NAND thermometer. Fiscal Q2 2026 revenue was $3.025bn, GAAP net income was $803m and non-GAAP EPS was $6.20. The company guided fiscal Q3 revenue to $4.4-4.8bn and non-GAAP EPS to $12-14. (Sandisk)

Western Digital has a strong HDD bottleneck and cash-flow story. Fiscal Q3 2026 revenue was $3.337bn, GAAP gross margin was 50.2% and operating cash flow was $1.12bn. But fresh-entry appeal is lower because much of the HDD scarcity premium is already recognized. (Western Digital)

Company	Core Exposure	View
Sandisk	NAND, data-center SSD	Conditional Buy candidate
Jeju Semiconductor	LPDDR/MCP legacy memory squeeze	Tactical Watchlist
FADU	AI eSSD controller / firmware	Watch / Wait
Western Digital	HDD capacity / cash flow	Less attractive for fresh entry

Final View

FADU is starting to be interpreted as Korea’s Sandisk beta. It is not yet a mature global pair like SK Hynix-Micron. But KRW 445.1bn of foreign net buying since May is too large to dismiss.

Setup	Action
Sandisk strong + FADU foreign buying continues + KRW 118,500 breakout	Recognize FADU as Korean AI-storage beta; conditional Buy upgrade
Sandisk strong but FADU foreign flow slows and institutions keep selling	Beta failure; watch for theme peak
Sandisk corrects but FADU foreign buying persists	Upgrade toward FADU-specific alpha
Break below KRW 105,900 + foreign net selling	Risk management

One sentence:

FADU is being discovered as Korea’s high-beta proxy for AI storage/NAND, but the right stance is Watch / conditional Buy, not blind chase.

Evidence Classes

[Fact]

FADU Q1 2026 revenue was KRW 59.5bn, operating profit was KRW 7.7bn, and disclosed new orders reached KRW 166.3bn by April. (StorageNewsletter)
TrendForce expects LPDDR4X ASP +70-75% QoQ and LPDDR5X +78-83% QoQ in 2Q26. (TrendForce)
Sandisk fiscal Q2 2026 revenue was $3.025bn, GAAP net income was $803m and non-GAAP EPS was $6.20. (Sandisk)
Western Digital fiscal Q3 2026 revenue was $3.337bn, GAAP gross margin was 50.2% and operating cash flow was $1.12bn. (Western Digital)
FADU-Sandisk flow/correlation work is based on Research OS local DB and yfinance data through 2026-05-29.

[Inference]

FADU is not a direct Sandisk tracker yet; it is a foreign-discovered Korean high-beta proxy for AI storage/NAND.
Jeju is a tactical low-power memory squeeze candidate, not the cleanest structural AI-storage bottleneck.

[Blocked]

FADU customer-level orders, controller ASP, gross margin and Gen6/CXL design wins.
U.S.-name investor-flow data comparable to Korean foreign/institution flow.
Jeju 2Q26 repeatability and working-capital quality.

KOSPI Foreign Ownership Is High, But Samsung and SK Hynix Are at 2026 Lows

Tue, 26 May 2026 10:30:00 +0900

This is a follow-up to Korea Foreign Investor Flow Analysis. That note showed that most 2026 foreign selling was concentrated in Samsung Electronics and SK Hynix. This note asks the next question: how much ownership-based selling pressure is left?

TL;DR

KOSPI market-cap-weighted foreign ownership is still high. On the user’s benchmark series, it rose from 36.02% at the start of 2026 to 38.56% as of May 22.
Yet the two stocks actually being sold are already near 2026 lows. Research OS local DB shows Samsung Electronics foreign ownership falling from 52.37% to 48.32%, and SK Hynix from 53.81% to 51.62%.
That divergence exists because KOSPI ownership is value-weighted. Foreigners can sell shares, but if their remaining Samsung/SK Hynix positions surge in price, their market-cap-weighted KOSPI ownership can still rise.
The practical trigger is now flow speed: first stabilization needs KOSPI foreign ownership below roughly 38.2% plus five-day average foreign net selling below about KRW 1tn per day.

The Core Data

Item	Start of Year	Current	Change	Read
KOSPI market-cap-weighted foreign ownership	36.02%	38.56%	+2.54pp	Foreign-owned market value remains high
Samsung Electronics foreign ownership	52.37%	48.32%	-4.05pp	2026 low; 2022+ local DB low
SK Hynix foreign ownership	53.81%	51.62%	-2.19pp	2026 low, but not a long-term trough
Samsung Electro-Mechanics	37.63%	38.21%	+0.58pp	Less ownership pressure
Daeduck Electronics	15.96%	17.23%	+1.27pp	Looks more like accumulation

Yonhap Infomax reported that Samsung Electronics’ 48.32% foreign ownership was the lowest since September 2013, while SK Hynix’s 51.62% was the lowest since May 2023. It also reported YTD foreign net selling of KRW 50.2027tn in Samsung and KRW 34.2279tn in SK Hynix. (Yonhap Infomax)

Samsung’s historical anchor is much higher. In May 2019, Korea JoongAng Daily reported foreign ownership at 57.33%, above the prior 2001 record of 57.30%. (Korea JoongAng Daily)

Why The Aggregate Still Looks High

The KOSPI foreign ownership ratio is a market-value measure, not simply a share-count measure. Seoul Economic Daily explained the gap clearly: foreign selling pulled down the ratio, but price appreciation in foreign-held shares added even more to the ownership value. (Seoul Economic Daily)

That means the aggregate KOSPI ratio says foreign portfolio value is still heavy, while Samsung/SK Hynix share-count ownership says the two key positions have already been reduced materially.

Flow Is Still Not Calm

The problem is speed. From May 18 to May 22, foreigners net-sold KRW 14.4477tn in KOSPI. SK Hynix and Samsung represented roughly 73% of that one-week selling. From May 7 to May 22, foreigners net-sold KRW 46.3383tn, with Samsung and SK Hynix together accounting for 82.9%. (Daum / Energy Economy)

Research OS local all-listed-stock proxy points in the same direction:

Period	Total Foreign Net Buy	Samsung	SK Hynix	Two-Stock Share
May 18-22	-KRW 13.16tn	-KRW 5.28tn	-KRW 5.35tn	80.7%
May 7-22	-KRW 43.31tn	-KRW 18.94tn	-KRW 19.59tn	89.0%
Jan 2-May 22	-KRW 89.20tn	-KRW 50.41tn	-KRW 34.39tn	95.1%

So the ownership level is getting closer to a clearing point, but the flow is still too fast to call the selloff over.

Practical Framework

Stage	Conditions	Interpretation
Selling pressure remains	KOSPI foreign ownership above 38.5% and five-day average selling above KRW 2tn/day	Rebalancing still active
First stabilization	Below 38.2% and five-day average selling below KRW 1tn/day	Selling speed has slowed
Meaningful stabilization	Below 38.0% and Samsung/SK Hynix ownership stops falling	Mechanical selling largely absorbed
Re-buying setup	37.5-38.0% stable, foreign net buying for 3-5 sessions, KRW stabilizes	Korea beta can be re-entered

Investment Read

For Samsung Electronics, the foreign share-count ratio already looks low. Below 48% would look like a late-stage reduction signal. But the stock still needs foreign selling speed to slow before new money gets a cleaner entry.

For SK Hynix, 51-52% is a lower-neutral zone, not a full trough. A move toward 50.5-51.0% with price resilience would make the risk/reward more comfortable.

The final line is simple: KOSPI-level foreign ownership still argues for rebalancing pressure, but Samsung and SK Hynix ownership already argues that the selling is closer to exhaustion than the headline flow suggests. This is not a buy-now signal. It is a wait-for-flow-to-slow signal.

Disclaimer: For research and information purposes only. Not investment advice.