Foreigners Sold 150T Won: Why SK Hynix Falls First

This piece is the follow-up to three earlier posts. The structure edition, Memory's Great Power Reversal, covered "why it went up," and the event edition, KOSPI 8000 Breaks: Real Crisis or a Fit?, covered "why everything fell together in June." This one asks: "since July, why does it keep cracking every time FX moves, and how low can it go?" The ADR thread continues from the report SK Hynix ADR: Can It Reach 5 Million Won?. Jargon is explained inline, with a glossary at the bottom. Figures are cross-checked against reporting and data as of July 2, 2026, and estimates are marked as such.

Today's Scene: Blue Again

The screen is blue again. The market confirmed today that June's synchronized crash was not a one-off fit but a trend.

Foreigners sold a net 1.12 trillion won on the KOSPI today alone. That makes three straight sessions and a cumulative net sell of over 2 trillion won. Across the first half, foreigners dumped about 150 trillion won on the KOSPI.

The trigger came overnight from the US. The Philadelphia Semiconductor Index fell 6.27%, with Micron down 10.57% and AMD down 6.89% in particular. When the memory bellwether dropped double digits in the US, Samsung and SK Hynix took the hit at the open.

And underneath it all lies the KRW/USD rate. After briefly dipping to 1,510 on news of US-Iran peace talks in late June, the rate bounced back to the 1,540 range. That is the highest level since the financial crisis (March 2009).

The observation that "every time FX moves, not just Hynix but Samsung too falls harder" is accurate. But why exactly when the won weakens, and why exactly these two names take the hardest hit? And is this about the ADR? Let's unpack it one step at a time.

Why the Instinct "It Falls Every Time FX Moves" Is Right

Bottom line first: won weakness and foreign selling move as one body, in a feedback loop. And at the center of that loop sit Samsung and SK Hynix.

Doing the Math From a Foreign Investor's Seat

When a foreign investor buys Korean stocks, they are really making two bets at once.

Suppose a US fund that brought dollars buys Hynix. Even if the stock is flat, if the won weakens 5% (FX rises), the dollar value of the asset shrinks 5%. In other words, won weakness is a "loss even standing still" for foreigners.

So when FX crosses 1,500 and jumps to 1,540, 1,550, foreigners move in two ways. First, un-hedged money sells stock to avoid the FX loss. It converts the won to dollars and exits, and in the process dollar demand rises and FX climbs higher. Second, hedged money has covered the risk with won futures and the like, but when FX volatility rises and won rates and hedging costs climb, the hedging cost eats the return, so they eventually trim the position.

In both cases the conclusion is selling. And selling pushes FX up, and the higher FX invites more selling. It is the vicious cycle of "FX up -> foreign selling -> won weaker still -> more selling." Foreigners selling over 2 trillion won across three straight days is the physical evidence of this loop.

Why Samsung and SK Hynix Specifically

When FX rises, not every stock takes an equal hit. There are three reasons these two get hit first and hardest.

First, foreigners own the most of them. Both names have high foreign ownership and heavy trading. When foreigners cutting risk sell "whatever they can sell," these are the first assets their hand reaches for.

Second, more than half the KOSPI is these two. Samsung Electronics, Samsung Electronics preferred, SK Hynix, and SK Square together approach 60% of the KOSPI's market cap. Selling the index (futures, ETFs) automatically sells these two, and when these two drop the index drops, and when the index drops they sell again. Selling begets selling.

Third, they are the assets that rose the most. Hynix rose about 13x over 19 months. In a phase of pulling liquidity back, the winner sitting on the fattest gains becomes the first cash machine. Dumping a losing small-cap raises no money, but selling Hynix instantly turns into a large pile of cash. This is exactly the mechanism the event edition called "the winner sells first in a crisis."

Why the Drawdown Is This Steep: Single-Stock Leverage as an Amplifier

One more thing gets layered on. It is not the force that sets the direction, but the amplifier that magnifies the drawdown and volatility: single-stock leverage.

The timing is exquisitely bad. On May 27, 2026, Samsung Asset Management listed single-stock leveraged and inverse ETFs that track Samsung and SK Hynix at plus or minus 2x the daily return. Among them, the KODEX SK Hynix Single-Stock Leverage ETF hit about 2.7138 trillion won in net assets within two weeks of listing, ranking first in retail net buying, trading value, and net assets alike. The moment the amplifier was installed (late May), the June-July volatility blowup arrived.

A 2x ETF rebalances daily, and it has the property of "selling when it falls." To keep 2x exposure, the fund resets its exposure near the close each day. It buys more on up days (buying the top) and must sell on down days (selling the bottom). So on a day like today when Hynix drops, this ETF pours mechanical selling near the close to reduce leverage, deepening the decline. In a rally it does the opposite, buying and inflating the gain. It is a procyclical structure that follows and amplifies the trend.

There is also volatility decay and path dependence. In a market that whipsaws up and down, this daily rebalancing compounds, so even when the underlying returns to where it started, the leveraged product takes a loss. The higher the volatility, the bigger the decay. So money trapped in leverage tires faster and is driven sooner into stop-losses and liquidation.

Margin loans and forced selling overlap here. With market-wide margin loans above 38 trillion won, a sharp drop in Hynix drives retail investors who bought on debt into margin calls and forced liquidation. The event edition's "you sell what you can sell, and the winner sells first" gets amplified by leverage.

The data leaves fingerprints too. In late June, Hynix daily candles had abnormally wide ranges. For example, June 23 drew about an 18% range in a day, from 2.536M to 3.002M intraday, and June 26 swung between 2.6M and 2.955M. These whipsaw candles are precisely the fingerprint of a market with leverage and single-stock ETF rebalancing layered on.

But keep the weighting sober. Single-stock leverage is not the cause but the amplifier. The direction (down) was set by FX, foreigners, and deleveraging, and the 2x ETFs and margin only make that move bigger and faster, especially near the close. Net assets of 2.7 trillion won (about 5 trillion won of notional exposure at 2x) are not "dominant" relative to Hynix's market cap and daily trading value, but they are a meaningfully large marginal supply at the close-rebalancing moment.

For reference, after the ADR lists on July 10, US-side leverage could attach too. The US already has plenty of single-stock 2x ETFs on names like Nvidia and Micron. If US single-stock leverage products appear on the Hynix ADR, one more channel opens for US overnight leveraged trading to transmit into the Korean common shares through arbitrage.

Are Foreigners Really Dumping Korean Hynix "Because of the ADR"?

This is the core question. Framed as a hypothesis: "The Hynix ADR lists on the US Nasdaq on July 10 anyway. So foreigners have no reason to buy Hynix in Korea while bearing FX risk. They can just buy it in dollars in the US. So aren't they pulling their Korean shares out ahead of time?"

Intuitively it sounds plausible. But dissected coldly, half of it is a misunderstanding and half of it is fact. Let's go in order.

The Misunderstanding: Buying the ADR Leaves FX Risk Fully Intact

This is the point the ADR report already stressed. The SK Hynix ADR price is roughly the Korean share price times FX.

That is because the ADR and the Korean common are interchangeable (convertible), and arbitrage keeps the two prices tied at "Korean share price times FX" at all times. So even buying the ADR in dollars in the US, the won risk is still inside it. When the won weakens, the ADR price falls too (in dollar terms).

Apple stock is a genuine US company that earns revenue in dollars, so it is truly a "dollar asset." But Hynix is a won company, so wrapping it in the ADR packaging leaves the contents still in won. In other words, the very logic of "abandoning the Korean market for the ADR to avoid FX risk" does not hold economically. Buy it in the US and the won risk follows you.

Going one step further, this listing is an event that could actually ease won weakness. Worth chewing on. Hynix is issuing 17.79 million new shares this time, raising up to 45.45 trillion won (about $29.6 billion). It is the largest ADR ever, surpassing Alibaba ($21.8 billion in 2014). If a large chunk of this money comes in as dollars and gets converted to won, dollar supply in the market rises and actually caps the upward pressure on FX. In fact, in late June the press and FX authorities viewed this listing as "a potential card to ease prolonged won weakness" (Korea Times). With KRW/USD having spiked to 1,542.7, the highest since the financial crisis, a $29 billion inflow could be a welcome rain in a drought.

So the causal chain "ADR listing -> foreigners exit the Korean market to avoid FX risk" is, at least on the grounds of FX-risk avoidance, wrong. The ADR does not remove FX risk, and the listing itself is if anything friendly to the won.

The Fact: There Really Is Short-Term Selling Pressure

That said, the instinct is not entirely wrong either. The path is different, but there are several channels through which selling really does pile onto the Korean common ahead of the ADR.

First, bookbuilding hedge supply. In the demand-forecasting (bookbuilding) process just before listing, foreign institutions seeking an ADR allocation sell Korean futures and cash (short) in advance to cover risk. It is a hedge to align the direction of "shares soon to be received as ADRs" with "the common they hold now." This shows up as short-term overhang (pending selling) in the days before listing.

Second, rotation into the new shares. Some global passive and institutional money finds it easier, for operational, custody, and mandate reasons, to switch Hynix exposure from the Korean common to the Nasdaq ADR. So there is money that sells the Korean common to participate in the ADR offering. Note, though, that these new shares use a third-party allotment, so retail cannot get in (they only bear the dilution), and the new shares are additional supply, not an exchange of existing stock.

Third, flowback risk. Korean ADRs have historically often traded cheaper than the common (at a discount) (KB, Shinhan, POSCO, etc.). If the ADR ends up cheaper than the common after listing, arbitrageurs convert the ADR into the common and sell it in Korea, a reverse flow that presses the common.

Fourth, the "peak event" psychology. A large part of the price surge was "ADR expectation." Once the listing is imminent, "buy the rumor, sell the news (sell-on-news)" profit-taking emerges. This psychology is mixed into the early-July decline too.

Then What About Liquidity? A Sharper Rebuttal

Let's go one step deeper. Even though "FX-risk avoidance" as a reason is wrong, there clearly are reasons, unrelated to FX risk, for foreigners to buy the ADR instead of the Korean common. And the point that, as these accumulate, Korean-market demand and trading (liquidity) for Hynix can thin out is correct. Let's go in order.

The real reasons, aside from FX risk, that foreigners choose the ADR. On access and convenience: they simply buy it in dollars, during US hours, through a US brokerage account. There is no friction of Korean foreign-investor registration (IRC), won conversion, or Korean settlement. On passive/ETF inclusion: US semiconductor ETFs and indices like SOXX can only hold the ADR (not the Korean common). US index money enters only through the ADR. On "US-listed only" mandates: many US funds cannot buy Korean stocks at all under their mandate. For them, the ADR opens the door to buy Hynix for the first time (new demand, if anything). Finally, liquidity begets liquidity. Trading crowds toward where trading is already active.

So the part where the point is right. If some of the new foreign demand goes to the US ADR rather than the Korean order book, the incremental buying that would enter the Korean common shrinks, and Korean-side trading and liquidity can thin out. Up to here it is accurate.

But why you must not oversimplify it into "Korean liquidity collapse -> price crash." Two forces block it. First, conversion and arbitrage reverse it. The ADR and the common are interchangeable (conversion is free in Korea), and arbitrage ties them at "common is roughly ADR divided by FX." When US demand crowds in and the ADR gets expensive, arbitrageurs buy the cheap Korean common, convert it to the ADR, and sell it. US demand flows back into the Korean order book, propping up the common's price and trading. So the price does not leak to one side. Second, it actually grows the whole pie. Like the "US-listed only" mandate money above, when money that could not buy before comes in fresh, the ADR grows demand rather than siphoning it, and part of that effect is transmitted to the Korean common through arbitrage.

The real remaining risk is the migration of "volume," not "price." Over time, existing holders convert the common to the ADR, and price discovery (the main trading) shifts to US overnight hours, so Korean volume can slowly thin. That said, past Korean ADRs (POSCO, KT, SKT, etc.) were mostly thinly traded and at a discount, and SKT delisted its London ADR in 2023 for lack of trading. That means the home market stayed the primary market. Conversely, Hynix is the largest ADR ever ($29 billion), an AI bellwether, and gets included in US semiconductor ETFs, so it is a candidate exception where volume migration could be larger than in the past. Even so, with the National Pension Service (about 8%), domestic institutions, and a thick retail base holding the common, Korea remains the primary market for the time being. Liquidity migration should be viewed as a gradual decline, not a cliff.

Separate the time axis from today's (7/2) drop. This liquidity migration is a structural story over months to years after listing. Today's -12.6% (on a closing basis) is not that; the cause is FX and deleveraging. "Long-term liquidity migration" and "today's crash" are questions on different clocks.

Bottom Line: The ADR Is Not the Culprit Behind Today's Drop

Weighing it honestly, here is how it looks.

In other words, the real force that broke Hynix and Samsung today is the "FX and deleveraging loop," and the ADR is one temporary overhang layered on top. The feeling of "isn't this because of the ADR" accurately captured the bookbuilding hedge and rotation supply, but that is not because of a motive to avoid FX risk, and after the listing the dollar inflow could actually help FX and reverse this pressure. Do not confuse the order of causes. FX comes first, and the ADR is a side branch.

The 4,700-Trillion-Won Capex: Good News That Weighs on the Market

The second question: "Samsung and Hynix keep talking about enormous capex. How much did that affect the market?" Here lies the most symbolic recent number.

The Black Hole Called 4,700 Trillion Won

The investment scale Samsung Electronics and SK Hynix have officially announced totals about 4,700 trillion won combined (on a mid-to-long-term cumulative basis, including the Yongin cluster, US fabs, and fab expansions). For 2026 alone, the two companies' DRAM and NAND capex is projected to rise about 19% year over year. The problem is that they cannot fund this out of their own cash alone. Bond issuance and bank borrowing are unavoidable, and Samsung has signaled fund deployment on the scale of hundreds of trillions of won.

Here is the real reason the market tenses up. To borrow Etoday's phrasing, Samsung and Hynix have become "a black hole soaking up the market's cash."

Crowding-out is the first of it. When these two top-credit companies soak up thousands of trillions of won in the bond market, lower-rated small and mid-sized firms are stripped of the very chance to issue corporate bonds. Money crowds only into large semiconductor projects. Sweeping up the back-end infrastructure is a problem too. Running semiconductor fabs and AI data centers requires power grids and water facilities, and even the project-financing (PF) money to build those is likely to be absorbed by the mega semiconductor projects.

In other words, capex has become an intake that soaks up the liquidity of the entire Korean capital market, beyond any individual company's financial issue. Layer on Hynix's 45-trillion-won ADR raise and record-sized IPOs, and the picture completes: "money gets sucked only into semiconductors while the rest of the market dries up." It dovetails exactly with the "raising cash (deleveraging)" phase the June event edition flagged.

Why Good News Reads as Bad News

Capex expansion is normally a good signal. It means "demand is overflowing, so we're building more factories." Yet the market now reads it as a burden. Three reasons.

First, it pulls forward future oversupply by its own hand. The memory cycle formula from the structure edition is always the same: record margins -> everyone expands -> supply explodes 18 to 36 months later -> price collapse. Today's 4,700-trillion-won expansion is the seed of 2027-2028 oversupply. The more the ADR-raised money accelerates HBM and Yongin expansion, the more it becomes a self-reinforcing loop that pulls the downcycle forward by its own hand.

Second, as capital intensity rises, the valuation logic wobbles. What broke Big Tech's golden age was precisely the collapse of the "low capex plus high cash flow" formula. Now that capex explosion is spreading to the three memory makers themselves. For now they hold up because margins are around 85%, but as depreciation swells and the cycle turns once, this enormous plant flips into a fixed-cost bomb.

Third, the demand side (hyperscalers) turns FCF-negative. Warnings keep coming that the Big Tech buyers of memory could flip to negative free cash flow (FCF) from the second half of this year due to AI investment. If rates rise too (the event edition's "funding squeeze"), the cost of the private credit and lease structures propping up that investment jumps too. "When the buyer's wallet dries up, the seller's (memory's) boom wobbles too." The market reads the 4,700-trillion-won capex news from this angle.

So How Low Can SK Hynix Go?

Now the most pressing question: "If this trend continues, how low?" Let's plot the coordinates the analyst way, confirm the support with data, then split into scenarios.

Current Coordinates and Technical Support

Reading the recent path from daily data:

The 7/2 close of 2.2295M already sits at the top of the June base (1.85M-2.16M), and the intraday low was pushed down to 2.197M. From here the natural first support is around 2.0M won. It is the psychological "2-million Hynix" and the top of the mid-June pullback. If that line breaks, the next is 1.85M won: the June intraday low, coinciding with the bear analyst target that comes up later. In other words, the current drop is strongly a "return to the June base." It gives back the entire surge spike (from 2.16M in mid-June to 2.96M on 6/22) and returns to where it started.

Compared to Samsung: What "Support" and a "Weak Signal" Tell Us

Samsung Electronics also broke -9.5% the same day, but the two sit in subtly different spots. Reading this difference makes the concept of "support" click.

Support is a price floor that buyers defended in the past, saying "we won't sell below here," and bounced the price off of. Its breaking means the buyers guarding the floor capitulated, which is read as a technically bad signal. Three reasons: demand exhaustion (no more backstop), the people trapped at that price waiting to sell at break-even so a supply wall forms overhead (support flipping to resistance), and no clear support below so it flows down faster on stop-losses and mechanical selling.

The two names sit differently now. Hynix's 7/2 low of 2.197M is the top of the June base (1.85M-2.16M). The whole cushion from 2.16M down to 1.85M remains below. Samsung tested its June low (283K) with a 7/2 intraday low of 281.5K. The decline has run further and touched the last line of defense, so at this moment Samsung is relatively weaker.

That said, it is not a "confirmed breakdown." Samsung briefly pierced its June low intraday but closed at 286K, above it. To count as a technical breakdown, it usually has to break on a closing basis and confirm the next day (follow-through). A wick (intraday) alone is a warning signal but unconfirmed.

And this is strictly a "price (chart) signal." It is not the leading indicators that come up next (spot prices, book-to-bill, HBM ASP). Even breaking the low does not change Samsung's earnings or the memory cycle. A support break means supply and sentiment weakened further, not that a downcycle has begun.

In short, "Samsung is a weaker signal than Hynix" means selling pressure has run further and is testing the last support (relative weakness plus a technical warning), not that fundamentals rolled over first.

Analyst Targets: The Largest Split Ever

Opinion on the Street has never been this divided. The split itself means "no one is sure about the cycle peak."

The gap between bull and bear is over 1.3M won. The bull case leans on "2027 HBM prices double again and server-memory demand rises 60%-plus (supply can't keep up)." The bear case leans on "already up too much, with capex, FCF, and FX burdens stacked on."

Four 12-Month Downside/Upside Scenarios (Subjective Estimate)

Combining the support, targets, and cycle mechanics above, split into four branches.

The direct answer to "if this trend continues, how low":

The most probable near-term floor is 2.0M won plus or minus (1.85M-2.15M). This is not because fundamentals broke, but the spot where FX and flows force the spike to be given back. From the 7/2 close of 2.2295M (-24.7% from peak), 2.0M is another -10% and 1.85M another -17% or so.

Below that (1.3M-1.7M) only opens if a "real downcycle" arrives. That is, not just the price but the leading indicators must actually roll over. We are not in that regime yet. Conversely, if FX stabilizes and the ADR draws strong demand, today's drop becomes "a healthy correction shaking out an expensive spike," and a retry of 2.90M-3.50M is also possible.

Why You Shouldn't Casually Call for Below 1.3M

In the past downcycle (2018 to 2019), Hynix operating profit fell from 20.8 trillion to 2.7 trillion won (-87%), and the stock fell 40% to 50% from its peak. Plug that in naively and the 2.96M peak times -50% gives a theoretical floor around 1.48M won.

But two things are different this time. First, supply is structurally blocked (advanced-process difficulty, power, and permitting bottlenecks). Second, multi-year HBM sold-out contracts raise the earnings floor above the past. So, as the structure edition concluded, "the cycle hasn't disappeared; it has lengthened and its trough has risen." Keep 1.3M-1.7M open as a worst-case tail scenario, but it is not the main scenario in the current phase.

Is This a Fit, or the Start of a Real Decline?

Let's close with the most important question. Is this -24.7% a "liquidity correction shaking out an expensive spike," or "a real decline that has topped out and started down"? How do you tell?

The answer is already in the event and structure editions. Watch leading indicators, not the price. If only the price is falling now, that is, if the cause is macro/liquidity like FX, foreign flows, and US contagion, and the leading indicators below stay intact, this is closer to a correction inside a bull market (① or ②). Conversely, if the leading indicators actually start to roll over, that is the real start of a downcycle (③).

The leading indicators to check:

1. Capex acceleration at the three memory makers plus China (CXMT). The seed of overshoot (already turning on: 4,700 trillion).
2. Hyperscaler capex guidance cuts / FCF turning negative. The root of demand (warnings emerging).
3. HBM/DRAM book-to-bill below 1. Orders failing to keep up with shipments.
4. Spot price rolling over before contract price. The most reliable leading signal for price.
5. HBM premium (ASP) narrowing. HBM no longer being scarce.

The current read (July 2, 2026): #1 (capex) and #2 (FCF warnings) have begun to turn on, but #3 to #5 (spot price, book-to-bill, HBM ASP) show no sign of rolling over yet. If anything, the view that 2027 HBM prices rise again is dominant. If so, the reasonable judgment is that this -24.7% is closer to scenario ① (liquidity correction). But if FX stays stuck above 1,500 and indicators #4 and #5 begin to turn on, it crosses into ③. Monitoring that turning point is the key.

In One Line

Hynix and Samsung cracking every time FX moves right now is not because of the ADR, but because won weakness (1,540, the highest since the financial crisis) triggers foreign selling, and that selling pushes the won weaker still, in a feedback loop. The ADR is just one temporary hedge overhang layered on top, and if anything a $29 billion inflow could help FX. The 4,700-trillion-won capex weighs on the market now as a black hole soaking up capital-market liquidity and the seed of future oversupply, but there is no sign the fundamentals (spot price, book-to-bill, HBM ASP) have rolled over yet. So today's -24.7% (7/2 close of 2.2295M) is reasonably viewed as a liquidity correction giving back an expensive spike (first support 2.0M, next 1.85M), and the real downcycle at 1.3M-1.7M only opens if leading indicators actually roll over. If FX stabilizes, the ADR draws strong demand, and the leading indicators hold, a retry of 2.90M-3.50M is also possible. It is a high-volatility phase where the direction is still undecided.

Glossary

Disclaimer

This piece is research collected and cross-checked from public reporting, statistics, and local data for informational purposes, and is not investment advice recommending the purchase, sale, or holding of any specific security. The companies mentioned (SK Hynix, Samsung Electronics, Micron, etc.) are analytical examples, and the author may hold some of them. The 7/2 close and moves may differ slightly on an intraday or reporting basis. Target prices and HBM price forecasts are Street estimates as of the writing date (July 2, 2026) and change frequently. Scenario price zones and probabilities are subjective estimates based on support levels, cycle mechanics, and precedent. All investing carries the risk of principal loss, and investment decisions and their consequences rest with the investor.

References

Today's market and flows

• Samsung, SK Hynix plunge at the open on US chip shock · Etoday
• Samsung and Hynix tumble the moment KOSPI hits 8000, foreign and institutional selling · Hankook Ilbo
• Foreigners sell SK Hynix and Samsung as FX spikes · Money Today
• Foreigners net-sell 150T won on KOSPI in H1 · Seoul Economic Daily

FX and ADR

• SK hynix ADR listing seen as potential support for won · Korea Times
• SK Hynix plans $29B Nasdaq ADR listing as soon as July 10 · CNBC
• SK Hynix Files for Nasdaq Listing, up to $29B ADR (17.79M shares) · Roic News
• Cheaper abroad, pricier at home: Hynix stock-split talk ahead of the ADR · Daum News

Capex

• Samsung and Hynix, a market black hole soaking up cash with 4,700T investment · Etoday

Target prices and outlook

• SK Hynix targets all over the map (Mirae/Eugene 3.2M vs. Kiwoom 1.9M/BNK 1.85M) · Opinion News
• Nomura lifts Samsung and Hynix targets · Money Today
• On the Hynix plunge, "a chance to add," only sentiment changed · Financial News

Single-stock leverage and margin

• Samsung AM launches 2x ETFs tracking Samsung and SK Hynix · Hankyung
• KODEX SK Hynix Single-Stock Leverage ETF (net assets, trading) · Samsung Asset Management
• Single-stock leverage ranks #1 in retail net buying, trading value, and net assets · Daum News
• How leveraged ETF daily rebalancing and volatility decay work · Investor.gov (SEC)

Data was cross-checked against local daily/weekly data for SK Hynix and Samsung Electronics and the KOSPI index (7/2 close of 7,830, etc.).