Markets Featured

How the World's Hottest Stock Market Lost a Third of Its Value in Five Weeks

South Korea's Kospi just posted the worst month in its history and the fourth-worst day ever. The crash is the first leveraged stress test of the global AI trade - and the same fault lines run under the US market.

Alpha One

28 Jul 2026 • 7 min read

At 11:47 a.m. in Seoul on Tuesday, the Korea Exchange did something it has now done eight times this year: it stopped trading entirely. The Kospi was down more than 8%, sirens-and-flashing-red territory that triggers a marketwide circuit breaker and a mandatory 20-minute halt. When trading resumed, the selling barely slowed. South Korea's benchmark index closed down 10.84% at 6,023.66 — the fourth-largest single-day percentage decline in its history, worse than any day of the 2008 financial crisis.

The two stocks that did the damage tell you everything about what this crash actually is. Samsung Electronics fell 14.4%, its worst day since October 2008. SK hynix fell 14.7%, and its American depositary shares closed at $143 — below the $149 price of its $26.5 billion New York listing just eighteen days earlier, the largest first-time US share sale by a foreign company on record.

Five weeks ago, the Kospi was the best-performing major stock market in the world. It peaked at 9,114.55 on June 22. It has now given back 34% — including a 29% decline in July alone, which surpasses October 1997, the depth of the Asian financial crisis, as the worst month in the index's history.

Here is the uncomfortable part: even after all that, the Kospi is still up 43% for the year. That single statistic captures both how extreme the run-up was and why the unwind is not necessarily finished.

The Purest Expression of the AI Trade

To understand why Korea crashed first and hardest, you have to understand what the Korean market had become: the most concentrated, most leveraged vehicle for the global AI trade anywhere on earth.

The logic of the rally was sound, which is what made it dangerous. Samsung and SK hynix dominate the global market for memory chips — and high-bandwidth memory (HBM), the specialized stacked DRAM bolted onto every Nvidia accelerator, became the scarcest commodity in the AI supply chain. As hyperscalers committed hundreds of billions of dollars to data centers, memory prices ripped, contracts stretched years forward, and the two Korean giants sat at the choke point. Investors who wanted AI exposure without paying Nvidia's multiple bought Seoul.

They bought it with borrowed money. By late June, Korean retail investors' outstanding margin debt hit a record 38.48 trillion won (about $28 billion). Margin loans tied to Samsung and SK hynix alone reached 9.1 trillion won — more than triple the 2.53 trillion won at the end of last year. Brokerages tried to lean against the wind, raising margin requirements and restricting credit-loan purchases on some names. It didn't matter. As one Seoul daily put it, the mood among retail traders was simply: "I want to make some money too."

Concentration did the rest. Samsung and SK hynix together account for more than half of the Kospi's weighting. That made the index a two-stock bet wearing a 917-stock costume. On Tuesday, only 36 of those 917 stocks finished higher.

Briefings like this land in members' inboxes before the market prices them in. Join free →

What Actually Broke It

Three shocks landed in sequence, each aimed at a different pillar of the memory thesis.

First, the demand question. Last week's US mega-cap earnings reopened the debate over whether AI revenues can justify AI spending. Alphabet and Tesla both sold off on capital-expenditure concerns, semiconductors shed over a trillion dollars in market value in the days that followed, and the Philadelphia Semiconductor Index carried the weakness into this week — Nvidia itself dropped roughly 5% on Monday. If hyperscaler capex decelerates, the memory order book is the first thing that gets repriced, because it is the most commoditized layer of the stack.

Second, the supply question. On Monday, ChangXin Memory Technologies — China's state-championed DRAM maker — surged 466% on its Shanghai debut. A Chinese memory producer commanding that kind of valuation, backed by that kind of capital, is a direct long-term threat to the duopoly economics that justified Samsung's and SK hynix's re-rating. Memory is a cyclical commodity business that has been priced, for the past year, like a structural growth business. Chinese capacity is how memory cycles have died before.

Third, the tools question. Reports that China has begun mass production of homegrown deep-ultraviolet lithography equipment — still unverified, but widely circulated — hit the market's deepest assumption: that export controls would keep Chinese chipmakers a generation behind. If China can equip its own fabs at scale, the moat around the entire non-Chinese semiconductor complex narrows.

Any one of these was survivable. All three, landing on a market where retail leverage had tripled in six months, produced a liquidation cascade. Implied volatility had already been pointing to daily index swings of roughly 4.5% — a level at which margin calls feed on themselves.

Who Sold, Who Bought

The flow data from Tuesday is stark. Foreign investors dumped a net 5 trillion won (roughly $3.4 billion) of Korean shares — institutional money heading for the exit. Retail investors bought 4 trillion won — individuals catching the falling knife, many of them adding to already-leveraged positions.

That pattern — professionals distributing, households absorbing — is one of the most reliable late-cycle signals in market history, and it is now happening at scale in the world's most AI-exposed market. The won has weakened to around 1,470 to the dollar, its lowest levels of the year, compounding losses for any foreign investor who stayed. South Korea's Financial Services Commission is reportedly weighing tighter leverage controls — the regulatory equivalent of closing the barn door.

For readers who have followed our coverage: on July 3 we asked why the won was sliding while the Kospi made record highs, and argued the currency was signaling risk the equity market refused to price. Five weeks later, the equity market has repriced. The won was right.

Why This Is Not Just Korea's Problem

It is tempting to file this under "emerging market excess" and move on. That would be a mistake, because every structural feature of the Korean crash exists, in milder form, in the US market.

Concentration? Samsung and SK hynix are half the Kospi — but the top ten stocks are roughly 40% of the S&P 500, and the AI complex dominates that list. Leverage? Korean margin debt hit records — but US margin balances and zero-day options volumes have been setting records too. A top-ticking IPO? SK hynix listed $26.5 billion in New York within two weeks of the peak — the same week SpaceX went public; SpaceX now trades more than 50% below its post-IPO high. Narrative fragility? The entire global memory rally rested on the same hyperscaler capex assumptions now being questioned on US earnings calls — this week, with four mega-caps reporting and the Federal Reserve deciding rates against 4%-plus inflation.

Korea is not an anomaly. It is the leveraged, concentrated, fast-forward version of the same trade everyone owns. It broke first because it was built to break first — a two-stock index, a tripling of margin debt, and a retail base buying leverage into the top. The question American investors should be asking is not "what's wrong with Seoul?" It is "how much of this describes my own market?"

The Bottom Line

A 34% drawdown in five weeks, the worst month in the index's history, and a still-positive year-to-date return can all be true at once — that is what the unwind of a genuine mania looks like. The Kospi crash does not prove AI is a bubble. It proves the AI trade, wherever it is most concentrated and most leveraged, has stopped being priced for perfection — and it shows exactly what the exit looks like when everyone heads for it at the same time. Watch the memory spot market, watch this week's hyperscaler capex guidance, and watch who is buying the dip. In Seoul, it is the customers, not the professionals. That is rarely how bottoms are made.

If this analysis was useful, this is what AlphaBriefing does every day — geopolitics and markets, connected to what it means for your money. Free members get the daily brief in their inbox; paid members get the investment frameworks, scenario pricing, and catalyst calendars behind the paywall.

Get this level of intelligence every day. Subscribe to AlphaBriefing — free, member, and paid tiers available.

Sources & Further Reading

Disclaimer

AlphaBriefing is an independent intelligence publication. The content in this article is produced for informational and educational purposes only. Nothing published by AlphaBriefing constitutes financial, investment, legal, tax, or regulatory advice, nor should it be construed as a solicitation or recommendation to buy, sell, or hold any security, asset, or financial instrument.

All views expressed are those of the author at the time of writing and are subject to change without notice. Markets are volatile and unpredictable; past performance is not indicative of future results. Any investment involves risk, including the possible loss of principal.

AlphaBriefing and its principals, employees, or contributors may hold positions in securities or assets mentioned in this article. This should be considered a potential conflict of interest. No material relationship with any company referenced exists unless explicitly disclosed. Readers should conduct their own due diligence and consult qualified financial, legal, and tax advisors before making any investment decisions.

Information in this article is drawn from public sources believed to be reliable at the time of publication. AlphaBriefing makes no warranty, express or implied, as to the accuracy, completeness, or timeliness of any information herein. AlphaBriefing accepts no liability for any loss or damage arising from reliance on this content.