Why KOSPI Crashed Twice, Then Surged 17.91% in a Day

Seoul’s stock market did not have a normal week between July 27 and July 31. It had five, compressed into five trading days: a small gain, a historic crash, a second consecutive crash, a smaller slide, and then the largest single-day rally in the exchange’s history. If you only read one number from that week, make it the last one — but the whole sequence is the story, so here’s all of it.

Key takeaways

• KOSPI closed Friday, July 31 at 6,595.45, up 17.91% (1,001.89 points) — the largest one-day percentage and point gain on record, recovering most of a three-day, 1,162-point rout.
• Foreign investors flipped from four straight sessions of selling (a combined ₩11.90 trillion, about $8.36 billion) to a single-day record of ₩8.79 trillion (about $6.17 billion) in net buying on Friday, concentrated almost entirely in two stocks.
• The trigger was offshore, not domestic: relief in the U.S. over AI infrastructure spending jumped straight into Korea’s two most AI-exposed chipmakers.
• Regulators used the same week to tighten the rules on the leveraged products that helped inflate — and then unwind — this run.

All figures below are drawn from the factsheet accompanying this brief, each with a primary or wire-service source. Where a claim is causal rather than a plain number, it carries an explicit hedge.

1. The number that mattered

17.91%. That’s how much KOSPI gained on Friday, July 31, closing at 6,595.45 — a jump of 1,001.89 points. It is also the largest single-day percentage and point gain in the index’s history. It came after three brutal sessions (July 28–30) had erased 1,162.19 points combined; Friday clawed back roughly 86% of that in one day.

Zoom out and July was still KOSPI’s worst calendar month on record. The index fell from 8,476.48 at the end of June to 6,595.45 at the end of July — a 22.19% monthly decline. Friday’s rally didn’t undo the month. It just made the final week of the month survivable. For why Korean equities carry a structural valuation gap to begin with, see how to read the Korea discount.

2. Foreign flows

Foreign investors — who can access the market through several routes — spent four straight sessions (July 24–29) selling KOSPI shares, a cumulative ₩11.90 trillion (about $8.36 billion at Friday’s close of ₩1,424/$1). The daily breakdown: ₩4.50 trillion sold on the 28th, ₩1.25 trillion sold on the 29th, then a reversal to ₩1.33 trillion bought on the 30th.

Friday broke the pattern entirely. Foreign investors bought a confirmed ₩8.79 trillion (about $6.17 billion) — a single-day record. Of that, ₩6 trillion landed in exactly two names: SK Hynix (₩3.61 trillion, about $2.54 billion) and Samsung Electronics (₩2.12 trillion, about $1.49 billion). This wasn’t diversified re-entry. It was a bet on two stocks.

3. What moved and why

Semiconductors. SK Hynix hit its daily upper trading limit, +29.95%. Samsung Electronics gained 26.81%, its own largest one-day move on record. The rebound is attributed to Microsoft’s earnings, where Azure revenue grew 43% on leaner AI-infrastructure spending. For chip investors, that read as evidence the AI capex slowdown feared earlier in the month wasn’t materializing.

KOSDAQ, more volatile than KOSPI itself. The junior index closed Friday at 719.76, up 11.63% — the second-largest daily percentage gain in KOSDAQ’s history, behind only a single day in October 2008. One local analyst is quoted describing non-semiconductor stocks as feeling “alienation” during the rebound. That reads as a signal that Friday’s buying stayed narrow rather than broad.

Leveraged ETFs got a regulatory haircut. Effective July 31, Korea’s Financial Services Commission tripled the basic deposit requirement for single-stock leveraged ETFs and ETNs, from ₩10 million (about $7,000) to ₩30 million (about $21,100). It also made the full amount cash-only, closing a loophole that had let up to 70% be posted in securities instead, and suspended new single-stock product listings pending stabilization. This is widely read as a response to the margin-call cascade behind the prior week’s crash. The regulator’s own release does not frame it that explicitly, though — the causal link here is inference, not confirmed fact.

Exports quietly backed up the chip story. Korea’s July trade data, released Saturday, August 1, showed exports of $98.89 billion (+62.8% year-on-year) and semiconductor exports specifically at $41.0 billion (+178.8% year-on-year). That produced a $30.32 billion trade surplus. Commentary frames this as underlying support for the memory-chip rally narrative. Worth noting, though: the export print reflects shipments already in the pipeline, not a forecast of where the stock rally goes next.

4. Lost in translation

English-language coverage of that week leaned hard on two narratives: foreign-investor whiplash and retail margin calls. Buried in the Korean financial press, and largely absent from the English wires, is a third narrative: Korea’s National Pension Service and other pension funds did not sell into the crash.

From July 1 through the afternoon of July 28, pension funds were net buyers of KOSPI shares. For the period, that is only about ₩15 billion (roughly $10.5 million) — a small number, but a positive one. On July 28 itself, the day of the 10.84% crash, pension funds were still net buyers, to the tune of roughly ₩23.5 billion (about $16.5 million). Korean commentary frames this as evidence the crash was a leverage and margin-call event rather than a broad institutional retreat. That’s a reasonable read, though still an interpretation — the underlying flow data doesn’t prove it on its own.

5. The week ahead

There is no Bank of Korea rate decision in the first half of August. The Monetary Policy Board’s next scheduled meeting is August 27. That’s worth flagging precisely because it’s easy to assume a week this volatile comes with a central-bank event attached. It does not.

Brokerages are using the volatility to reset their August targets rather than predict a direction: KB Securities is calling a 5,500–8,000 range, Samsung Securities 6,000–8,000. Read the width of those ranges as the real signal — nobody is claiming to know where this settles, and neither should you.

What is actually worth watching: whether the new leveraged-ETF deposit rule changes trading behavior now that it’s had time to bed in. Also worth watching: whether the July 31 foreign and pension buying was the start of a trend or just a one-day bet on two stocks. We could not verify a confirmed earnings cluster or statistics release for the first full week of August to a standard we’d put in front of you as fact. So none is claimed here.

---

This is general market information, not financial or investment advice. It does not constitute a recommendation to buy or sell any security. Figures are sourced as cited and current as of July 31–August 1, 2026; markets and rules referenced here can change without notice. Last updated: August 7, 2026.