Two Stocks Are the Korean Stock Market Now

There were two crashes in Seoul on Friday, 26 June, and which one you heard about depends on which clock you read.

At 12:10 in the afternoon the KOSPIKOSPISouth Korea's main stock-market index, roughly the equivalent of the S&P 500 in the United States; it tracks the value of the country's largest listed companies., South Korea's main stock index, the rough equivalent of the S&P 500S&P 500A widely watched index of 500 large US companies, often used as shorthand for 'the US stock market.', had fallen 731.97 points, or 8.19 percent, to 8,198.33. That breached the exchange's automatic 8 percent threshold and tripped a circuit breakercircuit breakerAn automatic, temporary halt in trading that kicks in when an index falls past a set threshold (here 8 percent), giving a panicking market a 20-minute pause to calm down., a mandatory 20-minute trading halt designed to let a panicking market catch its breath. That is the number that went around the world: Korea down 8 percent.

Then the market reopened, steadied, and settled. The official Korea Exchange close was 8,411.21, down 519.09 points, or 5.81 percent.

One number measures the panic; the other measures the settlement. The distance between them is itself a symptom of the thing we are about to measure.

Neither figure is wrong. Everything that follows is locked to the official close, because you cannot decompose a number that keeps moving. But hold on to the gap between 8.19 and 5.81.

Not 'why did Korea crash.' That is too big to answer honestly in one sitting, and half the answers would be guesses. Instead, a narrow, testable question:

This is answerable with grade-school arithmetic, which is the point. You do not need a model. You need two inputs per stock.

An index like the KOSPI is a weighted average. Every company counts in proportion to its market value, its 'weight.' Samsung is worth far more than a mid-size shipbuilder, so Samsung moves the index far more. The contribution of any single stock to the index's move is just its weight multiplied by its return that day.

A stock that is 25 percent of the index and falls 8 percent drags the whole index down by 2 percentage points, all by itself. That is the entire toolkit. Now plug in Friday's official figures.

By 19 June, the most recent precise reading from the Korea Exchange's market-cap tables (reported by Businesskorea), Samsung Electronics was 28.34 percent of the KOSPI and SK Hynix was 26.42 percent. Together, 54.76 percent. Two companies, more than half the index. On Friday Samsung closed down 5.30 percent (358,500 wonwonSouth Korea's currency; one trillion won was roughly 700 million US dollars in mid-2026. to 339,500) and SK Hynix down 8.36 percent.

• Samsung: 28.34% x (-5.30%) = -1.50 points of index, out of every 100
• SK Hynix: 26.42% x (-8.36%) = -2.21 points of index
• Together: -3.71 of the index's 5.81-point fall

Two stocks produced about 64 percent of the entire decline.

Read it slowly, because it is counterintuitive. The 830-odd other companies on the KOSPI, banks, carmakers, shipbuilders, biotech, retail, steel, every name that is not a memory-chip maker, lost 188 points between them. Two semiconductor stocks lost 331. The chip duo, fewer than 0.3 percent of the listed names, outweighed the other 99.7 percent combined.

It also means the '5.8 percent crash' overstated the day for almost everyone who owned almost anything else. Strip out the two chipmakers and the rest of the market fell, on a value-weightedvalue-weightedA way of averaging stock moves that gives bigger companies more influence, in proportion to their market value, rather than treating every company equally. basis, closer to 4.6 percent, and for the typical mid-cap name almost certainly milder still. The headline number was, to a first approximation, the price of memory chips wearing a national flag.

A claim worth printing has to be killable. Ours: the two chipmakers accounted for more than half of the point decline. The test: run weight-times-return against the official numbers, and if the duo comes in under 50 percent, the claim is dead.

It does not come in under 50 percent under any reasonable assumption. Use the older, lower weights from mid-May, when Samsung and SK Hynix were 28 and 23.5 percent of the KOSPI 200, which understates SK Hynix, the stock that fell hardest, and you still get roughly 59 percent.

The finding is robust precisely because the concentration is so extreme that the arithmetic has no room to escape it. And if anything it is conservative: some data vendors put Friday's per-stock falls steeper still (Trading Economics has Samsung at -6.28 percent and SK Hynix at -8.95 percent), which would push the duo's share toward 70 percent. We use the lower, prior-close-consistent figures so the claim rests on the weaker case.

Diversification on the KOSPI is, for practical purposes, an illusion: you are holding a memory-chip bet with a 45 percent tracking error.

Here a careful reporter has to separate two things that look identical and are not: market reaction (who is buying and selling, and how the plumbing routes their orders) versus economic effect (whether the world actually needs fewer memory chips). Three non-fundamental triggers landed on Seoul in the same week, and none is about chip demand.

1. MSCIMSCIA company that builds and maintains widely used global stock indexes; its decisions on which 'bucket' a country belongs in can move billions of dollars of fund money. said no, again. On 23 June (released 24 June Korea time) the index provider MSCI kept Korea in its 'emerging market' bucket and off the watchlist for promotion to 'developed' status, a designation that would have forced waves of automatic buying by global passive fundspassive fundsInvestment funds that simply track an index rather than picking stocks, so they mechanically buy or sell whatever the index tells them to.. Foreigners had front-runfront-runBuying in advance of an expected wave of other buyers (here, foreign funds betting on an MSCI upgrade) so as to profit when that buying arrives; if it never comes, the early bet unwinds. that hoped-for upgrade; when it did not come, the trade unwound. MSCI's sticking point was mundane plumbing: the won still cannot be freely settled offshore (it trades mostly through non-deliverable forwards), which makes the market hard for global funds to rebalance. Seoul is opening 24-hour onshore FX trading in July, but MSCI wants to see it work first.

2. A brand-new leverage machine. On 27 May Korea launched its first single-stock leveraged ETFs, 16 products engineered to move twice as much as Samsung or SK Hynix each day. They drew about 28 trillion won in three days. By late June these products accounted for roughly 31 percent of Samsung's daily trading volume and 38 percent of SK Hynix's. The country's top financial regulator publicly said he regretted not blocking them. Leverage is mechanical: when both stocks dropped near 13 percent on 23 June, the 2x ETFs had to dump about 6 billion dollars of shares in a single session just to reset their ratios. That is forced selling feeding on itself, not a verdict on AI.

3. Quarter-end, margined retail. Late June is when leveraged retail accounts get squeezed and forced to sell, regardless of what anyone thinks a stock is worth.

So the cleanest reading is uncomfortable for the panic narrative: the economic effect (chip demand) looks fine; the market reaction (positioning, leverage, index mechanics) did the damage. But notice what that does to our question. It does not refute the concentration finding, it sharpens it. If the selloff was plumbing rather than fundamentals, then the 55 percent chip weight is not the cause. It is the amplifier: the thing that turned an ETF-and-flows squeeze in two stocks into a national circuit breaker.

Explanation one: a fundamental AI wobble. The simplest story is that investors are repricing AI itself, if data-center spending slows, memory demand softens, and the two most AI-exposed stocks on earth get hit first. Support: the selloff clustered in AI-memory names, came amid a global tech pullback (the Nasdaq fell sharply on adjacent sessions), and the magnitude looks like fear about a forecast, not a technical blip.

Explanation two: a flows-and-plumbing squeeze wearing a chip costume. The competing read says nothing about chip demand changed; what changed was positioning. The MSCI non-upgrade unwound a front-run foreign trade, brand-new 2x single-stock ETFs forced mechanical selling, and quarter-endquarter-endThe close of a three-month accounting period, when funds and leveraged traders often adjust or unwind positions, sometimes adding selling pressure regardless of fundamentals. margin calls piled on. Support: Micron hit a record while SK Hynix cratered, HBM is contractually sold out for 2026, and the leveraged ETFs alone now move a third of the volume in both stocks. Under this reading the fundamentals are intact and the concentration is the amplifier.

The evidence leans toward the second, but the two are not mutually exclusive: fear about AI may have lit the match, while leverage and concentration poured on the fuel. Either way, the concentration finding stands, it is the transmission mechanism in both stories.

We have seen a national index become a single company before.

Nokia, Helsinki, 2000. At the peak of the telecom bubble Nokia was about 70 percent of the Helsinki exchange's market value. 'The Finnish stock market fell' and 'Nokia fell' were the same sentence, and when the bubble deflated the index spent years as a derivative of one company's fortunes. Korea's 55 percent is not yet Nokia's 70 percent, but it is the same structure: a national savings vehicle quietly reorganized around one theme.

Taiwan, now. The live cousin is the TAIEXTAIEXTaiwan's main stock-market index, heavily dominated by the chip foundry TSMC., where the chip foundry TSMCTSMCTaiwan Semiconductor Manufacturing Company, the world's largest contract chipmaker, which alone makes up a large share of Taiwan's stock index. alone runs around 40 to 44 percent. Korea is the more extreme case because its concentration is split across a duo that can both fall at once, which is exactly what happened on Friday.

Where the analogy breaks: Nokia was one company in one product cycle, and Finland was a small economy. Korea's duo sits at the center of the global AI supply chain with contracted, sold-out demand, so a fundamental collapse is harder to engineer than a positioning shake-out. The risk is less '1990s Nokia bust' and more that a real downturn, whenever it comes, would be amplified far beyond what a diversified index would feel.

The KOSPI roughly doubled in 2026 to become one of the hottest major markets on earth, it first cleared 9,000 on 18 June, only to give much of it back in a brutal final fortnight. As it rose, foreigners left in record size: net selling crossed 103 trillion won this year, more than the 62 trillion of the 2008 financial crisis and the 25 trillion of the 2020 pandemic combined.

What kept the index aloft was domestic. Korean households poured an estimated 70 billion dollars in, buying what foreigners sold.

The index is 55 percent two chipmakers. And the marginal buyer holding it up is the Korean saver.

Put the two findings together. Ordinary households have, mostly without choosing to, concentrated their wealth into a single wager on AI memory pricing, catching the falling knife from the foreign institutions who built the position and are now handing it back. Anyone holding a plain KOSPI index fund, pensions, retail accounts, passive savers, is far more exposed to two semiconductor stocks than the word 'index' suggests.

Base case (most likely): a choppy, range-bound stabilization. As the leveraged-ETF deleveragingdeleveragingReducing borrowed positions, often by selling assets; when many investors do it at once it can feed a self-reinforcing drop. completes and the regulator curbs the new products, forced selling fades while sold-out 2026 HBM contracts keep fundamentals firm. What would make this likelier: order books staying full (Micron and SK Hynix guidance intact) and no fresh foreign-outflow record.

Upside case: foreigners return. If Seoul's July launch of 24-hour onshore FX trading works and the won becomes easier to settle, the MSCI developed-market path reopens, reviving the passive-inflow trade foreigners had front-run. What would make this likelier: smooth FX reform, a concrete MSCI watchlist signal, and foreign net buying turning positive.

Downside case: the concentration bites for real. If AI capital spending actually slows and HBM contracts get renegotiated, the 55 percent chip weight would amplify a genuine fundamental decline into a deeper crash, with more circuit breakers and household losses concentrated in two names. What would make this likelier: any HBM order cancellation, a Micron or SK Hynix guidance cut, or another record week of foreign selling.

• The official close is the anchor. If a clean KRX decomposition put the chip duo under 50 percent of the point drop, the central claim fails. It does not, even on conservative weights.
• Intraday vs close. Measured against the 8.19 percent trough rather than the 5.81 percent close, the absolute contributions are larger but the proportions are similar; the clock you choose changes the drama, not the finding.
• Day-level flows disagree by source. The Korea Herald's intraday tally (as of 2:42pm) showed individuals net buying 3.77 trillion won against foreigners selling 3.11 trillion and institutions 0.73 trillion; other tallies run higher for the full session. We report the direction, heavy foreign selling, heavier retail buying, not a single contested figure.

Every wire framed Friday as 'Korea -8%, AI-bubble fears.' But the economic signal (HBM demand) is up, Micron at a record, 2026 capacity sold out. What actually fell was not a forecast about chips; it was the unwinding of a foreign upgrade-trade, magnified by month-old 2x ETFs, through an index where two stocks are 55 percent of the weight.

When the index is this concentrated, 'the stock market crashed' and 'memory chips had a bad day' stop being two events that happen to coincide. On 26 June they were the same event, told twice.