This week a company announced the most profitable quarter in its history — and within minutes, trading stopped on the whole national market. It was the second day in a row South Korea's market had to be halted. That has never happened before. On that market 917 companies are traded; to stop all of them, it took two.

The index had risen ~300% in about a year, then gave a third back in a month — and on the way down the emergency brake fired again and again. Nobody broke a law. Three reasonable-sounding rules did this: how you weigh an index, who may borrow for a bet, and when a falling market is paused.

Tuesday, 28 July: the KOSPI falls almost 11% in a single day; Samsung and SK Hynix each drop more than 14% — for Samsung, the worst day since 2008. Of 917 main-board stocks, 36 end higher. Halfway down, the automatic brake kicks in: index 8% down held for one minute → all trading stops for 20 minutes. This year alone, that rule has fired eight times.

Wednesday, 29 July: SK Hynix publishes the most profitable quarter in its history — the market expected more. The selling resumes, and the brake fires a second consecutive day: a first in the exchange's history. Two days, −18%; July on track for the worst month ever recorded — after a year as, by most counts, the best-performing major market on Earth. By Wednesday afternoon the finance minister had summoned the country's top financial authorities; in parliament, he apologised. What exactly for — that is the story.

Rule one: weigh every company by its size, with no upper limit. Watch what that does in a boom: the AI build-out made memory chips precious, and Korea makes the world's memory. Two years ago Samsung and SK Hynix were ~40% of the index. By May, 47. Mid-June, 55. End of June, 60% — two companies, six-tenths of a 900-company market. On the Nasdaq the top two are ~20%; on Japan's index they don't reach 10. Nobody decided this: the weighting rule converted the rally into dominance automatically. Success concentrates by design.

Concentration is fragility waiting for a direction change: two stocks at 60% falling 14% move the index more than 8% — the brake threshold crossed by two stocks alone. And one twist from outside: analysts warned that one more percentage point of weight would force foreign funds bound by American diversification rules to sell about $2 billion of Korean stock. A cliff in Seoul, written in Washington.

This spring Korea's regulators approved a retail product delivering twice the daily move of a single stock. It launched on 27 May — near the very top. Retail investors poured in about $10 billion in two months — five times the foreign inflow; borrowed money market-wide hit an all-time record. The mechanical part matters: a fund promising double the daily move must rebalance every day — forced to sell after every fall. By construction, it sells into the fall.

When chips turned, the machine ran in reverse: the 2× SK Hynix fund lost more than 80% from its June peak, the Samsung version 75%; brokers selling out accounts to recover loans unloaded roughly $1.7 billion of stock. In parliament a lawmaker said the country had turned into a casino; the minister said he was sorry — for introducing the product without careful consideration. A rare thing: the author of a rule naming the rule as the defect. The apology unwinds no one's losses.

Korea's ladder is precise: 8% down held a minute — everything pauses 20 minutes; 15% — again; 20% — the day is over. Brakes like this were designed after past crashes for a specific enemy: panic, a wrong price — cascades of automatic sell orders, a mistyped trade, a rumour. Twenty minutes of forced calm lets reality catch up.

But that does not look like Korea this week. This looks like the repricing of a belief — what a year of AI spending will actually earn — and a pause does not change what memory chips will earn. So the brake fires again: nine times in a year, twice in two days. Worse, a visible threshold creates its own gravity — the documented magnet effect: as the index approaches the trigger, selling accelerates, everyone out before the doors close. A brake that fires once a decade is an emergency device. Nine times a year, it is a metronome — setting the fall's rhythm in eight-percent slices.

Cap the weight. Index providers sell capped versions — no company above a fixed share. Done live: in 2023 the Nasdaq-100's giants outgrew its limits and the exchange stepped in and trimmed them, mid-year, by rule. The cost: Korea's economy genuinely is this concentrated — a capped index stops telling the truth; and every tracking fund must sell the national champions precisely because they succeeded. A tax on winning.

Gate the leverage. Europe, after its own wave of retail losses, capped leverage and banned the worst products; Korea is weighing position caps, lower ratios, professionals-only. The cost: the state decides who may risk their own money; demand migrates to offshore brokers and crypto copies with no protections at all — one such copy had already broken within days; and closing the counter reads as the elite pulling up the ladder. That is politics.

Redesign the pause. End halts with a reopening auction — one price set by everyone's collected orders, not a stampede; or one-way limits per step, America's redesign after the 2010 flash crash. The cost is written in the most famous experiment on this question: China's tight market-wide breaker of January 2016, scrapped in four trading days as the magnet effect made falls faster. Longer → a taller wall of waiting orders; rarer → why exist; tighter → a magnet. No neutral setting: the dial only chooses who panics, and when.

The most profitable quarter in a company's history helped stop a nation's stock market — twice. There is no villain to arrest: weighing by size is fair, investing with conviction is freedom, pausing a crash is prudence. The sum is a machine: success concentrates, concentration invites leverage, leverage cascades, and the cascade slams into a brake never designed for it. The minister apologised for one of the three dials; the other two are still set exactly where they were on Tuesday morning.

And this is not a Korean machine — every stock market on Earth runs on the same three dials: how success is weighted, how much conviction may borrow, when falling is paused. So when a market next halts, any market, any country, carry this question in: which of the three dials actually did it? The answer is almost never the one in the headline.