Two trillion dollars of maybe

In forty days, South Korea's stock market gave back about $2 trillion, most of the value it had spent the first half of the year accumulating. The Kospi, which had more than doubled in the first half of the year on AI memory-chip euphoria, touched a record 9,385 on 19 June, hit an intraday low of 5,263 forty days later (a peak-to-trough fall of 43.9 per cent) and closed at 5,593 on 30 July. SK Hynix, the maker of the high-bandwidth memory inside the world's AI accelerators, sits roughly 55 per cent below its June peak; Samsung Electronics roughly 44 per cent. The exchange tripped circuit breakers on consecutive days for the first time in its history. This is the moment the bubble-callers have been waiting for, and they are entitled to it. A book that scores the news against its claims owes them a straight answer when the news goes against it.

The claim on the table

The frontier-ledger claim (chapters 2, 6 and 9) says the AI build-out is history's largest private capital-formation event and that its gains are being banked as assets owned by the frontier, with no claim reserved for anyone else. The bubble objection, stated at full strength: if those gains are paper, the ledger is a ledger of nothing. A crash of this size says the market no longer believes the earnings the valuations assumed. It says the "largest capital formation event in history" may contain a large fraction of malinvestment. And it embarrasses the urgency premise: a citizen's stake built at the June peak would be down by nearly half, and the book's "before the gains are gone" would read as an invitation delivered at the top.

What the crash does not say

Look at what accompanied it. The selloff came amid doubts about the return on AI data-centre and chip capacity and, in Anadolu's telling, Chinese competition arriving in the memory market itself: the report names the market debut of ChangXin Memory, a Chinese memory-chip maker, among the proximate causes. That is not machine intelligence failing. That is more machine capacity arriving faster and cheaper than the incumbents priced. The supply shock accompanying the crash is the same force the book says is crashing the price of labour. The fabs did not un-build in July. The memory lines are still running. Cheaper chips make the automation of work more affordable, not less. The labour share does not recover because the ticker fell; workers were not in the June rally and are not made whole by the July collapse. A crash reprices who owns the future among owners. It moves nothing across the line between owners and everyone else.

Korea also shows what citizen exposure looks like when nobody designs it. The National Pension Service, the retirement claim of every Korean worker, is among Samsung's largest shareholders, with a stake reported at about 7.75 per cent in early 2026, so citizens were in this crash whether they knew it or not. At the other extreme, households held single-stock leveraged ETFs that double Samsung's and SK Hynix's daily moves, sold under the ETF label with none of an ETF's diversification, and critics lay their approval squarely at the government's door. The book's working examples (chapters 8 and 10) argue that design is the whole game: Norway's fund is built to ride a 40 per cent drawdown for decades; a two-times leveraged retail product is built to be destroyed by one. Korea has citizen exposure without citizen design, and July was the bill.

Verdict

Complicates. Honestly scored, this one lands on the book, not for it. The ledger claim assumed the gains were real enough to be worth pre-distributing, and a 43.9 per cent drawdown in the industry's heartland is genuine evidence that some of them were not. What survives is the part of the claim the crash cannot touch: the assets exist, the capacity keeps compounding and the question of who holds it is unchanged at the bottom of the chart. Whoever owns the machine in the trough owns the recovery, if it comes. The bubble-callers may be right about the price. The book's bet is that they are wrong about what the price was ever measuring.

Who owns the machine?