The Encore

In the autumn of 2012, a Korean semiconductor-testing company with no connection whatsoever to the music industry saw its shares spike in a matter of weeks — a rally Bloomberg reckoned at the time at nearly 500% from the lows, as the stock leapt from around 1,500 won to over 3,600. The company was DI Corporation. The reason was not a product, an earnings beat, or a contract. The reason was that its executive chairman, Park Won-ho, happened to be the father of a pop musician named Park Jae-sang, better known as Psy, whose video "Gangnam Style" had just become the first thing in human history to be watched a billion times on the internet. The market did the arithmetic that markets do at moments like that — the arithmetic of association rather than fundamentals — and bid a chip-tester's stock up severalfold because a man's son had made a funny video about a Seoul neighborhood. Its market capitalization ballooned from tens of millions of dollars into the hundreds of millions on no information about the business at all.

Then, as these things do, it came back down.

Fourteen years later, in June 2026, DI Corporation is once again one of the more talked-about tickers on the Korean exchange. The hook this time is not K-pop. It is artificial intelligence: through a subsidiary called Digital Frontier, DI makes the machines that test high-bandwidth memory wafers before they are sliced into the chips that feed AI accelerators — a real, if modest, link in the most important supply chain on earth. The stock is up about 40% on the year. And the reason this small company is worth your attention is not the 40%. It is that the very same stock, the very same casino chip, has now been swept up in two completely unrelated manias, a decade and a half apart, for two completely unrelated reasons — and that tells you something true and unsettling about the market it trades in.

It tells you the market is, once again, pricing the association rather than the thing. Last time the association was a music video. This time it is a three-letter acronym. The mechanism is identical, and so, history suggests, is the ending.

The index that became a chip

To understand why a Gangnam Style footnote matters, you have to see what has happened to the entire Korean stock market, because DI Corp is not an anomaly within it. It is a miniature of it.

Over 2025 and into 2026, South Korea's benchmark KOSPI index did something no major market is supposed to do: it nearly doubled, becoming the best-performing major index in the world, up roughly 93% on the year and crossing 7,000 for the first time in May. The engine was two companies. Samsung Electronics, which surpassed a trillion-dollar valuation in early May, and SK Hynix, which raced toward the same milestone — the two dominant makers of the high-bandwidth memory that the AI boom cannot get enough of. Their stocks ran up on the order of 115% and 185% respectively. And between them, the two now make up more than half of the entire index.

Sit with that. An entire nation's flagship equity benchmark — the thing held inside the retirement funds and index products of millions of people who have never heard of HBM — has quietly become a leveraged bet on two memory-chip stocks riding a single narrative. The iShares MSCI South Korea ETF, the $21 billion vehicle through which most foreign investors own the country, now holds Samsung and SK Hynix at roughly 55% of its weight. To buy Korea is, increasingly, to buy two chipmakers and call it diversification.

When more than half of an index is two stocks telling one story, the index is no longer a market. It is a position. And positions, unlike markets, can be exited all at once.

The black hole

If the concentration were the whole problem, it would merely be dangerous. What turned it combustible was leverage, sold to the public at the precise top.

In late May 2026, Korea launched its first single-stock leveraged ETFs — products that deliver two times the daily move of one company, in this case Samsung and SK Hynix. They were, in effect, a way to make an already-concentrated bet twice as concentrated, marketed to retail investors at an entry price low enough to feel like pocket money. The appetite was ferocious to the point of alarming. The funds triggered volatility halts on their very first day. Within their first week they were absorbing more than ten trillion won of trading volume a day; within three days they had pulled in some 28 trillion won. The Korean financial press, not given to hyperbole, began calling them a "black hole" swallowing the market's capital — a vortex pulling an ever-larger share of the nation's savings into a doubled-up wager on the two stocks that were already holding up the index.

This is the part of every mania that looks, in hindsight, like the engraved invitation to the crash. New, easy, leveraged instruments arrive to let ordinary people amplify a bet that the smart money has already made — and their arrival marks not the beginning of the move but the end of it. Leveraged ETFs carry a quieter cruelty, too: because of the way they reset each day, a stretch of volatility bleeds them through "negative compounding," so that even an investor who is eventually proved right about the direction can be ground to nothing by the chop along the way. The retail buyers piling into 2x Samsung at the end of May were not buying the rally. They were buying the right to lose twice as fast when it turned.

It turned.

The week the music stopped

The trigger, when it came, was almost insultingly small relative to the damage it did — which is the signature of a market priced for perfection.

On the evening of June 3, the American chip designer Broadcom reported earnings and guided to roughly $16 billion of AI-chip sales for the coming quarter — about 7% below the $17.2 billion Wall Street had penciled in. Not a collapse. A miss. A single company, an ocean away, falling a few percent short of one quarter's estimate. In a market grounded in fundamentals, it would have been a footnote.

In Seoul it was a detonation. On June 8 the KOSPI — the world's best-performing index a week earlier — sank more than 8%, plunging hard enough to trigger a circuit breaker that halted trading and forced an emergency meeting at the exchange. Samsung fell about 7%; SK Hynix more than 9%. The iShares Korea ETF, which had been up an astonishing 83% on the year as recently as June 5, lost nearly 13% of its value in a single day. The leveraged products built to double the upside did exactly what their design promised in reverse, with the 2x funds shedding around 20%. The narrative had not been disproven; it had merely been nudged, and a market leaning its full weight on that narrative discovered there was nothing behind it but more people leaning the same way.

Steve Brice, the global chief investment officer at Standard Chartered, put the polite version to clients: peak optimism around Korean equities, he warned, was "not too far around the corner," and it was time to take profits and diversify. The blunt version is the one the price action wrote on June 8. When the only thing holding an index up is the belief that it will keep going up, the first crack in the belief is the whole story.

The tell, in a chip-tester

Which returns us, finally, to DI Corporation, and to why a company most investors will never knowingly own is the most honest indicator in this entire market.

DI is the control group. It is a small, real business with a small, real connection to the AI supply chain — but its price has now been moved twice, fourteen years apart, by forces that had nothing to do with that business. In 2012 the force was a music video and a family tie, and the stock ran up severalfold and came back down. In 2026 the force is a three-letter macro story, and the stock is up 40% in its slipstream. The fundamental questions — how many wafer testers Digital Frontier actually sells, at what margin, to a customer whose own stock just fell 9% in a day — are, as in 2012, very nearly beside the point. What moves the stock is the association. What associations do is reverse.

None of this is to say the AI memory boom is fake. The demand for high-bandwidth memory is real, Samsung and SK Hynix are extraordinary companies, and DI's testers genuinely sit in the supply chain. The bull case has substance the 2012 Gangnam Style bubble never did. But substance is not the same as price, and a market that has put more than half of its flagship index into two stocks, handed its citizens doubled-up leverage to chase them at the top, and then cratered 8% on a 7% earnings miss from a company in another hemisphere is not trading on substance. It is trading on a song it has heard before — and humming along as if it does not know how this verse ends.

The first time, the chorus was about Gangnam. This time it is about AI. But it is, unmistakably, the same song, played a little louder, with a great deal more borrowed money in the room. And the thing about a sequel is that everyone in the theater already knows how it ends. They have simply decided, for the length of one more chorus, to pretend they don't.