The First Crack in the AI Bubble: Why South Korea's Near-Collapse Should Worry the Whole World
The Scare Didn’t Come From Where I Expected
I’ve been following this AI race long enough to be suspicious of any “the bubble just popped” headline. I’ve seen that false alarm go off dozens of times. But when I read what happened in South Korea, I stopped scrolling.
It wasn’t Wall Street. It wasn’t Silicon Valley. It happened on the other side of the planet, in Seoul, and it moved faster and hit harder than the 1997 and 2008 crises — two dates any South Korean economist still treats as a national trauma.
And the reason for the collapse is almost embarrassingly simple to explain: two companies, absurd capital concentration, and the invisible “fuel” behind this entire AI party.
Two Companies Carrying an Entire Country
I had to look at the numbers twice to believe them. Samsung Electronics and SK Hynix together make up more than 51% of the entire KOSPI index. Samsung alone is about 33.5%. SK Hynix, 18.4%.
In other words: more than half of South Korea’s stock market rides on the performance of two manufacturers of high-bandwidth memory (HBM) chips — the exact components sitting inside practically every AI server and GPU built in the world today.
It’s not a metaphor to say these two companies “carry AI on their backs.” That’s literally the structure of the index.
What unsettled me the most, though, was thinking about who was on the other side of that bet: millions of South Korean retail investors, deep in margin debt, betting everything they had on these two stocks because the hype seemed too good to fail.
How a Small Wobble Turns Into an Avalanche
The trigger was almost insultingly small. A minor dip in US semiconductor stocks, right before quarterly earnings, and that was it — the domino started falling:
- Samsung and SK Hynix dropped more than 9% in a single session.
- Since much of the market was leveraged with borrowed money, the drop triggered automatic margin calls.
- Brokerages started liquidating positions on their own to cover the debt.
- That pushed the entire index down more than 10% in a single day.
Within three weeks, more than 1.2 million accounts were hit with margin calls in South Korea. Between 320,000 and 360,000 accounts were fully liquidated by brokerages.
I try to imagine what it’s like getting that call from your broker, finding out your positions are gone, and I can’t think of anything more brutal than losing everything because of a “minor wobble” happening on the other side of the world.
Why This Should Keep People in the US Up at Night
What actually worries me isn’t Korea itself. It’s realizing the same risk structure that caused this panic exists — at an even bigger scale — in the United States.
| What to watch | South Korea (KOSPI) | United States (S&P 500) |
|---|---|---|
| Retail leverage | Record-high margin debt | Highest level in history, above the dot-com bubble (2000) and 2008 |
| Index concentration | 2 companies = over 51% | 10 companies (the big techs) = over 33% |
| What could pull the trigger | A wobble in semiconductor earnings | Dependence on the CAPEX of 4 tech giants |
When I look at that table, what jumps out is that Korea wasn’t an isolated accident. It was an involuntary stress test — and the world just watched, live, what happens when a highly concentrated, highly leveraged market meets a tiny trigger.
The One Domino Holding Everything Up
At the end of the day, this whole trillion-dollar machine boils down to a question I keep asking myself: how long will Microsoft, Google, Amazon, and Meta keep raising AI infrastructure spending, quarter after quarter, without ever pausing to breathe?
Because that’s what holds the entire chain together. From Nvidia to TSMC, through Samsung and SK Hynix — everything depends on that continuous flow of CAPEX. The day a single big tech decides to cut or pause that spending, even for “just one adjustment quarter,” the first domino falls.
And Korea just showed us, with brutal clarity, how much damage that domino does when it lands on a market leveraged to the hilt.
What I Actually Think
I don’t think this means AI is a fraud or that the hype ends tomorrow. The companies behind this — Nvidia, Microsoft, Samsung itself — are real companies with real revenue. This isn’t an entirely empty bubble.
But what Korea showed is that the financial structure around AI is more fragile than the optimistic narrative lets on. When half of a national index depends on two companies, and much of that position is bought with borrowed money, you don’t need a black swan to break the system. A “minor wobble” is enough.
I keep wondering: if this happened in a country with a sophisticated, regulated market full of people who genuinely understand technology, what’s stopping the same script from playing out in the US, where the concentration looks different but is just as dangerous?
I’m Left With This Question
I don’t have a tidy answer, and I’m suspicious of anyone who claims they do. What I know is that South Korea just handed us an exact map of what this kind of collapse would look like at global scale — and nobody should ignore that map just because it came from a country that isn’t California.
Do you think global markets could actually absorb a real slowdown in AI spending, or are we watching a rehearsal for something much bigger? Let me know what you think:
- Email: fodra@fodra.com.br
- LinkedIn: linkedin.com/in/mauriciofodra
Two companies, 51% of an index, 1.2 million accounts hit in three weeks. Korea just showed the violence of a domino the US hasn’t tipped over yet.
Read Also
- Forced Funding: How Your Pension Fund Is Feeding the AI Hype (Whether You Like It or Not) — If you own an index fund, you’re already exposed to this same concentration risk, just on the American side of it.
- Surfing the Bubble: How Not to Fall Behind in AI’s Accelerated Evolution — I wrote about how to ride the hype without getting swallowed by it. Korea’s case is exactly the kind of wave that can take you down.
- The AI Paradox: Between Market Panic and Real Skills in 2026 — The market panics, but the technology keeps evolving underneath the panic. Both facts coexist, and that’s exactly what makes any of this hard to predict.