South Korea’s AI Bubble Just Popped — And America Could Be Next

South Korea’s AI Bubble Just Popped — And America Could Be Next

South Korea just gave investors a preview of what an AI-driven market crash could look like.

Just weeks ago, the KOSPI was one of the best-performing stock markets in the world, surging nearly 200% over 12 months. Samsung had gained more than 500%, while SK Hynix had climbed over 1,000%.

Then everything changed.

The KOSPI plunged roughly 25%. More than 1.2 million investor accounts hit margin-call thresholds, while over 3 trillion won worth of investments were forcibly liquidated. At one point, the sell-off became so severe that South Korea's government called an emergency meeting to stabilize the market.

But this wasn't simply a Korean stock-market correction.

It was a stress test of the exact system that is increasingly supporting the U.S. market: AI spending, concentrated stocks, and enormous amounts of leverage.

And Korea may have just shown us what happens when that system starts moving in reverse.

---

The Entire Korean Market Was Riding on Two Stocks

South Korea has around 51 million people, and approximately 14 million of them are retail investors.

These investors have increasingly turned toward stocks as a way to build wealth, particularly as housing has become increasingly unaffordable for younger Koreans.

But they weren't simply buying stocks.

Many were borrowing money to buy them.

Others were using leveraged ETFs, which magnify both gains and losses.

And much of that money flowed into two companies:

Samsung and SK Hynix.

Together, the two companies represented more than half of the KOSPI.

That meant South Korea effectively had one enormous trade:

AI spending → memory-chip demand → Samsung and SK Hynix → rising Korean stocks.

As long as the AI boom continued, the machine worked.

Until investors started questioning it.

---

Then America Triggered the Sell-Off

The initial shock came from the United States.

U.S. chip stocks began selling off ahead of earnings from Micron, one of America's most important memory-chip companies and a major competitor to Samsung and SK Hynix.

The concern wasn't necessarily that AI spending had already collapsed.

It was that it might slow down.

That was enough.

The KOSPI fell around 4.6% in one session.

A week later, Samsung and SK Hynix both dropped more than 9% in a single day.

Then came what Korean investors began calling Black Tuesday.

The KOSPI plunged more than 10% in one day.

And because so many investors were using leverage, the selling didn't simply stop when investors decided they had sold enough.

It became automatic.

---

This Is How a Margin-Call Doom Loop Works

Imagine you buy $100 worth of stock using $50 of your own money and $50 of borrowed money.

If the stock rises, your returns are amplified.

But if it falls far enough, your broker doesn't care whether you believe the company will recover.

You get a margin call.

Deposit more money — or the broker sells your shares.

Now imagine millions of investors doing the same thing at once.

Prices fall.

More investors hit their margin thresholds.

Their brokers sell.

Prices fall again.

More investors get liquidated.

And the cycle repeats.

That's the margin-call doom loop.

During Korea's crash, the percentage of margin accounts being forcibly liquidated reportedly climbed above 10%, compared with roughly 2% under normal conditions.

Leveraged ETFs made the situation even worse.

A 3× leveraged fund doesn't just magnify your gains.

If the underlying stock falls 10%, you can lose roughly 30%.

And because leveraged funds have to rebalance, they can be forced to sell into falling markets, adding even more downward pressure.

The result is a market where falling prices create more selling, and more selling creates even lower prices.

---

And This Is Where America Gets Interesting

Because the United States is running a much larger version of the same experiment.

The U.S. stock market is heavily concentrated.

The top 10 companies account for around 36% of the S&P 500.

And many of those companies are tied to the same investment theme:

AI.

Microsoft, Google, Amazon and Meta are spending hundreds of billions of dollars building AI infrastructure.

That spending flows into Nvidia, Micron, data-center operators and semiconductor manufacturers.

Their revenues then become evidence that the AI boom is working.

Which encourages even more spending.

And the cycle continues.

But there's a problem.

The entire system increasingly depends on investors believing that this spending will eventually generate enormous profits.

---

The U.S. Is More Leveraged Than Ever

According to the source data, U.S. margin debt reached roughly 4.5% of GDP in June 2026, the highest level recorded.

That's higher than the leverage levels seen around the dot-com bubble, the 2007 financial crisis and the 2021 market boom.

And official margin debt doesn't capture every form of leverage.

It doesn't fully account for things like leveraged ETFs, options, portfolio margin and other modern forms of borrowing.

In other words, the headline number may not represent the full amount of risk sitting inside the market.

And history is not particularly comforting.

Extreme margin debt has appeared before major market corrections, including around 1987, 2000 and 2007.

That doesn't mean another crash is guaranteed.

But it does mean the system has become increasingly sensitive to falling prices.

---

The Bigger Problem Is AI

The most important question isn't whether AI is real.

It is.

The question is whether investors are paying too much for the amount of economic value AI is actually producing.

The biggest technology companies are spending extraordinary amounts of money on AI infrastructure.

But the returns are already becoming harder to justify.

Jim Chanos pointed to a major decline in the returns hyperscalers are generating on incremental AI investment.

The argument is simple:

If companies were previously generating around 40 cents of operating income for every additional dollar invested, but that figure falls toward 20 cents and eventually 10 cents, eventually someone has to ask:

"Why are we spending another trillion dollars?"

That's the question that could eventually break the entire AI investment cycle.

---

The Most Important Number to Watch

The biggest warning sign may not be Nvidia's stock price.

It may not be Bitcoin.

It may not even be the KOSPI.

It could be one number:

AI capital expenditures at Microsoft, Google, Amazon and Meta.

These companies are effectively the engine behind the AI buildout.

If they keep increasing spending, the AI machine continues.

But imagine an earnings call where one of them says:

"We're reducing our AI capital expenditures."

That could change everything.

Because the market has spent years rewarding these companies for increasing AI spending.

If investors suddenly reward a company for cutting spending instead, every other CEO gets permission to do the same.

And once one hyperscaler cuts CapEx, others could follow.

---

We Have Seen This Movie Before

During the dot-com bubble, companies were spending enormous amounts building internet infrastructure.

Then the spending slowed.

Coca-Cola, for example, reportedly needed thousands of Cisco routers during the boom.

Later, its requirements fell dramatically.

For companies selling the infrastructure, that wasn't a small adjustment.

It was a collapse in demand.

Cisco's earnings suffered.

The Nasdaq eventually fell roughly 78% from its peak.

The same mechanism could happen with AI.

If Microsoft cuts AI spending by 20%, Microsoft itself may survive perfectly well.

But that 20% reduction could become a much bigger problem for Nvidia, Micron, Samsung, SK Hynix and the entire AI infrastructure supply chain.

One company's cost is another company's revenue.

And when the spending stops, the revenue disappears.

---

Korea May Have Been the First Domino

There are two ways to interpret what just happened in South Korea.

The first is that this was simply a contained Korean correction.

Memory prices stabilize.

AI spending continues.

Microsoft, Google, Amazon and Meta keep spending through 2027.

The KOSPI eventually recovers.

That's entirely possible.

But there's another possibility.

South Korea was the first domino.

The Korean market was essentially a highly concentrated AI trade financed with leverage.

When investors questioned the AI spending story, the entire structure began unwinding.

The United States has a similar setup — except on a much larger scale.

Record margin debt.

Highly concentrated indexes.

AI-dependent valuations.

Massive capital expenditures.

And an enormous amount of investor confidence built around the assumption that the spending will eventually pay off.

---

So What Would Actually Pop the U.S. AI Bubble?

Probably not one bad earnings report.

The real trigger could be something much more subtle.

One major hyperscaler cutting AI CapEx.

AI companies struggling to raise additional capital.

Data-center financing becoming more expensive.

Corporate demand shifting toward cheaper AI models.

Or investors simply deciding that another trillion dollars of AI spending isn't worth the expected return.

Once that happens, the market doesn't need everyone to panic.

It only needs enough investors to stop believing the story.

And that's exactly what happened during previous bubbles.

The spending didn't necessarily stop first.

The belief stopped first.

---

Korea Just Showed Us What Leverage Does

The most important lesson from South Korea isn't that AI is fake.

It isn't.

It's that leverage turns a normal correction into something much more dangerous.

A stock falling 20% is one thing.

A stock falling 20% while millions of investors are forced to liquidate, leveraged ETFs are rebalancing, brokers are issuing margin calls and other investors are selling because they expect even lower prices is something completely different.

That's how a correction becomes a feedback loop.

And America's market may have more leverage sitting underneath it than ever before.

---

The AI Revolution Can Still Be Real

The AI revolution could ultimately create some of the most valuable companies in history.

But investors are currently paying for a future in which that success is already guaranteed.

That's the dangerous part.

AI doesn't have to fail for AI stocks to crash.

The technology can work.

Corporate productivity can improve.

New industries can emerge.

And investors can still lose enormous amounts of money if they paid too much for that future.

South Korea just demonstrated what happens when confidence in that future starts to crack.

---

The Bigger Picture

The Korean sell-off matters because it exposes the mechanics underneath a modern AI-driven market.

Concentration creates fragility.

Leverage creates forced selling.

AI spending creates interconnected demand.

And confidence holds the entire system together.

As long as investors believe that hundreds of billions of dollars in AI capital expenditure will eventually produce extraordinary profits, the machine can continue.

But if that belief changes, the reversal can be much faster than the rise.

South Korea may recover.

The U.S. AI boom may continue for years.

Or Korea may prove to have been an early warning.

Nobody knows.

But the question investors should be asking isn't whether AI will change the world.

It's much simpler:

**How much of today's market valuation is already assuming that it has?**