Global tech stocks plummet: Another stress test for the AI bull market

Original Source: Wall Street CN

June 23, Seoul, South Korea.

At 2:00 PM, the Korea Exchange triggered a circuit breaker. The KOSPI index plummeted 8%, suspending trading for 20 minutes. After resuming, it continued its decline, eventually closing down 9.99% at 8,203.84 points.

A few numbers illustrate the intensity of this sell-off—

Samsung Electronics fell 12.31%, SK Hynix dropped 12.47%. These two companies alone contributed approximately 71% of the KOSPI's decline that day. The Nikkei 225 fell 3.55%, losing the 70,000-point mark. Nasdaq 100 futures dropped 3.01%, and the Philadelphia Semiconductor Index closed down 7.7%. TSMC fell over 5% in pre-market trading, Micron Technology dropped over 8%, while AMD, Intel, Applied Materials, ARM, and ASML all fell over 7%.

The collapse of leveraged ETFs was even more staggering: the 3x Long Korea ETF plunged 32% in a single day, and the 3x Long Semiconductor ETF plummeted 17%.

The KOSPI's single-day decline ranks among the top five in South Korean stock market history. The last crash of a similar magnitude occurred in October 2008.

But 2008 was a clearly visible Great Recession. In 2026, the global economy is growing, the AI revolution is in full swing, and the KOSPI had been among the top-performing major global indices year-to-date—until this crash.

So the real questions are: what happened, and why.

The Trigger: A Confluence of Three Signals

Looking back at the timeline, the direct trigger for the crash was the confluence of three signals within 24 hours.

First: SK Hynix Slows HBM4 Expansion.

On the morning of June 23, South Korean media reported that SK Hynix was slowing its HBM4 capacity expansion, shifting focus to general DRAM. HBM (High Bandwidth Memory) is a core component for AI chips, with SK Hynix and Samsung being the only two suppliers capable of mass production. Market consensus on HBM4 supply and demand was almost overwhelmingly leaning towards "supply shortage."

HBM4 is one of the most certain bottlenecks in the global AI infrastructure race. Any time the market begins to question the tightness of this bottleneck, the result is often a repricing of faith.

Second: Profit-Taking Ahead of Micron's Earnings.

Micron Technology was set to report its quarterly earnings on Wednesday (June 25). Driven by news of a full-stack partnership between Micron and Anthropic, the stock had reached an all-time high, with year-to-date gains exceeding 300%. Goldman Sachs' trading desk noted, "Investor expectations have been pulled extremely high, creating conditions for early position reduction ahead of earnings."

In a market driven by "expectations," position adjustment before earnings can be more damaging than the earnings themselves.

Third: South Korean Regulator Warns Against Leveraged ETFs.

On June 22 (the day before the crash), Lee Bok-hyun, head of South Korea's Financial Supervisory Service, publicly stated that he "regretted" not preventing the issuance of single-stock leveraged ETFs tracking Samsung Electronics and SK Hynix, saying they "serve little purpose other than allowing securities firms to profit at the expense of retail investors."

The timing was almost cruelly precise. When a regulator publicly acknowledges structural problems in the market, it directly triggers a panic-driven stampede.

The Amplifiers: The Three Levers of the Korean Market

The reason the confluence of these three forces was so devastating is that it struck a market structure already deeply tied to leverage.

In this round of the Korean AI bull market, three amplifiers were operating simultaneously.

Amplifier One: Retail Margin Debt Hits Record Highs.

South Korean retail investors have long been known for their "gambling" spirit. But the leverage in this cycle has reached unprecedented levels. Retail margin balances in South Korea continued to climb to record highs before the crash. In its post-crash analysis, Goldman Sachs stated bluntly: "The Korean stock market's rally is increasingly dependent on retail investors as marginal buyers."

In a leverage loop where buying begets more buying, once marginal buyers turn, it triggers a reverse cycle of selling begets more selling.

Amplifier Two: Single-Stock Leveraged ETFs Balloon to $30 Billion.

This is the most unique problem in the Korean market. 16 onshore single-stock leveraged ETFs have assets of about $9.1 billion, while CSOP's 2x Long SK Hynix and Samsung ETFs listed on the Hong Kong Stock Exchange have a combined size of approximately $21 billion—totaling over $30 billion. Among onshore products, 92% of holdings come from retail investors.

Single-stock leveraged ETFs have a fatal structural characteristic: they require daily rebalancing. When the underlying stock price falls, the ETF needs to sell more stocks to maintain its leverage multiple, creating a self-reinforcing selling pressure during downturns. And when regulators hinted at possibly restricting such products, the first assets to be sold were precisely these products—and their underlying stocks.

The Financial Supervisory Service estimates that trading commissions from single-leveraged products range from $3 billion to $6.4 billion. Measures under consideration include raising investment thresholds for retail investors, strengthening trader education tests, imposing size limits on single-stock ETFs, restricting new product launches, and reinforcing suspension mechanisms when prices deviate from net asset value.

Whether or not these measures are implemented, the signal they send is clear enough: regulators believe a significant portion of this rally has deviated from fundamentally rational pricing.

Amplifier Three: The National Pension Service Becomes an Unexpected Seller.

South Korea's National Pension Service (NPS)—the country's largest pension fund—net sold approximately $1 billion of KOSPI stocks in the six days leading up to the crash, accumulating net sales of $1.5 billion in June so far, the largest monthly net selling record since April 2021.

NPS's operation was essentially a rebalancing move: the KOSPI's sustained rally had pushed its domestic stock positions above 30%, exceeding the approximately 28.8% cap.

But the key point is that in a market already heavily dependent on marginal retail funds, the sudden shift of NPS from buyer to seller—a core stabilizing buyer—meant there was no one left in the market to "catch" the sell orders.

As one Goldman Sachs analyst put it: "What was once a theoretical constraint has become an observable liquidity reality."

The Bubble Debate: When Will It Come, How Deep Will It Fall

After the crash, a debate about the "AI bubble" naturally emerged.

Chris Cha, Head of High-Touch Trading for Korea at Goldman Sachs, provided a clear diagnosis in his client note on the day of the crash—liquidity exhaustion: "I remain constructive on the memory cycle and continue to believe the KOSPI is undervalued. But this rally is increasingly dependent on technically sensitive buyers, making it more vulnerable to disruptions in liquidity momentum."

In other words: the medium to long-term logic hasn't changed, but the short-term market structure has become fragile to a critical point.

Bank of America's Quantitative Signal: The Nasdaq is Approaching Bubble Territory.

A few days before the crash, a team of BofA strategists published a report noting that their bubble risk indicator shows the Nasdaq 100 is approaching the 0.8 level, which "typically signals elevated short-term tail risks on both sides." Tech and semiconductor stocks have already "exhibited extreme bubble-like price movements."

But BofA also offered an interesting take: "The AI bubble could take years to fully form. History suggests this indicator is more useful for signaling periodic corrections rather than trend reversals."

Li Bei's Warning: Trigger Conditions Have Already Appeared.

Hedge fund veteran Li Bei also expressed caution about the AI sector in a letter to investors. In her view, the "trigger conditions for the bursting of the AI bubble have already appeared."

Three voices, pointing to three different time dimensions: Li Bei says "now," Goldman Sachs says "don't panic," and BofA says "it will rise more but first it must fall."

It's worth noting: In a market deeply bound by leveraged ETFs, retail margin debt, and momentum trading, a "periodic correction" and a "bubble burst" may be indistinguishable in terms of price action. A 10% daily drop has already triggered a circuit breaker. What would happen if another 10% followed?

Mirror Image: SpaceX Tells the Same Story

Shifting focus from Seoul to New York reveals a mirror image.

SpaceX has fallen for three consecutive days since its listing—down over 16% on June 19, another roughly 5% on June 22, and continuing to decline on June 23—its market capitalization evaporating by approximately $600 billion in three days, falling below its opening price of $150 to around $147.

More telling is this: amid the stock price crash, SpaceX announced its first-ever bond issuance—raising $20 billion for AI infrastructure construction. Typically, a company needs to issue debt when facing financial difficulties. But SpaceX's situation is the opposite: its AI capex story requires a constant supply of ammunition, and the falling stock price is closing the window for equity financing—debt is the last option to keep its narrative uninterrupted.

The resonance between SpaceX and Korean semiconductors reveals a common issue: the capital market narrative for the AI theme is shifting from "unlimited imagination" to a "calculating returns" phase.

When investors start doing the math—how much has HBM4 expansion slowed? How much have GPU rental prices dropped? When will AI revenue cover capital expenditures?—the market's pricing logic changes.

Looking Ahead: The Micron Verdict

After the crash, all eyes turn to one date: June 25.

Micron's earnings report—this is the most direct "verdict" facing tech stocks after this round of decline.

Bloomberg quotes Dilin Wu, a strategist at Pepperstone: "Micron's earnings this week are the real test. A strong performance would directly and positively transmit to Samsung and Hynix—this number will tell you whether the underlying logic of the AI hardware trade still holds."

Logically, there are two possibilities:

Scenario One: Micron Beats Expectations. If Micron delivers a strong report with optimistic guidance, the current sell-off could be quickly corrected—positions that exited due to uncertainty would be re-established once certainty arrives.

Scenario Two: Micron Misses Expectations. If Micron's guidance disappoints, the current sell-off will receive fundamental validation—the "liquidity issue" diagnosis will be overturned, leading to a collapse of faith.

In Scenario One, the KOSPI circuit breaker will be marked as "a technical stampede." In Scenario Two, it will be marked as "the turning point of the AI bull market."

The answer the market provides on Wednesday evening will determine the opening direction of Asian markets on Thursday morning.

Conclusion

The crash on June 23 can be explained technically as: a single SK Hynix HBM4 news + structural fragility of leveraged ETFs + NPS's unexpected selling for rebalancing + risk-off positioning ahead of Micron earnings—a confluence of four factors within 24 hours.

But this is only a technical explanation.

The deeper question is: when the AI bull market has reached a point where valuations are at historical extremes, the driving force has shifted from institutions to retail investors, the trading structure is deeply tied to leveraged products, and inflation and interest rate expectations are pushing up the risk-free rate—a structural breakdown was almost inevitable.

The KOSPI circuit breaker is a mirror.

What the mirror reflects is this: when all market participants are using leverage to bet on the same narrative, the correction in that narrative—no matter how fast or deep—should not be considered a surprise.

For investors, after Micron's earnings, there is only one question: How much of a drawdown are you willing to accept to answer the question, 'Should I continue to hold?'