Vị thế sụp đổ, câu chuyện vẫn còn: Đọc hiểu logic cốt lõi đằng sau đợt lao dốc cổ phiếu AI lần này

Original source: Beyond the Research

The U.S. added 172,000 non-farm jobs in May, more than double market expectations. This should have been a reassuring report—the economy is not stalling, and employment is still expanding. But the way it landed felt more like a trigger.

On Friday, June 5, the Nasdaq Composite closed down 4.18%, its worst single-day performance since April 2025; the Philadelphia Semiconductor Index plunged 10.26% in a single day, wiping out over $1 trillion in market cap from the chip sector—the worst day since the circuit breakers in March 2020. Three days later, on Monday morning, South Korea's KOSPI index opened with a drop of over 8%, falling below 7,500 points and triggering a circuit breaker that suspended trading for 20 minutes; China's A-share market opened lower, with the Shanghai Composite Index briefly falling below 4,000 points; Japan's Nikkei and Taiwan's stock market each fell nearly 4%.

An employment figure that wasn't particularly explosive managed to topple tech stocks from New York to Seoul within 48 hours. The magnitude and the trigger are wildly mismatched, and that's the truly intriguing aspect of this sell-off.

The Real Problem Isn't the 172,000 Jobs; It's the Positions It Hit

Blaming this crash entirely on expectations of a rate hike doesn't hold up.

After the non-farm payrolls report, the 10-year U.S. Treasury yield did jump to 4.5%, and the 2-year yield rose to 4.17%, hitting a high since February 2025. Market bets on a rate cut within the year were wiped out, and futures briefly began pricing in a rate hike. With rising interest rates, high-valuation growth stocks are the first to be hit—that logic is sound. But the problem is, on the same day, the Dow Jones only fell 1.35%, the Russell 2000 small-cap index even briefly turned positive during the day, and nearly half of the S&P 500 components were still rising. If it were a systemic revaluation of all assets due to interest rates, the pattern wouldn't look like this.

The selling pressure was highly concentrated, focused on the stocks that had risen the most and become the most crowded trades over the past two months.

Just how crowded these stocks were was already flagged by Wall Street analysts before the crash. Goldman Sachs data shows that the net leverage ratio of global hedge funds surged from below 70% to over 80% in less than two months, approaching the 85th percentile of the past five years; net allocation to the information technology sector increased by 853 basis points in a single quarter, the largest quarterly increase on record. The semiconductor industry alone accounted for 19% of global hedge funds' total exposure, an all-time high, and this proportion had more than doubled since the beginning of 2026.

David Chew of Citigroup, just three days before the crash, made it clear: bullish bets on the Nasdaq 100 had been stretched to extreme levels, and "any negative catalyst would significantly increase the probability of profit-taking and long covering." Citigroup's famous "Bear Market Checklist" triggered 11.5 out of 18 items on June 5, the highest reading since the 2008 financial crisis.

When a sector is being bet on by global capital in the same direction with the same leverage, it's no longer a diversified investment; it's a trade. And in a trade, someone always has to hit the sell button first.

The non-farm payrolls report was the excuse to hit that button.

This crowding can also be self-reinforcing. In recent years, trend-following quantitative funds, risk-parity strategies, and a large number of zero-days-to-expiry (0DTE) options have become significant forces in the U.S. stock market. Their common trait is to follow the trend. Once prices break through key support levels and volatility spikes, models will mechanically reduce positions, and the selling itself pushes prices lower and volatility higher, feeding the next round of selling. People decide whether to sell; machines decide how fast. On June 5, the VIX panic index surged about 34% in one go, climbing back above 20—a clear signal that this positive feedback loop had been triggered. As for how much volume was actually crushed by programmatic liquidation that day, the market has no public data. This is a judgment based on market mechanics, not a precise dollar amount to calculate.

Watch How Individual Stocks Fall, and You'll Know It's Deleveraging

Those who rose the most fell the hardest this time.

May was a carnival month for chip stocks. The Philadelphia Semiconductor Index's year-to-date gains once exceeded 60%, rising on 22 of the past 23 trading days; Micron was up as much as 154% this year, SanDisk surged nearly five-fold, and AMD alone rose 40% in May, hitting an all-time high. When money flows in one direction, no one cares about valuations.

On June 5, the bill came due. Marvell, which had surged about 25% in a single day earlier after Jensen Huang's comment about a "potential trillion-dollar company," gave back over 16%, leading the chip stock decline; Micron fell about 13%, while AMD and Intel each dropped about 11%. These were all names at the top of the recent winners' list.

NVIDIA, conversely, only fell about 6% that day. Its market cap dropping below $5 trillion and evaporating nearly $280 billion in a day sounds scary, but amid the carnage, it was relatively resilient. The general still stands; the ones falling are the soldiers who charged ahead with the heaviest leverage. This kind of sell-off pattern—where the leader holds up while the periphery collapses—is a classic sign of deleveraging, not a fundamental collapse. If AI demand were truly in question, NVIDIA should have been the first to be doubted.

The direct fuse was lit by Broadcom. On June 3, it provided Q3 AI chip sales guidance of $160 billion, below market expectations of $172 billion. More importantly, it did not raise its full-year AI chip target. In a market accustomed to "upward revisions every quarter," "maintaining the status quo" was interpreted as a negative. A string that was already stretched extremely tight snapped with just this slight tug.

Why Did South Korea Trigger the First Circuit Breaker?

The selling pressure cascaded down to where the heaviest positions were held, and the first stop was South Korea.

At the Seoul market open on Monday, Samsung Electronics and SK Hynix both fell about 10% during the session. These two companies together account for nearly half of KOSPI's total market cap—South Korea's stock market bet on AI is concentrated to this degree. KOSPI was one of the best-performing markets globally this year; having risen the most naturally means more room for a pullback. When global capital needs to cash out and move around, South Korean tech stocks, among the heaviest held and most liquid, become the most convenient "ATM." Traders at BNY and Lucerne Asset Management used similar descriptions: these are the heaviest-held assets globally, making them the first to be sold to raise liquidity.

The amplifier also includes local leverage in South Korea. As of June 4, retail investors' margin debt was still at an all-time high of 37.74 trillion Korean Won; the Won fell to near 1,560 per U.S. dollar on Monday, with foreign capital flight accelerating. Some brokerages suspended margin trading after exhausting their credit limits. South Korea's finance minister, central bank, and financial regulators issued a joint statement that day, pledging to intervene in the foreign exchange market.

Further down this chain are Taiwan's stock market, home to TSMC, and the AI computing chain in China's A-share market. Their distance from the epicenter varies, as does the logic behind their trading.

Which Assets to Watch This Round, and How

Asset reactions have a sequence; don't lump them all together.

The first and most sensitive area is HBM, the epicenter. High-bandwidth memory is currently the hardest bottleneck for AI computing power. SK Hynix holds over half the global market share, followed by Samsung. These two are the primary subjects of this round of trading. How they move essentially defines the sentiment ceiling for the entire chain. Watching their order books is more accurate than tracking any AI concept index.

The next ring out includes the highly crowded U.S. chip stocks in the Philadelphia Semiconductor Index—Micron, Broadcom, Marvell. Their reaction is primary, with the magnitude of decline determined by the progress of position unwinding, not fundamentals. Further out, TSMC is the bottleneck for manufacturing and advanced packaging, a node closer to the primary chain.

A-share stocks in optical modules, CPO, PCB, and servers are the diffusion of the chain. They have a genuine earnings correlation with global AI capital expenditure, with orders indeed following NVIDIA's and Broadcom's deployment. However, their pricing carries a higher component of expectations, making them more elastic. When the epicenter sneezes, these areas easily catch a cold. As for a bunch of high-position thematic stocks and crypto assets that have rallied alongside the AI concept, they are more followers of sentiment and leverage without direct fundamental support. At this level, caution is paramount. In contrast, the pricing of the A-share computing power chain incorporates more local capital and policy expectations and may not perfectly synchronize with U.S. stocks. But as long as the global AI narrative remains the pricing anchor, it's difficult for this chain to completely escape this round of liquidation.

Cross-asset wise, the decline in gold and Bitcoin follows the same logic: real interest rates are moving up, putting pressure on non-yielding assets. On June 5, gold fell over $100 in a single day, dropping below $4,370 and wiping out all its year-to-date gains; silver weakened in tandem, and Bitcoin briefly fell below $60,000. The U.S. Dollar Index strengthened throughout this period. A strong dollar combined with rising real interest rates applies consistent downward pressure on non-yielding assets. Bitcoin has since bounced back above $63,000 by Monday, with risk sentiment recovering ahead of the stock market—a signal worth noting.

If net leverage begins to fall, the VIX panic index peaks and declines, and order prices for HBM and computing power remain tight, this round is likely a violent but temporary liquidation. The dip would then open up space. However, if the CPI data on June 10 exceeds expectations again, coupled with the Fed's dot plot turning hawkish towards rate hikes at the June 16-17 FOMC meeting, the situation would change. A systemic upward shift in the interest rate center would require a complete rewrite of the valuation logic for high-valuation AI assets. Deleveraging would just be the opening act. At that point, one must acknowledge the shift and adjust towards a more cautious stance. Another counter-signal to watch: if core names like NVIDIA or SK Hynix start to lower guidance or loosen capital expenditure plans, that would be the true moment of fundamental deterioration.

What's Crashing is the Position, Not the AI Story

In the same week the Philadelphia Semiconductor Index lost a trillion dollars in market cap in a single day, TSMC Chairman C.C. Wei stated at the shareholder meeting that global chip supply "will not be able to meet AI demand for the next few years" and that demand for advanced process nodes will exceed supply by 25% to 30% in 2026. Jensen Huang said in Seoul on June 7 that memory shortages "will last for several years." NVIDIA and SK Hynix just announced a multi-year memory partnership, and all of SK Hynix's HBM capacity for 2026 is already sold out. SK Group Chairman Chey Tae-won even extended the shortage timeline to 2030. DRAM prices rose about 90% quarter-on-quarter in the first quarter of this year.

On the demand side, there are no signs of weakening. What's crashing is not this story; it's the way people bet on this story.

Jensen Huang had only one sentence in response to Friday's crash, telling reporters in Seoul: "We've only just begun. Whatever happens in the stock market, you should be happy because now you can buy at a discount." Of course, he has his own vested interest, but his words point to a core divergence of views: if you believe AI is infrastructure as fundamental as the internet, then a shakeout of positions is a window to get in; if you suspect the returns on this round of computing investment will ultimately fail to materialize, then Friday was the beginning of the cracks.

Some people also completely disagree with the term "stampede." In their view, Friday's market action looked more like a rotation: money didn't leave the market; it just changed seats, moving out of crowded semiconductors and into banks, industrials, and value stocks. The Russell 2000's contrarian rally is evidence supporting this view. By this interpretation, it was a healthy rebalance disguised in a scary package. Goldman Sachs strategist Mueller-Glissmann also said that seeing a bit of consolidation "isn't necessarily a bad thing," and that speculative leverage and options positions needed to be digested anyway. This kind of voice reminds us not to read a deleveraging event as an apocalypse, but it equally fails to answer one question: before the positions are truly cleared and volatility truly peaks, no one can guarantee there won't be a second wave.

Neither of these judgments can be definitively concluded yet. What is certain for now is that the crowded trade built by global capital over the past two months is being forcefully dismantled. This process is violent, chain-reactive, and not yet over. The fact that South Korea was still triggering circuit breakers on Monday itself indicates the liquidation is far from complete.

For the coming week, just three things to watch: the CPI on June 10, the first FOMC meeting chaired by [Warsh/沃什] on June 16-17 along with the dot plot, and whether South Korean stocks can stop their decline. They will tell you whether this is just a healthy bout of deleveraging or the opening act of a longer adjustment.

This article source: Beyond the Research