These are actually one type of market event that at a first glance seem like a simple earnings reaction. Broadcom releases quarterly results, no one can have imagined anything greater when it comes to AI chips and by the following Monday, the South Korean stock exchange has had to suspend trading, Japan's Nikkei has plummeted almost four percent and European investors have been frantically exiting out of what they were gloating about, just a week ago. This isn't a typical earnings response. It's a system that is under pressure, finding a crack.

The crack was opened on the evening of June 3rd when Broadcom plunged 12.6% on June 4th, offloading $280 billion in market value, one of the largest single stock wipeouts of the megacap era. While the company exceeded quarterly earnings estimates, its outlook on AI chip sales proved disappointing to investors following a strong rally in the stock, which weighed on the overall chip market. The message was less palatable: a company can be growing like gangbusters and the stock still tank if the market had anticipated something even quicker. So, here's to the AI trade in 2026.

It was a domino effect from there, and it spread throughout the world. The Nasdaq lost 4.18% on June 4th, closing at 25,709.43 its biggest drop since April 2025. The S&P 500 dropped 2.64% and ended at 7,383.74, while the Dow Jones Industrial Average lost 695 points, settling at 50,866.78. Micron Technology has fallen 17%, Intel plummeted 9% and AMD has declined 12.6% during the two trading sessions since Broadcom's report. These were not businesses with poor fundamentals, as they had seen huge gains on the AI infrastructure wave. It wasn't the business that changed, but the narrative.

The impact was felt first in Asia. The Kospi's opening was 5.54 percent lower than the previous Friday's level of 8,160.59. The shares of index heavyweights Samsung Electronics and SK Hynix fell by 6.40% and 9.92% respectively. On Monday, matters were in a dire state. The benchmark Kospi index tumbled almost 9% off the open to force a market halt for 20 minutes, before finishing 8.3% lower. The two companies make up over 40% of the index Samsung Electronics and SK Hynix ended Monday's session down 10.18% and 7.68%, respectively. Japanese tech investor SoftBank Group plunged 6.1%, while Tokyo Electron and Advantest were down 7.45% and 5.72%, respectively.

The Korean situation in particular needs to be examined. The overnight circuit breaker on the KOSPI didn't help either, as traders interpreted the Korean halt as a forced liquidation of chip stocks and a margin stress that hit the whole sector. There was one glaring thing in the derivatives market: Before the crash put option open interest in the leading South Korea ETF, EWY, was already up to some 880,000 contracts, the highest since last year and nearly quadruple from just weeks ago. Call options, on the other hand, were up just a bit. Well before the market broke, somebody had been buying protection on the downside in size. That's either sophisticated macro hedging, early risk recognition or something more worrisome, which regulators and analysts are still figuring out. One thing that is not at issue is that the positioning was directionally correct.

Europe, fresh on Friday morning after the initial shock to be led by the United States, escaped not. The Stoxx 600 technology index closed the day 2.8% lower, with Infineon Technology's share price falling 9.1%, ASML's price dropping 2.4% and Nokia's price dropping 5.9%. The UK's FTSE 100 rose almost 0.1%, outflanking major bourses in Paris, Frankfurt and Milan, which closed in the red. European chip stocks mirrored their Asian peers on Monday before staging a reversal; BE Semiconductor's shares rose 3.5%, ASML closed 3.6% higher, Infineon advanced 2.3% and STMicroelectronics jumped 7.4%. But that bounce had a caveat in that it was driven in part by the same risk-off repricing that had mauled stocks, with investors coming back to beaten-down positions.

The question that has been rumbling around all this time on Wall Street, but which no one really wanted to address: Have the AI stocks been based on fundamental assumptions that these businesses were never going to match?

Broadcom's Q2 2026 earnings saw revenue of $22.19 billion, beating the consensus estimate of $22.13 billion. On the bottom line, Non-GAAP EPS was $2.44 compared to $2.39 expected. The company beat. It still fell 12.6%. The difference between delivery and expectation is the issue. The market had factored in every quarter with an AI bias to drive forward estimates. That beat stopped (even a little) and the repricing was brutal.

D.A. Davidson managing director Gil Luria noted that the broader market is worried about inflation, indicating that investors are concerned that "the data-center buildout could decelerate if companies back off on AI spending. This is the 2nd order concern: not only do individual businesses fall short of expectations, but that the broader capex cycle powering semiconductor demand slows down if businesses start to doubt the payoffs of investing in AI. The expected expenditure on AI in 2026 is estimated to be $539 billion according to Goldman Sachs. However, McKinsey indicates that 88% of businesses are using AI regularly, but that the adoption could be plateauing with employees still fearful of being replaced in the process, resulting in a disconnect between investment and utilization.

The Fed adds its own wrinkle. Goldman Sachs has raised its rate cut forecast for the Fed, predicting no rate moves for this year. The bank had earlier planned to cut interest rates by two quarter points in December 2026 and March 2027, but instead those cuts have been postponed to June and December 2027. That was because of a higher than expected May jobs report. Goldman's head of the US economics team, David Mericle, pointed to the strong labor market, and Goldman raised its estimate of the chance for small rate hikes to 20% from 10%. In a note, Goldman Sachs predicted its last two rate cuts would be postponed to June and December 2027, "as the labor market has been stronger than expected.

That has a bearing on the chip stocks in a direct and mathematical manner. High growth technology companies are valued based on future earnings discounted back to today's value. If rates remain high, it means that those future earnings are not worth as much in today's dollars. AI and semiconductor firms are highly valued on future earnings. As rates increase, the value of those future profits declines in today's dollars, and that's particularly bad for high-growth stocks such as Samsung, SK Hynix and Micron.

Geopolitics added to the mix. The Middle East situation, namely the escalating tensions between Iran and Israel, was not only a diplomatic background. Lately, there has been huge volatility in the global markets, due to the military confrontations between Israel and Iran. The ongoing Iran conflict has halted helium output in Qatar, responsible for about one-third of the world's supply of the gas used to manufacture computer chips. The rise in cost of energy is affecting input costs at data centres, which are now being attacked. But the rising price of energy and the scarcity of raw material are hardly esoteric issues for chip fabs they are in the margins.

Then, recovery starting Tuesday, at least in Asia. South Korean shares rose strongly on the Asian trading day of June 9, as memory chip makers took the lead to lift the KOSPI index by 8.28%. Samsung Electronics and SK Hynix posted substantial increases, indicating the steady investor confidence. Two factors aided the recovery: Nvidia CEO Jensen Huang said the recent sell-off in semiconductors was a buying opportunity, and Nvidia strengthened its ties with major South Korean companies, with SK Hynix making plans to increase HBM4 manufacturing capacity. Support was further added by an improvement in risk sentiment following Israel's and Iran's agreement to a ceasefire. As is the way of markets, a ceasefire, and a Jensen Huang endorsement were enough at least for now to turn a multi-day rout around.

But, the bounce should not overshadow the structural issues that the selloff raised. Goldman Sachs upped its estimate of AI capex in 2026 from $465 billion to $527 billion, and now estimates that AI capex will hit 0.8% of GDP, which is still close to 1.5% of GDP that occurred during similar tech booms over the last 150 years. The honest reading of that data point is that it's not that extreme yet, but it's certainly priced in that way already. Goldman Sachs estimates that the total capital expenditure for AI will come in at approximately $7.6 trillion from 2026 to 2031, with annual expenditure rising from $765 billion this year to $1.6 trillion by 2031. That's a staggering amount. So, whether or not the AI buildout is real is a moot point it's there. The question is whether the velocity of capital can keep up or not with the velocity of return on capital.

Bank of America strategist Michael Hartnett cautioned that the AI bubble is akin to what was seen during the dot-com era, and advised bonds and defensive stocks as his "post-bubble roadmap. But that's not the majority opinion. But the week of June 4th was warning enough for even the bulls to overlook: when the world's most valuable chipmaker beats earnings and loses $280 billion in market value during a single session, even the bulls are saying that expectations have risen above fundamentals. Semiconductors continue to be a strong business. Perhaps the story with them developed too far.

The week proved to be one of the times when those who trade in the global equity markets need to remind themselves of something they may have known intellectually but not always felt so: there is no geography in the technology sector anymore. Within 72 hours a guidance miss in Palo Alto, a jobs report in Washington, a missile exchange in the Middle East can wipe out a trillion dollars in paper money in Seoul, Tokyo, Amsterdam and Frankfurt. Interconnection is the feature. It is also the risk when the market is overvalued.