Broadcom Inc's (NASDAQ:AVGO, XETRA:1YD) fiscal second-quarter earnings report, released after the bell on Wednesday, was not a disaster by any conventional measure.
Revenue climbed 48% year on year to $22.19 billion, net income rose 88%, and the company guided for third-quarter revenue of $29.4 billion, comfortably ahead of Wall Street expectations.
And yet the stock fell 13% in Thursday trading, dragging Asian markets with it and sending Nasdaq futures sharply lower, because in the world of AI-driven investing, merely meeting expectations is no longer enough.
The specific disappointment was narrow but telling: chief executive Hock Tan declined to raise the company's full-year artificial intelligence chip revenue forecast, leaving the $100 billion target unchanged, and the company's enterprise software division, much of it inherited from its 2023 acquisition of VMware, missed estimates.
For a stock that had risen close to 40% in 2026 alone and multiplied nearly ninefold since ChatGPT ignited the generative AI boom at the end of 2022, the bar was simply set too high.
The ripple effects were swift and severe. Broadcom dragged CrowdStrike and Micron Technology with it into the pit of fear, before concerns fanned out globally.
South Korea's Kospi index dropped more than 5%, with SK Hynix, one of the world's leading producers of high-bandwidth memory chips used in AI accelerators, plunging 8.4%, and Samsung Electronics shedding 5.4%.
Japan's Nikkei fell 1.4%, led by chip equipment maker Tokyo Electron, which dropped 7.2%.
The scale of the reaction illustrates how tightly wound the AI trade has become, with valuations across the semiconductor supply chain priced for continued hypergrowth, leaving little margin for anything short of an upgrade.
The central question now is whether Broadcom's stumble is a genuine inflexion point or a speed bump on a road that remains intact.
The bull case remains, albeit slightly dented.
Tan confirmed that Broadcom counts six major custom chip customers, including Google, Meta, Anthropic and OpenAI, and said AI revenue more than doubled year on year to $10.8 billion in the quarter, with a further tripling expected in the current period to $16 billion.
That is not the language of a cycle rolling over.
The bear case, however, is that the investment required to sustain AI infrastructure at the scale currently envisaged is beginning to strain even the most committed balance sheets, and that the gap between capital commitment and commercial return is widening.
Tan himself acknowledged as much, noting that customer bookings were not for immediate delivery, and that buyers still needed to "align quite a few other things" before orders could be fulfilled, a carefully phrased admission that the AI buildout is more complex, and potentially more drawn out, than the market's pricing implies.
The honest answer is that one quarterly print proves very little in either direction.
What it does prove is that the AI trade is now so finely balanced, and sentiment so stretched, that even a whisker of ambiguity from one of its most important bellwethers is enough to send markets into a tailspin from Seoul to Tokyo to Wall Street.
That fragility, more than anything in Broadcom's results, is worth paying attention to.
