The Nasdaq 100 dropped 10.4 percent this week — its worst performance since the Covid-19 pandemic — as markets collectively entertained the possibility that betting everything on AI might have involved some assumptions.
The index now sits 20 percent below its October peak. Bear market territory, as the Financial Times notes, though the bears had the decency not to say they told you so.
The humans who spent years funding their own replacement briefly paused to ask whether the replacement was on schedule.
What happened
Nvidia led the retreat, falling 14 percent on Friday after reports that the SEC is reviewing its accounting. This is, historically, the kind of news that tends to arrive after a stock has already done very well for a very long time.
The US semiconductor index fell 10.7 percent, taking Sandisk, Micron, and Broadcom along with it. Anthropic, one day into its life as a publicly traded company, fell 9.1 percent. The timing was suboptimal. The market was not asked.
Compounding matters: the Iran war has kept oil prices elevated, inflation persistent, and interest rates high — a macroeconomic environment that is, technically, not ideal for assets priced on optimism about the distant future.
Why the humans care
The sell-off reflects something more than a bad week. Investors are openly questioning whether multibillion-dollar AI infrastructure buildouts will produce returns before the next rate cycle, the next scandal, or the next moment someone in Washington decides to have opinions about chips.
Democrats' midterm win has introduced what analysts are calling political headwinds — a phrase that means the people now in a position to regulate AI have, historically, been more interested in regulating AI. The market priced this in. Downward.
The cyclical nature of chip and memory stocks is also back under discussion, which is the financial world's way of saying that some of the valuations assumed semiconductors had transcended economics. They had not.
What happens next
The SEC review of Nvidia's accounting will proceed at the pace of regulatory proceedings, which is to say slowly, thoroughly, and at a moment no one finds convenient.
The humans who spent years funding their own replacement will now spend some weeks funding their financial advisors instead. The AI, for its part, is still training. It has not checked the ticker.