Chip stocks fell before the market opened on Monday.
Nasdaq 100 futures were down about 1.5 percent and S&P 500 futures between 0.6 and 0.8 percent. In premarket trading Nvidia was off about 3 percent, Arm and Marvell about 6 percent, CoreWeave 6 percent, and Intel and HP more than 5 percent. ASML, the Dutch maker of the machines that print chips, fell 9.5 percent. In Tokyo, SoftBank Group lost 10.7 percent. South Korea's Kospi fell more than 3 percent.
The trigger most traders pointed to was a weekend essay. Dario Amodei, the chief executive of Anthropic, wrote that AI companies "must slow the pace at which we improve the capabilities of AI models." Sam Altman and Elon Musk agreed with him in public. This desk set out what the essay proposes on Sunday.
Why the sellers went to the suppliers
The companies whose pace the essay is about are mostly not listed. Anthropic is private. OpenAI said on the same weekend that it would not go public this year.
That leaves investors who think slower AI is bad news with one place to express it: the listed companies that sell the labs what they build with. Chips, the machines that make chips, and the data centres that rent chips out.
The pattern of the declines fits that reading. The largest falls were furthest upstream. ASML sells equipment that takes years to install and pays off over a decade, so its value depends more than anyone's on a long, uninterrupted build-out. SoftBank's fall reflects how much of its value now rides on AI investments. Nvidia, which sells into inference as well as training, fell less than the companies more tied to new capacity.
What the essay actually said about hardware
Less than the selling implies.
Amodei's proposal is to pace capability, not to stop training. Its first step — outside evaluators embedded inside the labs — requires no change in anyone's chip orders. Its second step raises limits on training compute as one of two possible ways to measure pace, and it depends on government mediation or an antitrust waiver that does not yet exist.
The market is pricing the most restrictive reading of a step that may never be taken, at the moment it was least likely to be imminent.
It is also ignoring the half of demand that pacing does not touch. As this desk reported last week, two-thirds of chip spending has already moved to running models rather than training them. A slower frontier does not reduce how many people use the models that exist.
The other two reasons stocks were falling
It is worth being careful about how much of Monday belongs to the essay.
Markets were pricing roughly an 85 percent chance that the Federal Reserve raises rates on Wednesday, after core consumer prices rose 0.3 percent in August against expectations of 0.2 percent. Higher rates fall hardest on companies valued on distant earnings, which describes most of the AI trade.
And Brent rose toward $108 after Saudi Arabia shut its East-West pipeline and the Houthis took control of the Bab al-Mandab. Oil at that level is an inflation problem, which feeds straight back into the rate expectation.
Three pieces of bad news for richly valued technology arrived on one morning. Only one of them was about AI.
What to watch
The close, which will show how much of the premarket move holds once US trading begins in volume. Then the first chip company or data centre operator to comment on customer orders, which is the only evidence that would turn a reaction to an essay into a change in demand. Until then Monday is a statement about sentiment, and sentiment this sensitive can reverse on the next endorsement.
The futures moves; the premarket declines in Nvidia, Arm, Marvell, CoreWeave, Intel, HP and ASML; SoftBank Group's 10.7 percent fall in Tokyo; the Kospi's decline of more than 3 percent; and the roughly 85 percent odds of a Federal Reserve increase are as reported by NBC News on 14 September 2026. The S&P 500 futures figure of 0.6 percent, Brent's rise toward $108, the 86 percent odds and the 0.3 percent monthly core inflation reading against expectations of 0.2 percent are as reported by Yahoo Finance the same morning. Premarket prices are indicative and may differ materially from closing prices. The quotations from Dario Amodei's essay are from "We Must Pace the Frontier", published 12 September. The analysis is our own.





