Nvidia closed 3.3% lower in New York on Monday. AMD fell 4%, Micron Technology and Sandisk 5% each, and the Nasdaq ended the day down 0.5%. The selling began after Anthropic chief executive Dario Amodei spent the weekend telling the AI industry to slow down. Investors started pricing in a scenario they had left out until now: an industry that decelerates on purpose, and therefore finds it harder to pay for the hundreds of billions of dollars of infrastructure it has already committed to.
The damage spread across every market that sells into the buildout. SoftBank, a heavy investor in OpenAI, dropped 13%. South Korea's Kospi, which leans on memory makers supplying AI companies, fell 3%. Taiwan Semiconductor Manufacturing Company slipped 1.2%. In Europe, ASML — the continent's most valuable company and a critical supplier to the semiconductor industry — lost 6%.
The more revealing trade was on the other side. WPP, the advertising group, gained 5% in London. Relx, the analytics company, also gained 5%. Both are businesses that AI is supposed to eat, and both had a good day because AI had a bad one. Relx in particular fell sharply earlier this year after Anthropic introduced a set of new data and automation tools. On Monday the market took Amodei's warning as a partial reprieve for his own victims.
Amodei's argument is that the current pace is reckless, and that a swarm of AI agents could do hundreds of billions of dollars of damage if it seized the entire internet. Some experts disputed the claim. It nonetheless drew an unusual set of endorsements: Sam Altman of OpenAI, Demis Hassabis of Google DeepMind and Elon Musk of SpaceX all backed Amodei's essay on slowing AI development, "We Need to Withstand the Pace of Frontier Development." Altman went further and said he was prepared to match Amodei's commitment by placing external evaluators inside OpenAI to check that safety measures are being followed. Anthropic co-founder Jack Clark told the BBC that preventing a catastrophe might require an AI kill switch that a third party could activate.
Donald Trump rejected all of it. In a social post on Monday the president wrote that the only control or "guardrail" AI needs is "a strong and smart president with a high IQ," and that the United States already has one. He said his administration had stopped people in AI who might have been doing bad or potentially dangerous things, naming Dario Amodei and Anthropic specifically. He described a "sick conspiracy" operating against AI and data centres that benefits only China, and said the country that wins the AI race wins everything. He made the same case by phone to Nvidia's Jensen Huang, who was on stage at an AI conference in California at the time; over the speakerphone, Trump called the fears around AI a hoax and said robots would not take over the world.
Governments are moving in the opposite direction. On Monday a cross-party group of British MPs and members of the House of Lords set out a list of human rights threats from AI and concluded that no country in the world yet has laws adequate to contain them. In China, the official responsible for intelligence warned that adversaries' use of AI could threaten the country's political and social security: on Sunday Chen Yixin wrote in a state publication that frontier American models, including Anthropic's Mythos and OpenAI's GPT-5.5-Cyber, could pose a serious threat to Beijing's critical information infrastructure, and called for a comprehensive strengthening of AI security. The UN Security Council plans to hold a session on AI next week, Agence France-Presse reported on Monday, as international concern builds.
Set against that noise, the market move was small. A 3.3% day for Nvidia is not a repricing of the AI trade; it is a flinch. Nothing changed on Monday about demand, supply, or a single signed compute contract. What changed was the rhetoric, and the sell-off is best read as the market discovering that "the people building this say it is dangerous" is now a headline risk with no obvious hedge. The rotation into WPP and Relx makes the point: this was not a judgement that AI will be worth less, only that it may arrive more slowly.
Jim Reid of Deutsche Bank does not expect even that. Competition in the sector, he argues, will not let companies cut investment sharply; rivalry between firms and between countries remains intense, and it is hard to imagine any of them retreating voluntarily while rivals keep moving, or to expect China to stop. He suggested that leaders discussing risk in public may be a way of demonstrating how transformative they believe the technology to be while advertising what their own products can do. His read is that these statements mean not less spending but a larger share of it going to safety, monitoring and governance, while the compute buildout carries on regardless.
That is the question the weekend left unanswered. Every remedy on the table — external evaluators, a third-party kill switch, a slower frontier — is something the labs would grant themselves, on their own timetable, with no named authority to enforce it and no penalty for walking it back. Altman's offer of in-house evaluators is a commitment to host an inspector he selects. Clark's kill switch has no third party attached to it. The UK parliamentarians' finding, that no country has sufficient laws, is the whole problem stated plainly: the only people currently able to slow AI down are the people being asked to.
The timing sharpens it. Amodei's warning landed just before reports that Anthropic could turn a profit as soon as this quarter; the company has told investors its adjusted operating profit will be positive for a second consecutive quarter, a significant result ahead of a possible US listing this year, according to the Financial Times. OpenAI has hinted at its own listing plans, but Altman said over the weekend that the company will not go public in 2026, citing concerns about the safety of the technology. So one lab is asking the industry to decelerate while preparing to sell shares to public investors, and the other is citing safety as a reason not to sell them at all. Whichever of those is the sincere position, the public markets are about to price the difference.