Why Semiconductor Stocks Sold Off After Anthropic’s AI Slowdown Call

Semiconductor stocks fell sharply after Anthropic CEO Dario Amodei’s We Must Pace the Frontier essay gave investors a new reason to question the near-term pace of the AI buildout. The central market concern was not that AI would disappear, but that a slower path for frontier-model capability gains could postpone the releases and data-center investments driving demand for chips, memory, servers and power infrastructure. A jump in oil prices and bond yields made that reassessment more painful for expensive growth stocks. 1512

What Amodei proposed—and why markets focused on it

Amodei argued that frontier AI companies should slow the rate at which they improve model capabilities, allowing safety work and oversight to keep up. His proposal included ongoing access for independent evaluators, coordination among leading AI developers and broader international arrangements around the most dangerous systems. Anthropic said it would provide third-party evaluators with permanent, employee-level access to its systems. 354041

That is materially different from a blanket suspension of AI training. The proposal was about pacing capability progress and applying safeguards, which leaves room for continued investment and development. But markets tend to price the marginal change: if frontier labs release or advance models more slowly, they may defer part of the enormous near-term spending on computing capacity. 53136

Sam Altman said he agreed that the frontier should be paced, and Elon Musk responded, “Dario is right.” Reporting also said Microsoft CEO Satya Nadella welcomed deliberate pacing and embedded evaluators. 3738

The response was not universal. President Donald Trump criticized the effort as a “SICK conspiracy going on against AI and Data Centers,” according to contemporaneous reporting. China’s Foreign Ministry rejected “fear mongering,” confrontation and destructive competition, instead calling for open and inclusive AI development and global cooperation. 33342

How the AI trade reacted

The most immediate losses were concentrated in companies tied to AI training and the data-center expansion. Market reports described a roughly 5% early drop in an ETF tracking major chip stocks, while Nasdaq-100 futures fell about 1.5% as traders worried that AI restraint could weigh on the capital-spending boom. 12

During the session, reporting put AMD down about 5.9% and the Philadelphia Semiconductor Index down about 5.6%; Nvidia was down roughly 3% to 3.5%. 35 Intel, Micron and other memory or chip names also fell sharply, with several reports putting early declines around 6% for Intel and Micron. 1517

The selloff reached beyond chip designers. CNBC reported steep declines among data-center and AI-infrastructure names, including HPE, Dell, Oracle, CoreWeave and Vertiv. 5 The Nasdaq Composite was down as much as 1.3% intraday before recovering some ground; it ultimately closed down 0.56%. 1821

Not every software category fell. Cybersecurity stocks rose on the expectation that increased attention to AI safety and control could bring more security spending; CNBC reported gains of more than 13% for Palo Alto Networks and CrowdStrike. 5

Why the move looked like a repricing and positioning unwind

The trading pattern did not establish that demand for leading AI accelerators had suddenly deteriorated. Instead, it looked more like a rapid repricing of a crowded AI capital-expenditure theme.

One clue was the breadth of the decline. Intel, which is less directly exposed to frontier-training GPU demand than Nvidia, was hit at least as hard in several intraday accounts. That is consistent with indiscriminate selling across chips, memory, servers and data-center infrastructure rather than a precise judgment that demand for the highest-end accelerators had collapsed. This is an inference from market behavior, not proof of sustained demand. 31517

A second clue was the split within software. The cybersecurity rally suggested investors were not abandoning AI-related spending altogether; they were rotating toward businesses perceived as potential beneficiaries of more stringent safety, security and governance requirements. 5

Finally, the policy message itself mattered. A call to slow capability advances could change the timing and governance of compute purchases, but it does not inherently mean long-run demand for AI infrastructure disappears. The supplied reporting does not provide sufficiently reliable evidence to establish the claimed implications of any Anthropic-AMD MI450 partnership, so it should not be used as evidence that the selloff contradicted a specific accelerator contract.

Oil, yields and Fed expectations made growth stocks vulnerable

The AI-specific catalyst arrived during a broader risk-off move. Reuters reported that oil rose as attacks on Saudi energy infrastructure and risks to Gulf shipping heightened supply concerns; Brent was up about 3% to $107.84 a barrel after gaining almost 9% the prior week. Other market coverage put Brent above $108. 112

Higher oil prices can intensify inflation concerns. Reuters reported that, after a hotter-than-expected U.S. consumer-price report, markets priced an 86% probability of a 25-basis-point Federal Reserve rate increase that week. 1 Rising yields and expectations for tighter policy generally weigh more heavily on high-valuation companies whose expected cash flows lie further in the future—an important backdrop for AI, semiconductor and software shares.

The bottom line

The Monday decline was a collision of two repricing forces. Amodei’s proposal, reinforced by support from rival AI leaders, led investors to question whether the frontier-AI investment cycle could be paced more slowly. At the same time, oil-driven inflation fears and higher yields reduced investors’ appetite for growth and technology risk. 1512

That combination explains the severity of the selloff better than a simple conclusion that end demand for AI compute had already failed. The evidence from the session points to uncertainty over the schedule and intensity of AI spending—and a broad unwind of richly valued AI-linked positions—rather than a confirmed collapse in the long-term AI infrastructure thesis. 5