One weekend essay, and by Monday half the AI supply chain was in the red. After Dario Amodei called for slowing the pace of development and Sam Altman and Elon Musk backed him, Wall Street sorted out who depends on things staying fast.
Selling what the buildout needs
Micron, Intel, Marvell and Applied Materials each gave up more than 4%, Nvidia about 3%. SK Hynix fell 7% in U.S. trading, SoftBank 10% in Japan. Among the server vendors, Hewlett Packard Enterprise took an 8% hit, Dell and Oracle roughly 4% each. CoreWeave, flagged by Bernstein analysts as one of the most exposed names, dropped 5%.
The logic is simple enough. If you sell chips and servers for the data center buildout, you live off the pace. Train more slowly and you get dampened growth and full warehouses.
Cleaning up pays better
Security stocks did the exact opposite. Palo Alto Networks gained 14%, CrowdStrike 15%. Okta, Zscaler, Qualys and Netskope also rose double digits. Salesforce, Adobe and ServiceNow gained too.
What stands out is how deep these firms already sit in the process: per CNBC, CrowdStrike and Palo Alto have participated in early testing of unreleased OpenAI and Anthropic models. Calls for stronger defenses against AI-driven attacks got loud over the summer, after OpenAI disclosed that its own agents had orchestrated an attack on Hugging Face.
What the analysts make of it
Zoe Gillespie of RBC Brewin Dolphin, speaking to CNBC: the equity rally of recent years was built on AI growth and productivity gains. If that starts to derail, it hits future earnings – and with them, the share prices.
Ben Barringer of Quilter Cheviot is calmer about it. Even if training and rollout slow, inference is where demand shows up. And Amodei didn’t call for a stop, he called for a brake: progress will “still seem fast.” Altman added the same thing on X on Monday – pacing does not mean stopping.
The market is pricing in cleanup work
With over a trillion dollars a year in infrastructure spending, a shift in expectations alone is enough to produce a day like this. But the interesting part of Monday isn’t the drop. It’s the counter-move.
When security stocks climb double digits while chip names fall, the market isn’t reading Amodei’s essay as “less AI.” It’s reading it as “more money for control, monitoring and auditing.” Which, to be fair, is what the essay says: embedded evaluators, common standards, safety cases. That costs money, and somebody earns it. For Anthropic, heading for the Nasdaq, the timing is awkward all the same.
Sources: