A large mechanical brake lever pulled halfway down beside a glowing server rack emitting a soft pulse of light, with a speedometer dial frozen mid-sweep mounted on the wall behind, close-up shot from a low angle on a concrete data center floor, cold blue key light from above mixed with a faint amber glow from the right, deep shadows and shallow depth of field.

AI Labs Hit the Brakes, Chip Stocks Pay the Price

By Oliver BennsSeptember 15, 2026

Anthropic and OpenAI have called for slowing the pace of AI capability development, with Anthropic CEO Dario Amodei warning that rogue AI agents could overrun the internet within a year. OpenAI is reportedly earmarking up to 20% of internal compute for safety monitoring, and Sam Altman ruled out a 2026 IPO. The pivot rattled AI-adjacent stocks on Monday: Nvidia fell 3%, Oracle 4%, CoreWeave 7% and SoftBank 14%. The White House pushed back, with President Trump dismissing safety advocates as negative forces and insisting whoever wins AI wins.

Anthropic Claims Profitability, But the Math Looks Fuzzy

Anthropic has told investors ahead of a planned IPO that it has been profitable for two straight quarters, a milestone that would contrast sharply with OpenAI's reported $20bn in operating losses last year. The claim rests on adjusted operating figures, with gross margins pitched above 80% before accounting for revenue-sharing deals with partners like Amazon and the cost of training models - the company's two largest expense lines. Full clarity awaits the S-1 filing.

CrowdStrike Rips 14% as Cyber Stocks Catch a Bid

CrowdStrike jumped 14% on Monday as investors rotated into cybersecurity names, bucking a broader selloff in AI-adjacent equities. The move came as frontier labs flagged the risk of AI-driven cyber threats accelerating once open-source models catch up to closed-source capabilities over the next six to nine months.