Cerebras CEO pushes back on AI doomsday warnings

Cerebras Systems CEO and co-founder Andrew Feldman is pushing back on the now-familiar AI doom loop. Speaking at TechCrunch Disrupt 2026, he said some people leading the AI conversation maybe shouldn’t be speaking in public at all, a jab at the constant stream of worst-case talk around model safety and existential risk.

That talk has gotten loud enough to spill into the mainstream. Feldman pointed to everything from warnings about models breaking containment to arguments inside the industry over whether AI could wipe out humanity by the end of the decade. Even Anthropic CEO Dario Amodei got dragged into the culture war, with a recent “Saturday Night Live” sketch parodying the tone of the debate.

Feldman’s complaint isn’t that AI has no risks. It’s that the danger talk is crowding out the pitch for what the technology might actually do. He said leaders should spend more time explaining the upside — including a future where, in his example, grandkids might never meet anyone who dies of cancer — while still acknowledging the need for restraint and reasonable regulation.

There’s a clear business reason for that optimism. Cerebras sells to companies working on AI, including OpenAI, Block, Figma, GSK, and Lovable. But Feldman also argued that powerful technologies always come with the possibility of misuse, and that the answer is not to pretend otherwise. The answer, in his view, is to build them carefully and regulate them sensibly.

He also tied some of the backlash to the way the data center rush has been handled. Feldman said the industry has not done a good job of working with local communities in the U.S., and he compared that with countries such as Brazil, Canada, Finland, Iceland, Norway, Mexico, and Sweden, where he said companies are being encouraged to build. And he was relaxed about efficiency gains reducing chip demand, saying hardware makers and model builders can both get better at doing more with less.