Andrew Feldman, CEO and co-founder of AI chipmaker Cerebras Systems, pushed back on the AI doomsday narrative during his TechCrunch Disrupt 2026 appearance, arguing that the negativity isn’t doing the industry any favors. Feldman joked that “some of the leaders in the AI world maybe shouldn’t be allowed to talk in public,” referencing warnings that have spread through the industry — from models breaking containment to AI insiders publicly debating the percentage chance that AI could kill all of humanity by the end of the decade. The argument has become mainstream enough that Anthropic CEO Dario Amodei was recently parodied on Saturday Night Live, where a fictional character summed up the discourse: “We do not condone what we are doing.”

Feldman’s position is that AI leaders should spend more time telling people about the upside. He suggested messaging around a future where your grandchildren “never meet anybody who dies of cancer,” even if those same leaders hold private concerns about AI risk. The article notes Feldman has a direct stake in AI’s success: Cerebras customers include OpenAI, Block, Figma, pharmaceutical giant GSK, and Lovable. Still, he framed his view as pragmatic rather than dismissive. “I don’t think there are powerful technologies that don’t come with the challenge of ‘what happens when powerful technologies are in the hands of bad people?'” he said, adding that “you can’t make one and not have the other.” He acknowledged the technology is powerful, said the industry must be thoughtful about it, and called for reasonable regulation.

Feldman also tied some of the negativity around AI to how the data center industry handled the AI-driven buildout. He admitted the industry “wasn’t super good for the past few years at working with local communities, particularly in the U.S.,” and that it failed to engage communities the way the power, oil, and gas industries have. He contrasted U.S. NIMBY-ism with countries like Brazil, Canada, Finland, Iceland, Norway, Mexico, and Sweden, where governments are actively encouraging companies to build data centers.

Finally, Feldman said he isn’t concerned that efficiency improvements would reduce demand for Cerebras chips, or for chips from other Nvidia competitors. “We can do our job as hardware vendors to make our compute more efficient, and then the model makers can do improved work in making their models more efficient to achieve the same goals,” he said. The combined gains, in his view, are what produce meaningful strides forward, rather than a zero-sum trade-off between hardware and model efficiency.