AI leaders should discuss the technology’s risks without letting warnings crowd out its potential benefits, Cerebras Systems CEO and co-founder Andrew Feldman argued at TechCrunch Disrupt 2026. He also pointed to community engagement around data centers as one way the industry can address public concerns.
Make room for AI’s possible benefits
Feldman criticized what he sees as an overly negative narrative about artificial intelligence. He joked that “some of the leaders in the AI world maybe shouldn’t be allowed to talk in public.” His comment referred to concerns ranging from models breaking containment to debate among AI insiders about whether AI could kill all of humanity by the end of the decade.
That debate has become familiar enough to be parodied on Saturday Night Live. In a skit featuring a fictional version of Anthropic CEO Dario Amodei, the character summed up the discourse by saying, “We do not condone what we are doing.”
Feldman’s counterpoint is that the possible gains deserve attention too. He said AI leaders should tell people about the potential for a future in which their grandchildren will “never meet anybody who dies of cancer.” The example is a hoped-for outcome, not a claim that the technology will deliver it.
His position is shaped in part by Cerebras’s place in the AI industry. The chipmaker’s customers include OpenAI, Block, Figma, pharmaceutical giant GSK, and Lovable. Feldman acknowledged the need to weigh the rewards of AI against the risks that come with powerful technology.
Power brings responsibility
Feldman said powerful technologies can be misused when they fall into the hands of bad actors. In his view, that challenge is inseparable from developing powerful tools in the first place.
“I think what we have to say is that this technology is powerful,” he said. “We have to be thoughtful about it. We have to regulate it reasonably.”
That framing makes room for both optimism and caution. Acknowledging AI’s potential does not remove the need to consider how it could be used, while discussing danger alone can leave the technology’s possible contributions out of the public conversation.
Data centers need community engagement
Feldman also connected some public resistance to AI with the way the data center industry has handled new construction driven by AI demand. He said the industry had not worked effectively with local communities, particularly in the U.S.
He compared the industry’s approach with established parts of the power sector, including oil and gas, saying data center companies had not engaged communities as they should have. In his account, that failure has been a problem and contributes to the “not in my back yard!” response to proposed projects.
Feldman contrasted U.S. resistance with markets he named as encouraging companies to build data centers: Brazil, Canada, Finland, Iceland, Norway, Mexico, and Sweden. His remarks put local engagement alongside AI safety as a part of the broader debate over how the technology should expand.
Efficiency can advance across the stack
Feldman said he was not worried that efficiency improvements would reduce demand for Cerebras chips or those made by other companies competing with Nvidia. He described progress as a shared task for hardware vendors and model makers.
Hardware companies can make computing more efficient, while model developers can improve their models to reach the same goals with greater efficiency. Feldman argued that both kinds of work need to come together to produce meaningful progress.
Together, his comments offer a view of AI’s future that holds several concerns at once: powerful technology can carry serious risks, its potential benefits remain worth discussing, and the infrastructure behind it needs better relationships with the communities where it is built.