Cerebras Systems chief executive Andrew Feldman is pushing back on the most catastrophic public warnings about artificial intelligence, arguing that the industry’s focus on danger is obscuring the technology’s possible benefits. His comments were made at TechCrunch Disrupt 2026 and reported by TechCrunch on October 13.

Feldman criticized what he sees as an overly negative narrative among prominent AI leaders. He suggested that executives should spend more time describing the benefits they expect from the technology, including the possibility of major medical advances. His prediction that future generations might no longer encounter deaths from cancer was aspirational, not a documented outcome or a timetable for a cure.

Abstract medical research lattice emerging from AI hardware with an unresolved gap.
Feldman cited possible medical advances, but his cancer-related prediction remains an aspirational forecast.

The position comes with an obvious commercial interest. Cerebras sells AI computing systems, and TechCrunch noted that its customers include companies working across software and pharmaceuticals. Greater adoption of advanced models could increase demand for the infrastructure Cerebras supplies.

Feldman did not argue that AI is harmless. He said powerful technologies inevitably create risks when placed in the hands of malicious actors. His proposed response was to recognize the technology’s power, approach it thoughtfully and regulate it reasonably rather than allowing worst-case scenarios to dominate the entire discussion.

He also linked part of the backlash to the physical expansion required by AI. Feldman said the data-center industry had not worked closely enough with local communities in the United States during recent construction campaigns. In his account, insufficient engagement helped create opposition to new facilities.

Generic data center and community space linked by an unfinished bridge.
Feldman said data-center developers need stronger engagement with the communities that host new infrastructure.

Feldman contrasted that resistance with countries he described as more welcoming to data-center investment, including Brazil, Canada, Finland, Iceland, Mexico, Norway and Sweden. TechCrunch reported his argument, but the article did not provide comparative data on approvals, energy costs or local opposition in those markets.

The Cerebras chief also rejected the idea that improvements in computing efficiency would reduce the market for AI chips. He said hardware makers and model developers can each make their systems more efficient while jointly enabling more capable products. That is an industry forecast from a chip vendor, not proof that demand will continue growing at its current pace.

Feldman’s remarks place him in a familiar but consequential debate: how AI companies should communicate uncertain risks without minimizing them or presenting speculative benefits as guarantees. His answer favors a risk-and-reward message, paired with regulation and community outreach. Whether that approach builds public trust will depend less on rhetoric than on measurable safety practices, useful products and the local consequences of the infrastructure behind them.