The central question for AI's future is whether compute, energy, and infrastructure can keep pace with model capabilities. Cerebras CEO Andrew Feldman will address this at TechCrunch Disrupt 2026, drawing on his company's decade-long bet on wafer-scale computing.
Cerebras's approach differs from conventional chip design: instead of cutting wafers into individual chips, it builds a processor on the wafer itself for AI workloads. The company has scaled commercially, raising $5.5 billion in its May IPO and signing a multiyear deal with OpenAI for 750 megawatts of systems from 2026 through 2028. In August it launched the CS-4, its latest wafer-scale infrastructure.
The company is also investing in the physical side of scaling. It reports more than 600 megawatts of data center capacity live or under contract by end of 2027, plans to increase manufacturing capacity more than tenfold in 2026, and expects to bring its first European data center capacity online this year, expanding to 200 megawatts there by end of 2027. Feldman's session is part of Disrupt, running October 13-15 at Moscone West in San Francisco.