Cerebras Systems’ Andrew Feldman on whether AI can keep scaling at TechCrunch Disrupt 2026

AI model capacity continues to rise, and each advance places heavier demands on compute capacity, energy supply, cooling systems, and manufacturing throughput. At TechCrunch Disrupt 2026, Cerebras Systems CEO Andrew Feldman will address those system-level constraints and explain how his company’s wafer-scale approach differs from conventional chip architectures.

Cerebras rejects the common practice of slicing a silicon wafer into many small chips. Instead, the company has built processors at wafer scale, a design it says is optimized for large AI workloads. Feldman and his team have scaled that architecture through on-premise systems and a cloud platform; they also introduced the CS-4 generation in August.

The company’s recent commercial activity underscores the scale of demand. Cerebras reported more than 600 megawatts of data-center capacity either live or under contract for delivery by the end of 2027, said it will expand manufacturing capacity more than tenfold during 2026, and plans to bring its first European data-center capacity online this year with a target of 200 megawatts there by the end of 2027. Cerebras also disclosed a multiyear agreement with OpenAI to deploy 750 megawatts of systems from 2026 through 2028 and completed a $5.5 billion IPO in May.

Feldman’s Disrupt session aims to put those numbers and engineering choices into context for founders, investors, and infrastructure planners: where compute demand could head next, what physical and logistical limits might appear, and how different hardware approaches could change the calculus. The talk is one of more than 200 sessions at the conference, which expects thousands of attendees and hundreds of speakers and exhibiting startups.