Andrew Feldman, CEO and co-founder of Cerebras Systems, will discuss whether AI infrastructure can continue scaling at TechCrunch Disrupt 2026 in San Francisco. His October 13-15 appearance follows Cerebras' May 2026 public listing, $5.55 billion in fundraising and disclosure of a $25 billion backlog. Cerebras' WSE-3 wafer-scale engine (a processor built from an entire silicon wafer) contains roughly 4 trillion transistors and 900,000 cores, positioning the company around high-speed inference (running trained models to produce outputs) rather than conventional GPU-led training. The company has a multi-year inference contract with OpenAI and a September 29, 2026 partnership with General Compute for agentic coding systems scheduled to come online in Q1 2027. The backlog represents commitments rather than recognized revenue, while customer concentration remains a business risk. Cerebras is competing as data centers face power, cooling and land constraints, alongside Google, Amazon and Groq. The General Compute deployment will offer an early test of whether Cerebras' architecture can translate commercial commitments into operating infrastructure.