“OpenAI has only two AI accelerator compute vendors in production today, Cerebras and Nvidia,” Cerebras CEO Andrew Feldman says. Four days after Cerebras went public, Feldman joined Bloomberg Intelligence’s Kunjan Sobhani to discuss the company’s next chapter and the rapidly shifting AI infrastructure landscape. Feldman breaks down the OpenAI deal, the strategic AWS partnership around disaggregated inference and why Cerebras believes fast inference is becoming the industry’s defining battleground. He explains how Cerebras evolved from building the world’s largest chip to operating one of the fastest inference platforms, why disaggregated inference could reshape hyperscale AI deployments and how the company is navigating power, memory and data-center constraints. The episode also explores the competitive landscape beyond GPUs and Feldman’s broader perspective on the next phase of AI compute.