OpenAI Inks $10B Cerebras Deal for 750 MW AI Compute

OpenAI has signed a multiyear agreement with AI chipmaker Cerebras Systems to deploy up to 750 megawatts of computing power, valued at more than $10 billion, in a move aimed at accelerating response times for its AI services through 2028.

The deal, announced Wednesday, calls for Cerebras to provide its waferscale processor systems starting in early 2026 and scaling up over multiple phases. These systems specialize in low-latency inference, the process of generating quick outputs from AI models, which OpenAI officials say will enable faster interactions on platforms like ChatGPT. Deployment will occur across Cerebras data centers in the United States and internationally, with the company planning to build or lease additional facilities to support the capacity.

OpenAI's compute infrastructure lead, Sachin Katti, described the partnership as adding a "dedicated low-latency inference solution to our platform," noting it would lead to "faster responses, more natural interactions, and a stronger foundation to scale real-time AI to many more people." Cerebras CEO Andrew Feldman echoed the potential impact, stating that "just as broadband transformed the internet, real-time inference...