OpenAI has entered a multi-year agreement with AI chipmaker Cerebras Systems to significantly expand its compute capacity, deepening its infrastructure stack as demand for real-time AI continues to surge. Under the deal, Cerebras will supply 750 megawatts of compute beginning in 2026 and running through 2028. A source familiar with the transaction confirmed the agreement is valued at more than $10 billion.
The partnership is designed to accelerate low-latency inference across OpenAI’s products, enabling faster responses for workloads that currently require longer processing times. Andrew Feldman, co-founder and CEO of Cerebras, has positioned real-time inference as a structural shift in how AI systems are delivered, comparable to the impact broadband had on the internet.
For OpenAI, the agreement supports a diversified compute strategy aligned to specific workloads. Sachin Katti, who leads infrastructure initiatives at OpenAI, has emphasized that Cerebras adds a dedicated inference layer to improve responsiveness and scalability across OpenAI’s platform.
