OpenAI signs multi-year compute deal with Cerebras

On Wednesday, OpenAI announced that it has reached a multi-year agreement with AI chipmaker Cerebras to secure compute. In the context of artificial intelligence, compute refers to the processing power required to run and train complex models. Under the terms of the partnership, Cerebras will deliver 750 megawatts of compute to OpenAI. The delivery of this processing power is scheduled to begin this year and continue through 2028. According to a source familiar with the details, the deal is worth over $10 billion, a transaction value that was also reported by Reuters.

The partnership is designed to enhance OpenAI’s capabilities in inference, which is the process of running a trained AI model to generate predictions or responses for users. OpenAI representative Sachin Katti explained that the company’s compute strategy is to build a resilient portfolio that matches the right systems to the right workloads. Katti detailed how the partnership supports this strategy: “Cerebras adds a dedicated low-latency inference solution to our platform. That means faster responses, more natural interactions, and a stronger foundation to scale real-time AI to many more people.” This low-latency solution provides minimal delay in processing or response times, which is critical for real-time applications. Cerebras asserts that its systems are faster than GPU-based systems—computing systems utilizing Graphics Processing Units, such as those offered by Nvidia. Andrew Feldman, the co-founder and CEO of Cerebras, emphasized the importance of this speed, stating that just as broadband transformed the internet, real-time inference will transform AI.

The agreement comes as Cerebras continues to raise capital, following the launch of ChatGPT in 2022. Although Cerebras filed for an IPO—an initial public offering—in 2024, the company has continued to raise capital in the private market. Cerebras is currently in talks to raise another billion dollars at a valuation of $22 billion. While OpenAI CEO Sam Altman is an investor in Cerebras, the compute agreement is a corporate partnership between the organizations rather than a personal transaction.

Why it matters

OpenAI is diversifying its infrastructure beyond standard GPU providers to optimize for low-latency inference, signaling a strategic shift in how the company scales real-time AI capabilities.