OpenAI announced a multi-year agreement with AI chipmaker Cerebras on Wednesday. Cerebras will deliver 750 megawatts of compute to OpenAI starting this year and continuing through 2028 in a deal valued at over $10 billion, aimed at providing faster outputs for OpenAI customers.
The agreement requires Cerebras to supply substantial computing power equivalent to 750 megawatts over the four-year period from this year to 2028. A source familiar with the details confirmed the deal’s value exceeds $10 billion to TechCrunch, while Reuters also reported this figure. This compute capacity targets improvements in processing speed for OpenAI’s services.
OpenAI stated in a blog post that the new systems will accelerate time-consuming responses, speeding outputs that currently require more time to process. Cerebras co-founder and CEO Andrew Feldarthritis described the partnership’s potential by saying, “just as broadband transformed the internet, real-time inference will transform AI.” Both companies emphasized that the deal focuses on delivering faster results to OpenAI users.
Cerebras has operated for more than a decade, with its prominence increasing significantly after the launch of ChatGPT in 2023 and the ensuing AI boom. The company designs chips specifically for AI applications and claims its systems outperform GPU-based alternatives, such as those from Nvidia, in terms of speed.
Cerebras filed for an initial public offering in 2024 but has postponed the offering multiple times. Despite these delays, the company has raised substantial capital through other means. Reports on Tuesday indicated that Cerebras is in talks to secure an additional $1 billion in funding at a $22 billion valuation.
OpenAI CEO Sam Altman is an existing investor in Cerebras, and OpenAI previously considered acquiring the chipmaker. These connections precede the current compute supply agreement.
Sachin Katti of OpenAI explained the company’s compute strategy as building a resilient portfolio that matches specific systems to particular workloads. He added, “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.”
