OpenAI has signed a multi-year partnership with AI chipmaker Cerebras under which Cerebras will supply 750 megawatts of computing capacity to OpenAI from 2026 through 2028, the companies announced on Wednesday. The agreement, valued at more than $10 billion, marks one of the largest infrastructure deals yet in the rapidly intensifying race to secure AI compute.
Both companies said the partnership is aimed at significantly improving the speed and responsiveness of OpenAI’s models, particularly for tasks that currently require longer processing times. In a blog post, OpenAI said the Cerebras systems would accelerate complex inference workloads and reduce latency for end users. Cerebras co-founder and CEO Andrew Feldman compared the shift to a fundamental technological leap, saying that “just as broadband transformed the internet, real-time inference will transform AI.”
The partnership builds on an existing relationship between the two companies. OpenAI has previously evaluated Cerebras’ technology, considered acquiring the company, and OpenAI CEO Sam Altman is already an investor in the startup, according to people familiar with the matter. These prior connections helped lay the groundwork for the current deal, which formalizes Cerebras’ role as a major infrastructure supplier to OpenAI.
Founded more than a decade ago, Cerebras designs specialized chips and systems optimized for artificial intelligence workloads. The company argues that its architecture can outperform traditional GPU-based systems, such as those made by Nvidia, especially for large-scale inference and model deployment. Its profile has risen sharply since the launch of ChatGPT in 2022 and the subsequent surge in demand for AI services.
Despite filing confidentially for an initial public offering in 2024, Cerebras has repeatedly delayed its IPO amid volatile market conditions. Instead, it has continued to raise private capital. Earlier this week, reports said the company was in talks to raise an additional $1 billion at a valuation of around $22 billion, underscoring investor appetite for alternative AI hardware providers.
