OpenAI Reportedly Agrees to Spend Over $20 Billion on Cerebras Chips in a New Deal

OpenAI appears to be doubling down on its AI infrastructure strategy, with a new report suggesting a massive expansion of its partnership with Cerebras Systems. This comes at a time when demand for AI compute is only getting higher.

OpenAI's Cerebras deal could scale up significantly

The news first reported by The Information (paywalled) suggests that OpenAI has agreed to spend more than $20 billion over the next three years to use servers powered by Cerebras chips (via Reuters). That's a notable jump, especially considering the company had already committed to a deal worth over $10 billion earlier this year.

In fact, back in January, OpenAI signed an agreement to buy up to 750 megawatts of computing capacity from Cerebras. Now, the newer commitment appears to be roughly double that size, hinting at how aggressively OpenAI is scaling its infrastructure.

The deal is said to focus heavily on inference workloads, which is essentially the process where AI models generate responses. With usage growing across products like ChatGPT, it's not surprising that OpenAI is locking in more compute capacity.

Equity stakes and IPO plans come into play

The report also goes on to talk about OpenAI receiving warrants in Cerebras, potentially translating into an equity stake of up to 10%, depending on how much it spends over time. On top of that, OpenAI has reportedly also agreed to invest around $1 billion to support the development of new data centers that will run these AI workloads. If the numbers hold, total spending could reach as high as $30 billion over the three-year period.

That being said, this partnership seems to be playing a key role in Cerebras' broader plans. The company is reportedly targeting an IPO in the second quarter and is also said to be planning a $3 billion raise at a valuation of around $35 billion. It's still unclear how much of this deal will be officially disclosed, but if accurate, it could mark one of the largest AI infrastructure commitments yet.

In another company news, OpenAI yesterday also announced a new biology-focused AI model: GPT-Rosalind and has also expanded Codex beyond coding, a move that comes as a response to Claude Opus 4.7's launch.