OpenAI plans to buy Google’s Tensor Processing Units, The Information reported on Tuesday, citing a person familiar with the matter. The report did not disclose the size of the purchase or any other terms. OpenAI declined to comment, and Google did not immediately respond to a request for comment, according to Reuters reporting of the story.
The plan, if it proceeds, changes the nature of the relationship. Reuters reported in June 2025 that OpenAI had begun renting Google’s AI chips through Google Cloud. A direct purchase would put Google silicon on OpenAI’s own balance sheet, alongside the Nvidia graphics processors OpenAI buys in large volume and the custom chip it is developing separately.
There is one piece of supporting evidence outside the unnamed source. The Information pointed to OpenAI job postings for engineers to lead work optimising its frontier models for TPUs, and described that optimisation effort as still at an early stage. Porting frontier models to a second architecture is expensive engineering, and companies rarely fund it for hardware they do not intend to run at meaningful scale.
The report lands the same week Google widened TPU distribution on two other fronts, with Oracle set to offer TPUs on its cloud from 2027 and OpenAI itself making its frontier models available through Google Cloud. Until OpenAI or Google confirms a purchase, the story should be read as a reported intention, not a signed deal.