OpenAI reportedly plans to buy Google TPUs after renting capacity

OpenAI reportedly plans to buy Google’s artificial intelligence chips after previously renting access to Tensor Processing Units. The reported move would shift the companies from a capacity-rental relationship toward direct hardware procurement, although the available evidence does not establish that they have signed or completed a transaction.

Neither company has publicly confirmed the plan. The potential price, chip volume, delivery schedule, TPU generation, deployment location and commercial structure have not been disclosed.

In June 2025, OpenAI was reported to have begun renting Google TPU capacity to power ChatGPT and other products. The newer claim therefore marks a reported shift from renting compute capacity toward buying chips, not confirmation that a purchase has closed.

There is separate first-party evidence that OpenAI is preparing its software for TPU hardware. An OpenAI engineering job listing seeks someone to lead optimization of the company’s frontier models on TPUs across model code, kernels and hardware execution. The listing does not identify a chip supplier or procurement agreement.

TPUs are Google-designed accelerators for AI workloads. Google and Oracle describe the chips as supporting both frontier-model training and large-scale inference, the process of running trained models to produce answers. DataPhoenix has separately covered Oracle and Google’s plan to bring TPUs to OCI in 2027. In an October 27 announcement, Oracle said the deployment would be the first time Google TPUs were offered through a hyperscale cloud provider outside Google.