OpenAI reportedly considers Google TPUs to reduce Nvidia dependence

Bloomberg reported on October 27 that OpenAI is reviewing adoption of Google Tensor Processing Units and considering buying the chips. Neither company confirmed the discussions, and scale and terms were not disclosed. OpenAI posted a job listing seeking an engineer to lead TPU optimization of its frontier models. Google began developing TPUs more than a decade ago for search. Richard Ho leads OpenAI’s in-house chip team and previously worked on Google’s TPU program. OpenAI has committed to spend about $600 billion on Nvidia servers through 2030. Nvidia agreed in August to provide more than $100 billion in credit support for OpenAI’s Ohio data center project. OpenAI is also developing its own chip, Jalapeno, with Broadcom, planned for later this year. An OpenAI spokesperson previously said the company “has no plans to use TPUs specifically.” On October 27 Google and Oracle announced that Oracle will offer Google TPUs through its cloud infrastructure. Oracle is buying the TPUs directly; an end user shared by both companies has not yet been identified. OpenAI last week announced it would offer model services through Google Cloud. Whether any TPU purchase closes, and on what terms, remains unconfirmed.