Overview
- OpenAI reportedly entered private talks on Wednesday, Oct. 28, 2026, to buy Google's tensor processing units, known as TPUs, as part of a push to add more AI compute quickly.
- No financial terms or final deal details have been disclosed and both OpenAI and Google declined to comment on the reported discussions.
- OpenAI is also building its own Jalapeno chips with Broadcom and guidance from former Google engineers, and the company has said it plans to deploy those processors later in 2026.
- Oracle announced on Oct. 27 that it will offer Google TPUs through its cloud, a move Google framed as supporting a broader need for diverse compute for the next generation of AI.
- Google developed TPUs more than a decade ago for search and later adapted them for AI workloads, and wider access to TPUs could broaden suppliers beyond Nvidia and speed how fast companies scale large models.