News that OpenAI plans to buy Google's TPU chips emerged on 28 October 2026, in a Reuters report citing The Information. According to the report, OpenAI plans to purchase AI chips from Google, with financial terms and other details undisclosed. Google did not immediately respond to a request for comment, while OpenAI declined to comment. The story therefore remains a single-source account of a plan, not yet confirmed by either company.
What the report says
According to the report, OpenAI is one of the largest buyers of Nvidia graphics processors, which it has long used to train models and run inference. Google's Tensor Processing Units (TPUs) are seen as an alternative, and Google began selling TPUs directly to some customers this year. Reuters reported in June 2025 that OpenAI had already begun renting Google's chips. The report also mentions that OpenAI is hiring an engineer to optimise its frontier models to run on TPUs, a step often read as an early signal of a purchasing relationship — though a job posting alone does not prove a deal has been reached.
Why now
Last week, OpenAI announced that its frontier models would be offered through Google Cloud. Two companies that compete directly in products and search are deepening their commercial ties. For OpenAI, an additional compute route brings bargaining power on price and supply, and reduces dependence on a single vendor. For Google, winning a buyer like OpenAI would be the most significant endorsement yet for its external TPU sales business. Alphabet reports quarterly results on 29 October, when progress in cloud and TPU orders will also be in focus.
Who is affected
Nvidia faces the largest short-term impact. Losing exclusivity is not the same as losing orders: OpenAI remains one of Nvidia's most important customers, and TPU purchases would more likely supplement than replace GPUs. For other model developers, if OpenAI does move some workloads onto TPUs, the credibility of TPUs as a viable alternative to Nvidia would rise, and cloud compute pricing could loosen. Three questions come next: whether either company confirms a deal, how large any purchase would be, and which workloads would move to TPUs first. Until those are answered, this should be read as a plan, not an established fact.
