- OpenAI plans to purchase Google’s Tensor Processing Units to expand its AI training and inference capacity, according to a report by The Information.
- The move would reduce OpenAI‘s reliance on Nvidia, which has long been its primary GPU supplier.
- OpenAI has been renting Google TPUs via Google Cloud since 2025 and is adapting its models to run on them.
- Financial terms and other details of the planned purchase have not been disclosed.
What Happened
OpenAI plans to buy Google’s Tensor Processing Units to expand its AI model training and inference capacity, Analytics India Magazine reported (Oct 28, 2026), citing a report by The Information.
Why It Matters
The purchase would strengthen Google’s TPU business as a direct competitor to Nvidia’s GPUs, following Anthropic’s own multi-gigawatt Google TPU deal. It marks a further diversification of compute supply among frontier AI labs away from single-vendor dependence on Nvidia.
Technical Details
OpenAI has been renting Google TPUs via Google Cloud since 2025 and has been adapting its models to run on the chips ahead of the planned purchase. OpenAI has historically been one of Nvidia’s largest customers for training and inference workloads, and Nvidia has itself invested billions of dollars tied to OpenAI’s compute buildout.
Who’s Affected
The deal would affect Nvidia’s standing as OpenAI’s dominant chip supplier and strengthen commercial ties between OpenAI and Google even as the two compete directly in AI products. Google Cloud stands to gain a larger enterprise customer for its TPU hardware business.
What’s Next
Financial terms and other details of the proposed purchase remain undisclosed. Neither OpenAI nor Google has confirmed the report publicly, and it is not yet clear when any TPU shipments would begin.
