Loading page…
Google’s chips offer an alternative to Nvidia hardware, but the prospective purchase has no disclosed financial terms or demonstrated cost savings.
Listen to this story
OpenAI’s reported move would turn its existing access to Google’s TPUs from rentals into a prospective hardware purchase, adding a supply path beyond Nvidia GPUs. The talks are unconfirmed, and no volume, price, or delivery timing has been disclosed. Google began selling TPUs directly to some customers in 2026; a purchase could diversify OpenAI’s compute and potentially lower inference costs, but the report provides no evidence of savings or a completed order.
Reuters reported that OpenAI began renting Google’s AI chips in June 2025, making a purchase a shift from access to ownership.
Google began selling TPUs directly to some customers in 2026, creating a path for the proposed order.
OpenAI uses Nvidia GPUs for both model training and inference, so TPUs could add to its hardware mix rather than replace an established supplier.
OpenAI is preparing to buy AI chips from Google, The Information reported on October 27, 2026, citing a person familiar with the matter, according to Reuters. The prospective purchase concerns Google’s Tensor Processing Units, or TPUs—processors used for AI computing—and would give a major Nvidia customer another route to hardware.
The purchase is not confirmed. Financial terms and other purchase details were not disclosed. OpenAI declined to comment, while Google did not immediately respond to Reuters’ request outside regular business hours.
Reuters reported in June 2025 that OpenAI had begun renting Google’s AI chips. The newly reported step is buying the hardware, rather than first gaining access to it. Google began selling TPUs directly to some customers in 2026, providing the backdrop for a prospective OpenAI order.
OpenAI is one of Nvidia’s largest buyers of graphics processing units, or GPUs. It uses those processors both to train AI models and for inference—the computing that runs trained models for users. That makes the reported TPU plans relevant to a company already buying substantial amounts of AI hardware.
Google’s TPUs have gained traction as an alternative to Nvidia chips as companies seek cheaper options and hardware suited to running AI models. The appeal is not just access to more processors, but the possibility of a different cost profile.
Finimize frames adding TPUs alongside Nvidia hardware as a possible way to diversify OpenAI’s supply and reduce inference costs. Those are potential benefits, not measured results from this proposed purchase.
Loading discussion...
Join the conversation
Explain whether another hardware option outweighs closer reliance on a rival.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.