AI Affairs, home

Wednesday 28 October 2026

Markets

OpenAI plans to buy Google AI chips after renting them, Reuters reports

The proposed purchase would extend a chip-supply relationship between two AI rivals. OpenAI is also one of Nvidia’s largest GPU buyers.

OpenAI's headquarters building in San Francisco, with palm trees and overhead wires
Photo: Coolcaesar, CC BY 4.0, via Wikimedia Commons (cropped)

OpenAI, one of Nvidia’s largest GPU buyers, plans to buy Google’s AI chips after a rental move reported in June 2025, Reuters reported on 27 October, citing technology publication The Information’s account of a person familiar with the plan. A purchase would extend OpenAI’s chip-supply relationship with Google, a rival in AI products and search.

Key points

  • The Information reported a planned purchase of Google’s Tensor Processing Units, citing a person familiar with the matter.
  • Reuters reported in June 2025 that OpenAI had begun renting Google’s AI chips.
  • OpenAI uses Nvidia GPUs to train models and run inference computing.
  • Google began selling TPUs directly to some customers in 2026.

OpenAI’s proposed TPU purchase

The chips in the reported plan are Google’s Tensor Processing Units, or TPUs. The proposed transaction would take the relationship beyond renting chips, a practice OpenAI had begun by June 2025. Google and OpenAI also compete in AI products and search.

Reuters reported that The Information gave no financial terms or other details of the planned purchase. OpenAI declined to comment, while Google did not immediately respond to a Reuters request made outside regular business hours. The reported plan concerns a purchase, rather than a completed transaction.

OpenAI already uses Google’s cloud services for products as well as rented chips. It has brought GPT-6 Astra and Codex to Google Cloud, AI Affairs has reported. The proposed TPU purchase concerns another part of the relationship: the hardware used for AI computing.

Google expands direct TPU sales

Reuters reported that Google’s TPUs have gained traction as an alternative to Nvidia’s AI chips as companies seek cheaper options suited to running AI models. Google began selling TPUs directly to some customers in 2026, adding purchases to the ways outside companies could obtain its chips.

Oracle plans to offer Google TPUs on its cloud platform in 2027, AI Affairs has reported. That plan concerns access through Oracle’s cloud, while The Information’s reported OpenAI plan concerns buying chips from Google.

For Google, a sale to OpenAI would involve a customer that also competes with it in AI products and search. The two companies already have a chip-rental relationship, according to Reuters’ June 2025 reporting. The Information’s account places a proposed purchase alongside that earlier use of Google hardware.

Nvidia remains an OpenAI chip supplier

OpenAI is among Nvidia’s largest GPU buyers. It uses those processors both to train AI models and to run inference computing, the work performed when a model responds to requests. The planned Google purchase would add another chip supplier to a computing operation that already relies heavily on Nvidia hardware.

OpenAI has a separate chip-financing arrangement involving another industry supplier. Broadcom arranged more than $50 billion for OpenAI chip financing, AI Affairs has reported. That financing is a separate development from the proposed Google purchase reported by The Information.

Alphabet will report its quarterly results on 28 October, with investors looking for signs that its heavy AI investment is continuing to drive growth in its cloud business.

Sources

Topics: Chips, Data centres, Foundation models