OpenAI plans to purchase Google’s artificial intelligence chips as it expands the computing capacity needed to train and operate its models, according to a person familiar with the matter. The planned deal would deepen ties between two companies that compete in AI products and search.
The processors, known as Tensor Processing Units or TPUs, were developed by Google more than a decade ago to make its search engine faster and more efficient. Google later adapted the technology for AI workloads, creating an alternative to the graphics processing units made by Nvidia. Google began selling TPUs directly to some customers in 2026.
OpenAI has already rented Google chips
OpenAI began renting Google’s AI processors by June 2025 and remains one of the largest buyers of Nvidia GPUs, which it uses for both training and running AI models. Buying Google’s chips would extend a relationship that has already included cloud rentals and other commercial agreements.
- OpenAI plans to purchase Google TPUs after previously renting them.
- Google began selling the processors directly to selected customers in 2026.
- OpenAI is developing a separate custom chip with Broadcom for deployment by the end of 2026.
The companies have also expanded their cooperation even as they compete in important parts of the technology market. OpenAI said last week that it would offer its leading AI models through Google Cloud. Oracle is also planning to make Google TPUs available through its cloud infrastructure, giving customers another way to access the processors.
A separate chip effort
OpenAI is developing its first custom AI processor with Broadcom under the name Jalapeno. The company has unveiled the chip and said it intends to deploy it by the end of 2026 as part of a multigenerational chip program.
The internal project has links to Google’s own processor work. Richard Ho, who leads OpenAI’s internal chip team, previously worked on Google’s TPU program, and Google veterans helped create OpenAI’s in-house chips. The planned Google purchase therefore sits alongside, rather than replaces, OpenAI’s effort to build its own hardware.
Google expects most of the revenue from its TPU sales agreements to arrive in 2027. The company’s chips have become a larger part of the infrastructure market as AI developers seek computing capacity beyond Nvidia’s GPUs.
