OpenAI Plans to Procure Google TPU AI Chips

en.Wedoany.com Reported - On October 28, U.S. artificial intelligence company OpenAI plans to procure Google's self-developed Tensor Processing Unit (TPU) chips. The Information disclosed this procurement plan on October 27, citing people familiar with the matter. OpenAI is currently a major customer of Nvidia's AI GPUs, and this procurement plan involves adjusting the chip supply sources for its AI computing infrastructure.

OpenAI had previously rented TPU computing power through Google Cloud to support the operation of its AI products. This plan involves directly purchasing chips, which differs from the previous approach of obtaining computing power through cloud services. OpenAI has also posted job listings for TPU model optimization engineers, whose responsibilities include optimizing its advanced AI models for Google TPUs, covering model training and inference workloads.

Google began directly selling TPU chips to external customers in 2026. On October 27, Google and Oracle announced a partnership to deploy TPU computing power on Oracle Cloud Infrastructure (OCI) starting in 2027, with Oracle providing supporting data center power, cooling, and network infrastructure. This is a separate partnership project for expanding Google TPUs to other major cloud platforms.