Google TPUs Are Set to Join Oracle Cloud in 2027 Under New Deal

Google and Oracle will make Google’s Tensor Processing Units (TPUs) available through Oracle Cloud Infrastructure (OCI) in 2027, adding a hyperscale cloud channel beyond Google’s own platform.

The companies announced the agreement Oct. 27. OCI will host the TPUs and provide power, cooling and data-center networking. OCI is set to be the first hyperscale cloud outside Google to offer the accelerators, giving customers another cloud option for accessing them. The announcement did not specify an initial TPU model, launch quarter, chip count, pricing, customer orders or expected revenue.

“The next generation of AI will require a diverse range of compute,” Amin Vahdat, Google’s senior vice president and chief technologist for AI and Infrastructure, said.

The cloud deal follows Alphabet’s earlier TPU system agreements for specialized, high-scale workloads using on-premises infrastructure. Alphabet began recognizing revenue from a limited number of those agreements in the second quarter of 2026 and expected the significant majority of their revenue in 2027. Those agreements involve supplying systems for on-premises workloads; the OCI deal makes TPUs available through a cloud provider that supplies the hosting infrastructure.

Oracle is also expanding its Nvidia offerings on OCI. Nvidia Vera Rubin NVL72 systems and Vera CPUs are generally available on the platform. Oracle plans to expand OCI Superclusters using Vera Rubin systems to multi-gigawatt scale.

TPU availability on OCI is scheduled for calendar year 2027, with capacity set to ramp over multiple years.