Oracle Cloud Infrastructure (OCI) will begin offering Google's Tensor Processing Units (TPUs) in 2027, becoming the first major cloud provider outside of Google to host these chips. The partnership was announced on October 27, 2026, at the Oracle AI World event in Las Vegas.
The rollout will be phased and is expected to scale over several years. Oracle will provide access to Google’s latest generation of hardware, specifically the TPU 8t for AI model training and the TPU 8i for inference tasks. The TPU 8t is designed to deliver nearly three times the compute performance per pod compared to the previous generation.
This initiative aims to provide customers with greater architectural choice within a single cloud platform. While Oracle continues to collaborate with Nvidia on its Vera Rubin systems and maintains partnerships with Cerebras Systems and SambaNova Systems, the addition of TPUs serves as a new differentiator for OCI.
Following the announcement, Oracle shares rose by nearly 5%. The collaboration builds on an existing multicloud relationship between the two companies, which previously included the integration of Oracle Database services into Google Cloud.
