We are the Frontier
U.S.

Google TPUs to Join Oracle Cloud Infrastructure in 2027

The partnership will offer Google's custom AI hardware via Oracle's cloud platform to support high-scale compute requirements.

Updated on Oct. 28, 2026 in Data Centers

Google TPUs to Join Oracle Cloud Infrastructure in 2027

Live Poll

Will expanded cloud availability of specialized AI hardware significantly lower costs for business users in 2027?

Google and Oracle announced a partnership on October 27, 2026, to bring Google Tensor Processing Units (TPUs) to Oracle Cloud Infrastructure by 2027. This move marks the first time Google will provide its proprietary AI hardware on the Oracle platform, which currently supports Nvidia Vera Rubin systems.

Why it matters

The collaboration aims to address the rising demand for diverse high-scale compute architectures required by next-generation artificial intelligence. By integrating Google TPUs into its infrastructure, Oracle diversifies its high-performance cloud offerings beyond existing Nvidia hardware.

Oracle Cloud Infrastructure will provide the necessary power, cooling, and networking to host Google TPU hardware. This complements Oracle’s current high-performance catalog, which already features Nvidia Vera Rubin systems and Vera CPUs.

The players

Google

A subsidiary of Alphabet that develops proprietary Tensor Processing Units for high-scale machine learning workloads.

Oracle

A cloud service provider that operates Oracle Cloud Infrastructure and currently hosts Nvidia Vera Rubin and Vera CPU systems.

The details

Oracle Cloud Infrastructure will host Google’s Tensor Processing Units—specialized hardware accelerators designed specifically for large-scale machine learning tasks. While Alphabet currently supplies these systems through on-premises agreements, this new arrangement transitions the hardware to a cloud-hosted model. Oracle will manage the underlying physical environment, including the power and data-center networking required to sustain these compute-intensive workloads.

Timeline

  1. Q2 2026: Alphabet began recognizing revenue from existing on-premises TPU agreements.

  2. October 27, 2026: Google and Oracle announced the cloud hosting partnership.

  3. 2027: Google TPUs are scheduled to become available on Oracle Cloud Infrastructure.

  4. 2027: Alphabet expects the majority of revenue from on-premises TPU agreements.

The Tech Race

Oracle is expanding its OCI Superclusters toward multi-gigawatt scale to compete with dominant cloud providers. This integration of Google TPUs provides a strategic alternative to the Nvidia Vera Rubin systems that currently anchor Oracle’s high-performance AI offerings.

Cloud architects and machine learning engineers will gain access to Google's TPU architecture within Oracle’s network by 2027. This provides a new hardware option for scaling AI workloads without the need for on-premises infrastructure procurement.

The takeaway

This partnership signals a shift toward heterogeneous computing environments where cloud providers offer multiple proprietary AI accelerators. Watch for further details on technical performance benchmarks and regional availability as the 2027 launch date approaches.

Further reading

For broader trends in enterprise hosting, visit Data Centers.

Live Poll

Will expanded cloud availability of specialized AI hardware significantly lower costs for business users in 2027?