Oracle and Google announced on October 27 that Google's Tensor Processing Units — custom chips designed specifically for AI workloads — will be made available through Oracle Cloud Infrastructure (OCI). This makes OCI the first hyperscale cloud platform outside of Google's own network to offer the chips.
TPUs come in two current forms: a training chip and an inference chip built for the iterative, latency-sensitive work of running AI agents and large models in production. Oracle will supply the power, cooling, and data center networking needed to deploy them at scale. A senior OCI executive described the arrangement as a multi-year, multi-gigawatt investment, with the expectation that new TPU generations will be available on OCI on the same day they appear on Google Cloud. Availability is expected to begin in 2027 and ramp over several years. Oracle also announced at the same event that it plans to add chips from Cerebras Systems and OpenAI to OCI's compute lineup.
For organizations running AI workloads on Oracle Cloud, the practical effect is more choice in compute architecture without moving to a different cloud provider. Businesses that rely heavily on AI training or inference may want to monitor when TPU availability opens and how pricing compares to existing GPU options on OCI and other platforms. Those in long-term cloud contracts should also review whether their agreements include provisions for new compute types.
Sources
Does this affect your business?
Our advisors can help you figure out the right move, at no cost.