Google and Oracle are partnering to bring Google TPUs to Oracle Cloud Infrastructure. OCI will be the first hyperscaler outside Google to offer TPUs. Deployment will start in 2027. TPUs are co-designed for performance per watt and cost efficiency. They sustain over 97% goodput on large-scale training runs. The latest TPUs deliver up to 80% better performance per dollar for serving. TPUs will run alongside GPUs in OCI for workload flexibility. Customers can place AI workloads next to Oracle enterprise data. Oracle’s data centers are engineered for AI density and liquid cooling. Oracle Acceleron provides secure, high-performance networking for TPUs.
What changed
TPUs were only available on Google Cloud.
What it unlocks
Running Google TPUs on Oracle Cloud Infrastructure for AI training and inference.
- TPU v8 delivers 12.6 PFLOPS per chip
- 97% goodput on large-scale training
- 80% better performance per dollar for serving
- TPU deployment on OCI starts in 2027
What you need to act on it
- TPU capacity deployment on OCI beginning in 2027
Sources