Summary
On October 27, 2026, Oracle and Google announced that Google TPUs will be offered through Oracle Cloud Infrastructure (OCI), making OCI the first hyperscaler outside Google to provide TPUs; the service is expected to be available in 2027 with 8th-generation TPU 8t training and TPU 8i inference chips.
Event Overview
- Date: October 27, 2026 (local time), at Oracle AI World in Las Vegas.
- Parties: Oracle and Google.
- Event: In an official press release, Oracle said Google TPUs will be available on OCI, which will become the first hyperscaler outside Google to offer Google TPUs; the service is expected to be available in calendar year 2027 and ramped over a multi-year period.
Key Information
| Item | Detail |
|---|---|
| Venue | Oracle AI World, Las Vegas, October 27, 2026 |
| Positioning | First hyperscaler outside Google to offer Google TPUs |
| Availability | Calendar year 2027, ramping over a multi-year period |
| Chips | 8th-generation Google TPU: TPU 8t (training) and TPU 8i (inference) |
| TPU 8t performance | Nearly 3x the compute performance per pod versus the previous generation |
| Development cycle | Frontier-model development cycles shortened from months to weeks |
| Infrastructure split | OCI provides power, cooling and datacenter networking for large-scale TPU deployment |
Background and Impact
- Per Oracle: Oracle Cloud president Mahesh Thiagarajan said OCI will provide power, cooling and datacenter networking to deploy TPUs at scale within its infrastructure.
- Chip details: Per Oracle, the 8th-generation TPU has two architectures — TPU 8t for training and TPU 8i for inference; the TPU 8t delivers nearly 3x the compute performance per pod versus the previous generation, cutting frontier-model development from months to weeks.
- Procurement and customers: Per The Information as relayed by Wallstreet CN, Oracle is directly purchasing the chips, and the end users are major customers shared by the two companies; their identities were not disclosed.