Where to run demanding AI workloads: Google TPUs on Oracle Cloud

Enterprises building large AI models and production inference pipelines face a hard choice: how to match each workload with the right accelerator without locking into a single cloud. Oracle and Google have now addressed that constraint by bringing Google Tensor Processing Units (TPUs) to Oracle Cloud Infrastructure (OCI) — the first time Google TPUs will be offered on a hyperscale cloud outside Google. Softprom is the official partner of Google.

What was announced

At Oracle AI World in Las Vegas on October 27, 2026, Oracle and Google announced that Google TPUs will be available through OCI. OCI will provide power, cooling and datacenter networking to run Google TPUs at scale within its infrastructure. Availability is planned for calendar year 2027, ramping over a multi-year period.

Google's eighth-generation TPU introduces two purpose-built architectures: TPU 8t for training and TPU 8i for inference. TPU 8t delivers nearly 3x the compute performance per pod over the previous generation and shortens frontier model development cycles from months to weeks.

Bringing Google TPUs to OCI gives customers access to multiple leading AI compute architectures in one cloud

Why this matters

For CIOs, CTOs and procurement leaders, the deal expands supplier choice for AI compute at a time when GPU capacity and economics are a board-level issue. TPUs are co-designed with open software to cover the full AI lifecycle: frontier training, large-scale inference and the multi-step reasoning required by agentic AI. Running them inside OCI lets organizations align each workload with the most cost-effective accelerator while keeping data and operations in a single hyperscale environment.

Technical details

• Platform: Google TPUs delivered as part of OCI AI compute portfolio
• Architectures: TPU 8t for training, TPU 8i for inference
• Performance: nearly 3x compute per pod vs previous TPU generation (TPU 8t)
• Workloads: dense and mixture-of-experts (MoE) models, reasoning, agentic and long-context inference
• Infrastructure: OCI power, cooling and datacenter networking
• Availability: calendar year 2027, multi-year ramp

FAQ

When will Google TPUs be available on OCI?

Google TPUs on OCI will be available in calendar year 2027, with capacity ramping over a multi-year period to match customer demand.

What is the difference between TPU 8t and TPU 8i?

TPU 8t is optimized for large-model training and delivers nearly 3x compute per pod over the prior generation. TPU 8i is designed for production inference, including agentic and long-context workloads.

Which AI workloads benefit most from TPUs in a hyperscale cloud?

Frontier model training, mixture-of-experts models, large-scale inference and agentic AI with multi-step reasoning benefit most, where per-watt efficiency and chip-to-chip bandwidth drive cost and latency outcomes.

Softprom and Google

Softprom is the official partner of Google. Our team helps enterprises evaluate Google Cloud and TPU-based AI infrastructure, plan migrations and build procurement scenarios for hybrid and multicloud AI stacks via Google.

Prepared by the Softprom editorial team based on vendor materials (Softprom DistriFlow). Source: original article.