Oracle Cloud to Offer Google TPUs From 2027

Oracle will offer Google's Tensor Processing Units on Oracle Cloud Infrastructure starting in 2027, the two companies said on October 27 at Oracle AI World in Las Vegas. Oracle says OCI will be "the first hyperscaler outside Google to offer Google TPUs," giving its cloud customers a second family of AI accelerators alongside Nvidia GPUs.

Key takeaways

• Google TPUs will be available on OCI in calendar 2027 and ramp up over several years; Oracle has not said how many chips it will deploy.
• Oracle will host the TPUs in its own data centers and supply the power, cooling and networking.
• The deal covers Google's eighth generation: TPU 8t for training and TPU 8i for inference.

How Google TPUs on Oracle Cloud will work

According to Oracle's announcement, OCI will provide the power, cooling and data center networking needed to run TPUs at scale. Oracle pitches them for training and post-training of large dense and mixture-of-experts models, and for production inference in reasoning, agentic and long-context applications.

"Bringing Google TPUs to OCI gives customers access to multiple leading AI compute architectures in one cloud," said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure. Amin Vahdat, Google's chief technologist for AI and infrastructure, said, "TPUs give customers a powerful, purpose-built option for demanding training and inference workloads."

Oracle did not disclose how many TPUs it will start with, The Register reported. Pricing was also not announced.

TPU 8t and TPU 8i specs

Google introduced the eighth-generation chips at Cloud Next in April. In a Google blog post dated April 22, the company said a TPU 8t superpod scales to 9,600 chips and 121 exaflops, with 2 petabytes of shared high-bandwidth memory and nearly three times the compute per pod of the previous generation.

TPU 8i carries 288 GB of high-bandwidth memory and 384 MB of on-chip SRAM, three times the prior generation. Google claims 80% better performance per dollar for TPU 8i and up to twice the performance per watt of Ironwood for both chips.

Context: TPUs leave Google's own cloud

Until now, customers rented TPUs through Google Cloud. The largest outside commitment came from Anthropic, which said on October 23, 2025, that it would expand to as many as 1 million TPUs in a deal worth tens of billions of dollars, bringing well over a gigawatt of capacity online in 2026. Reuters, citing The Information, reported this week that OpenAI plans to buy Google TPUs; neither company has confirmed it.

Oracle is not dropping Nvidia. At the same event it announced general availability of Nvidia Vera Rubin NVL72 systems on OCI, and expanded its OpenAI partnership with Oracle AI Workspace, which connects ChatGPT to Oracle Fusion, NetSuite and Oracle's database. "Different AI workloads have different compute needs," Thiagarajan said, as quoted by The Register. Hyperscaler GPU orders keep growing, as Amazon's tripled Nvidia commitment to 2 million GPUs showed.

What's next

The first OCI TPU capacity is due in 2027. The open questions are scale, price and which regions get the chips first, details Oracle has yet to publish.

FAQ

When can customers use Google TPUs on Oracle Cloud? Oracle says in calendar 2027, with capacity growing over several years.

Which TPUs will OCI offer? Google's eighth generation: TPU 8t for training and TPU 8i for inference.

Does this replace Nvidia GPUs on OCI? No. Oracle launched Nvidia Vera Rubin NVL72 systems at the same event and presents TPUs as an added option.

References

• Oracle Brings Google TPUs to Oracle Cloud Infrastructure — Oracle (Oct. 27, 2026)
• Our eighth generation TPUs: two chips for the agentic era — Google (Apr. 22, 2026)
• Expanding our use of Google Cloud TPUs and Services — Anthropic (Oct. 23, 2025)
• Big Red gets cozy with Google to offer TPUs to AI customers — The Register (Oct. 27, 2026)