What happened
On 27 October 2026 Google Cloud and Oracle announced a partnership to bring Google Tensor Processing Units (TPUs) to Oracle Cloud Infrastructure (OCI). The announcement, published on the Google Cloud blog and written by Oracle and Google Cloud executives, says OCI will be the first hyperscaler outside Google to offer TPUs and that capacity will be deployed on OCI “starting in the calendar year 2027”.
The same day, NVIDIA and Oracle announced plans to expand OCI Superclusters with NVIDIA Vera Rubin NVL72 systems to multi-gigawatt scale and to bring NVIDIA Vera CPUs to OCI.
Key details
Both pages were read on 29 October 2026. The figures are the companies’ own claims.
- Google TPUs on OCI: deployment is planned from calendar 2027. The post lists four benefits: AI-density data centers, Oracle Acceleron networking, proximity to enterprise data, and a choice of architectures, with TPUs alongside GPUs. TorchTPU, which the post describes as open source, is part of the plan. The post gives no prices.
- Claimed TPU figures: the post says eighth-generation TPUs deliver up to 12.6 PFLOPS of native FP4 compute per chip at up to twice the performance per watt of the previous generation, over 97% goodput on large training runs across thousands of chips, and up to 80% better performance per dollar for low-latency serving.
- NVIDIA and Oracle: NVIDIA says Vera Rubin NVL72 systems and Vera CPUs are now generally available on OCI, and that it plans to introduce HGX Rubin NVL8 systems. Each NVL72 rack links 72 GPUs over NVLink. The release gives no gigawatt figure, price or date for the expansion.
- OpenAI on OCI: NVIDIA’s release says OpenAI trained and served “GPT-6 Astra” in Abilene on an OCI Supercluster with more than 100,000 NVIDIA Blackwell GPUs, and quotes OpenAI executives on using Vera Rubin and Vera CPUs.
Why it matters
Capacity for training and serving AI models is still the bottleneck for many labs, and these announcements add options on both sides: a rival chip family on a cloud that has been closely tied to NVIDIA hardware, and a large Vera Rubin build-out on the same cloud. Customers would get more choice only from 2027 in the TPU case, so the near-term effect is limited to the NVIDIA systems the companies say are already available.
All performance numbers come from the vendors. They are claims about their own hardware, not independent measurements.
Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article. Anthropic is also a Google Cloud customer; this article is based on the announcements by Google, Oracle and NVIDIA, read on 29 October 2026.
Sources
- Bringing Google TPUs to Oracle Cloud Infrastructure Primary source , Google Cloud Blog (27 October 2026)
- NVIDIA and Oracle Expand NVIDIA Vera Rubin to Multi-Gigawatt Scale and Bring NVIDIA Vera CPU to Oracle Cloud Infrastructure Primary source , NVIDIA (27 October 2026)