Signals›Chips & compute
Chips & computeShipping

Nvidia's next data-center platform, Vera Rubin, is now generally available on Oracle's cloud.

Rubin GPUs and Nvidia's new Vera CPU can now be rented on Oracle Cloud. The "multi-gigawatt" expansion is a plan; the racks are real.

ZN ZilverNet EditorialPublished · 4 min read · 1 source Generally available Oct 27, 2026
The Oracle and NVIDIA logos side by side
Image credit: Nvidia
The short versionShipping
What it is
Vera Rubin NVL72 — racks connecting 72 Rubin GPUs through Nvidia's NVLink — and Nvidia's new Vera CPU, both generally available on Oracle Cloud Infrastructure.
How far along
ResearchPrototypePilotShippingEverywhere
Shipping on Oracle Cloud as of Oct 27. A "multi-gigawatt" expansion and smaller eight-GPU servers are announced as plans.
Why it matters
For most companies, the cloud is where a new GPU generation first becomes something you can actually use.
Still unproven
The announcement gives no performance figures. Vera CPU benefits such as "reduce GPU idle time" come without numbers.

What's available now

On October 27, Nvidia and Oracle said "NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPUs Are Now Generally Available on OCI." Each NVL72 rack connects 72 Rubin GPUs through NVLink so they can work as one large system, linked into Oracle's wider network of switches and Nvidia's BlueField-4 data processors.

Vera is Nvidia's new CPU. Nvidia says it offers "high single-thread performance" and helps "reduce GPU idle time" — keeping the expensive GPUs fed with work. The announcement doesn't quantify either.

What's planned

  • Multi-gigawatt scale. Oracle plans to expand its Vera Rubin clusters "to multi-gigawatt scale." No figure is given.
  • HGX Rubin NVL8. Smaller servers with eight Rubin GPUs, two 64-core Intel Xeon 6 CPUs, 4 TB of memory and 6.4 Tbps of networking.
"AI factories must be productive, durable and fungible." — Ian Buck, VP of hyperscale and HPC, Nvidia

The customer in the room

OpenAI features heavily in the announcement. It says it "has trained and served GPT-6 Astra on an OCI Supercluster with more than 100,000 NVIDIA Blackwell GPUs" in Abilene, Texas — note that is Blackwell, the current generation, not Rubin. OpenAI is an early-access customer for the Vera CPU.

"We are bringing Vera into production on OCI to support these workloads at scale." — Uday Ruddarraju, CTO of compute, OpenAI

Where all this compute goes is the subject of Gartner's forecast: this year, spending on running AI overtakes spending on training it.

Update log
  • Published.