Back to the stories

NVIDIA and Oracle expand Vera Rubin AI infrastructure on OCI to multi-gigawatt scale

Score 9.2,

AI Infrastructure and Compute

Designing smarter models is only part of the AI race. Providers also need enough computing power to run them.

Oracle and NVIDIA plan to expand Oracle Cloud Infrastructure’s AI superclusters. They are large groups of connected computers. The planned expansion reaches a scale requiring multiple gigawatts of power. Vera Rubin systems and Vera processors are now generally available on Oracle’s cloud.

Why does that matter? Training AI, answering user requests, and running AI agents all depend on access to computing capacity. When a cloud provider adds more, organizations can build and run those workloads without owning the infrastructure themselves.

Think of a supercluster as a factory floor for AI: many computing machines work together instead of relying on one oversized computer. Oracle and NVIDIA say they can add different kinds of computing power independently, rather than expand everything at once.

For organizations using Oracle’s cloud, the new systems are available now. But the much larger expansion remains a plan, not capacity customers can assume is online today.

AI’s race is not only about software anymore. It is also a build-out of physical infrastructure. The open question is how quickly that planned scale becomes dependable capacity customers can actually use.