NVIDIA and Oracle Plan Multi-Gigawatt OCI Superclusters With Vera Rubin Systems

Vera Rubin in OCI is the subject of a new joint announcement from NVIDIA and Oracle. On October 27, 2026, the companies said they plan to expand OCI Superclusters with NVIDIA Vera Rubin NVL72 systems to a multi-gigawatt scale. They also announced the general availability of Vera Rubin NVL72 systems and NVIDIA Vera CPUs in Oracle Cloud Infrastructure (OCI).

NVIDIA and Oracle published the announcement through their official channels. It is a joint plan by both companies. The companies did not disclose a specific schedule, cloud data center regions, or the investment value of the planned expansion.

Vera Rubin in OCI

According to the announcement, Vera Rubin NVL72 is part of the infrastructure Oracle is making available in OCI together with NVIDIA Vera CPUs. The announcement also connects this infrastructure with AI agent workloads.

Oracle also plans to deploy NVIDIA HGX Rubin NVL8. This is an eight-GPU configuration in an x86-architecture server. However, the published information does not state when this configuration will become available to OCI customers.

The general availability of Vera Rubin NVL72 and Vera CPUs means, according to the announcement, that they are available in Oracle Cloud Infrastructure. The multi-gigawatt expansion of the Superclusters, by contrast, remains a plan rather than completed capacity.

Multi-gigawatt plan without implementation details

The planned scale signals further expansion of OCI’s high-capacity AI infrastructure. If the plan is fulfilled, OCI would gain capacity for training and operating large models at a multi-gigawatt scale.

The published materials do not specify how much capacity would be added in each region, what the energy and construction requirements would be, or what supplier commitments would be needed for implementation. Without this information, the project’s actual financial or operational scope cannot be determined.

In the announcement, NVIDIA and Oracle also discuss benefits for efficiency, GPU utilization, training, inference, and agentic AI. These are vendor claims that have not been independently quantified in the available materials.

What to watch next

• the timing and regions of the planned multi-gigawatt OCI expansion,
• the specific capacity and technical parameters of the Superclusters being built,
• the availability of the NVIDIA HGX Rubin NVL8 configuration to Oracle customers,
• information about energy, construction, and supplier commitments.

The joint announcement also contains statements about OpenAI’s use of this infrastructure. However, separate confirmation from OpenAI regarding the deployment or extent of use of Vera Rubin systems was not available in the sources reviewed.

Verified and updated: 10/28/2026 06:21