Oracle launched NVIDIA Vera Rubin NVL72 and Vera CPU cloud services; expansion to multi-gigawatt cluster scale remains planned.

Oracle and NVIDIA announced the availability of Vera Rubin NVL72 systems and Vera CPU cloud services. Separately, they said OCI Superclusters using Vera Rubin NVL72 would be expanded to multi-gigawatt scale; the source does not establish that this planned scale is already deployed.

Oracle also plans to introduce HGX Rubin NVL8 systems containing eight Rubin GPUs for frontier-model training, inference, and agentic applications. Vera CPUs, Rubin GPUs, and the named system configurations are distinct components and offerings, rather than interchangeable measures of compute capacity.

A broader report says Oracle intends to use its own Acceleron networking chips in some NVIDIA servers, replacing some NVIDIA networking products. The companies describe improved GPU utilization and scalable AI compute as objectives, but the supplied reporting contains no measured utilization improvement.

Oracle had not disclosed financing arrangements for these chip purchases, according to the report. It also cites Oracle’s previously maintained fiscal 2027 capital-expenditure target of approximately $90 billion–$95 billion and management’s statement that customers were increasingly bearing chip-purchase costs; neither establishes funding for a specific announced deployment.