NVIDIA and Oracle Plan Multi-Gigawatt Vera Rubin Expansion

What happened

NVIDIA and Oracle announced on October 27 plans to expand Oracle Cloud Infrastructure Superclusters with NVIDIA Vera Rubin NVL72 systems to multi-gigawatt scale. The companies say the expansion is intended to support growing customer demand for AI training, inference and agentic applications.

Alongside that expansion plan, the companies announced general availability of Vera Rubin NVL72 systems and standalone NVIDIA Vera CPUs on OCI. The distinction is important: the hardware offerings are available, while the broader multi-gigawatt buildout remains a plan.

The announcement covers both GPU capacity for training and running advanced models and CPU capacity for agents working between model calls. VentureBeat describes those CPU tasks as executing code, retrieving information and using software tools. Unite.AI reports that Oracle announced an initial deployment of 150,000 Vera CPUs across its regions.

Why it matters

• Agent infrastructure

  The announcement combines GPU infrastructure for model workloads with standalone CPU capacity for agents executing code, retrieving information and using tools between model calls.

• Deployment choices

  The planned eight-GPU NVL8 configuration would offer a smaller deployment option than the available 72-GPU rack configuration.

What we know

NVIDIA says OCI Superclusters combine NVL72 rack-scale systems with Oracle Acceleron networking and NVIDIA Spectrum-X Ethernet, including Spectrum-6 switches and BlueField-4 DPUs, plus gigawatt-class power and cooling. Each rack connects 72 GPUs through NVLink, while OCI’s multiplanar network scales across hundreds of thousands of GPUs, according to NVIDIA.

The planned expansion builds on OCI infrastructure that already includes NVIDIA GB200 and GB300 NVL72 systems. NVIDIA says its common CUDA software foundation and continued software optimization help customers support evolving models and workloads across hardware generations.

Oracle also plans to introduce HGX Rubin NVL8, a smaller configuration with eight Rubin GPUs connected through NVLink in an x86-based server. NVIDIA says the planned platform includes two 64-core Intel Xeon 6 CPUs, 4TB of DDR5 memory and 6.4 Tbps of RDMA networking for training, fine-tuning and inference. Unlike NVL72, NVL8 was described as planned rather than generally available.

For Vera CPUs, VentureBeat reports that Oracle cites NVIDIA’s claim of up to 80% faster performance on agent-related CPU workloads compared with leading x86 processors. That claim comes from NVIDIA platform benchmarks, not tests of the OCI instances themselves, so it should not be read as a measured performance result for the newly available cloud offering.