Oracle and NVIDIA make Vera Rubin NVL72 and Vera CPUs available on OCI

Oracle and NVIDIA announced that Vera Rubin NVL72 systems and Vera CPUs are generally available on Oracle Cloud Infrastructure. The launch gives OCI customers access to the rack-scale GPU systems and Vera processors, though Oracle has not named the exact regions where Vera CPU instances can be ordered immediately. One Oracle product post says select-region availability will occur this quarter.

According to Oracle, OCI Compute with NVIDIA Vera CPUs is offered as Bare Metal and Flex Virtual Machine instances. The company says an initial 150,000 Vera CPUs will be made available across its regions. It does not provide a region-by-region allocation or clarify whether that figure counts physical processors, provisionable instances or another unit of capacity.

Separately, the companies said they plan to expand OCI Superclusters to multi-gigawatt scale and add an HGX Rubin NVL8 configuration. Those are future offerings, not capacity that Oracle says is already deployed. The announcements give no deployment date or location for the Superclusters and no availability date for NVL8. For context, DataPhoenix has separately covered Vera Rubin capacity plans from NVIDIA cloud partners in Europe.

Oracle lists each Vera processor as having 88 Olympus cores, 176 threads, up to 1.2 terabytes per second of memory bandwidth and 3.4 terabytes per second of bisectional fabric bandwidth. It also cites an NVIDIA platform benchmark claiming up to 80% faster agentic CPU performance than leading x86 CPUs. That is a vendor benchmark, not an independent OCI instance or workload test.

The Vera Rubin NVL72 offering connects 72 GPUs in each rack through NVIDIA’s NVLink interconnect. For the planned Superclusters, the companies describe a design that combines NVL72 systems with Oracle Acceleron networking, NVIDIA Spectrum-X Ethernet, Spectrum-6 switches and BlueField-4 data processing units.

The planned HGX Rubin NVL8 system would put eight Rubin GPUs connected by NVLink in an x86 server with two 64-core Intel Xeon 6 processors, 4 terabytes of DDR5 memory and 6.4 terabits per second of RDMA networking, according to the joint NVIDIA announcement.

Oracle and NVIDIA say independently scaling CPU and GPU resources will reduce idle capacity, improve GPU utilization and lower operating costs. The opened evidence includes no independent OCI workload benchmark, utilization test, pricing or total-cost comparison establishing those outcomes.

OpenAI compute chief technology officer Uday Ruddarraju said the company is bringing Vera into production on OCI for agentic sandbox workloads. That establishes a named customer deployment but does not independently validate Oracle’s performance or utilization claims. Oracle also cautions that future product direction is not a delivery commitment and that development, timing, release and pricing may change.