It adds Vera Rubin NVL72 systems and Vera CPUs so customers can scale GPU and CPU capacity separately.
NVIDIA and Oracle announced plans to expand Oracle Cloud Infrastructure (OCI) Superclusters with NVIDIA Vera Rubin NVL72 systems to multi-gigawatt scale to support demand for AI training, inference, and agentic applications. Oracle and NVIDIA also expanded their AI infrastructure collaboration with the general availability of NVIDIA Vera Rubin NVL72 and NVIDIA Vera CPUs on OCI.
The expanded collaboration provides customers with infrastructure for both GPU-intensive model training and inference, as well as CPU-intensive tasks that agents perform between model calls, such as executing tools, testing code, and retrieving data. Customers can scale CPU and GPU resources independently based on workload needs, which may help reduce idle capacity and improve infrastructure utilization.
OCI superclusters help organizations bring frontier AI to market faster at multi-gigawatt scale
OCI Superclusters combine NVIDIA NVL72 rack-scale systems, Oracle Acceleron networking with NVIDIA Spectrum-X Ethernet networking featuring NVIDIA Spectrum-6 switches and NVIDIA BlueField-4 DPUs, and large-scale power and cooling infrastructure to train and serve AI models. Each rack connects 72 GPUs through NVIDIA NVLink scale-up networking, while OCI’s network is designed to scale across hundreds of thousands of GPUs.
The planned Vera Rubin expansion builds on OCI’s existing NVIDIA-based infrastructure, including NVIDIA GB200 and GB300 NVL72 systems. NVIDIA’s CUDA software foundation and ongoing software optimization are intended to help customers support changing models and workloads across hardware generations.
OCI also plans to introduce NVIDIA HGX Rubin NVL8, which provides eight Rubin GPUs connected through NVLink in an x86-based server configuration.
The planned platform includes two 64-core Intel Xeon 6 CPUs, 4TB of DDR5 memory, and 6.4 Tbps of RDMA networking, offering a deployment option for AI training, fine-tuning, and inference. HGX Rubin NVL8 gives customers access to Rubin GPU technology in an eight-GPU configuration while maintaining compatibility with x86-based applications.
This setup allows customers to adjust their AI infrastructure as models and workloads change, using accelerated computing for both training and inference.
OpenAI has used this infrastructure for frontier AI work. In Abilene, OpenAI has trained and served GPT-6 Astra on an OCI Supercluster with more than 100,000 NVIDIA Blackwell GPUs.
NVIDIA Vera CPUs on OCI accelerate the work behind AI agents
AI agents do more than generate responses. They call tools, execute code, and retrieve information to complete tasks. Slower execution of these CPU-intensive operations can delay results and leave GPUs waiting between model calls.
OCI Compute with NVIDIA Vera CPUs combines single-thread performance with memory bandwidth to help agents complete CPU-intensive tasks more quickly. Higher single-thread performance can speed up individual tool calls and code execution, while memory bandwidth helps keep CPU cores supplied with data during concurrent tasks. These capabilities can help reduce GPU idle time and support consistent performance as workloads scale. To further support these workloads, OCI Kubernetes Engine lets customers scale CPU and GPU resources independently, while Oracle Acceleron provides hardware-based isolation for workloads running untrusted, AI-generated code.
Customers with early access to NVIDIA Vera CPUs on OCI, including OpenAI, are evaluating them across a range of AI workloads.
For more information, visit nvidia.com.
