NVIDIA and Oracle Expand NVIDIA Vera Rubin to Multi-Gigawatt Scale and Bring NVIDIA Vera CPU to Oracle Cloud Infrastructure

NVIDIA and Oracle have announced plans to scale Oracle Cloud Infrastructure (OCI) Superclusters powered by NVIDIA Vera Rubin NVL72 systems to multi-gigawatt capacity, addressing increasing customer demand for AI training, inference and agentic applications. The companies have also expanded their AI infrastructure partnership by making NVIDIA Vera Rubin NVL72 systems and NVIDIA Vera CPUs generally available on OCI.

The expanded collaboration will provide customers with infrastructure to support both GPU-intensive workloads, such as training and running advanced AI models, and CPU-intensive tasks performed by AI agents between model interactions. These tasks include executing tools, testing code and retrieving data. Customers will also be able to scale CPU and GPU resources independently according to their workload requirements, helping minimize idle capacity and improve infrastructure utilization.

OCI Superclusters Help Organizations Bring Frontier AI to Market Faster at Multi-Gigawatt Scale

Oracle Cloud Infrastructure (OCI) Superclusters integrate NVIDIA NVL72 rack-scale systems, Oracle Acceleron networking with NVIDIA Spectrum-X Ethernet technology, NVIDIA Spectrum-6 switches and BlueField-4 DPUs. Combined with gigawatt-scale power and cooling capabilities, the infrastructure is designed to train and run the world's largest AI models. Each rack connects 72 GPUs through NVIDIA NVLink technology, while OCI's resilient, multiplanar network supports deployments across hundreds of thousands of GPUs.

The planned Vera Rubin expansion builds on OCI's existing NVIDIA-powered infrastructure, which includes NVIDIA GB200 and GB300 NVL72 systems. Supported by NVIDIA's unified CUDA software platform and ongoing software optimisations, customers can run evolving AI models and workloads across different hardware generations.

OCI also plans to introduce NVIDIA HGX Rubin NVL8, an x86-based server configuration featuring eight Rubin GPUs connected through NVLink.

The platform will feature two 64-core Intel Xeon 6 CPUs, 4TB of DDR5 memory and 6.4 Tbps of high-speed RDMA networking. This configuration offers a flexible solution for AI training, fine-tuning and inference, allowing customers to adopt next-generation Rubin GPU technology in a more compact eight-GPU system while maintaining compatibility with existing x86-based applications.

This flexibility allows customers to adjust their AI infrastructure to accommodate different models and workloads as requirements change. By using accelerated computing resources across both training and inference, organisations can improve infrastructure utilisation and maximise their technology investments.

OpenAI has demonstrated the capabilities of this infrastructure through its work in Abilene, where it has trained and served GPT-6 Astra on an OCI Supercluster powered by more than 100,000 NVIDIA Blackwell GPUs.

“As our frontier models advance, we need infrastructure that can keep pace,” said Sreekanth Pothanis, engineering leader, compute infra at OpenAI. “NVIDIA Vera Rubin on OCI gives us the computing performance and scale to accelerate model training and serve increasingly demanding AI workloads.”

NVIDIA Vera CPUs on OCI Accelerate the Work Behind AI Agents

AI agents perform more than just generating responses. They also use tools, execute code and retrieve data to complete tasks. Delays in these CPU-intensive operations can slow down overall performance and leave GPUs idle while waiting for the next model call.

OCI Compute powered by NVIDIA Vera CPUs combines high single-thread performance with high memory bandwidth to accelerate these operations. Strong single-thread performance helps execute individual tool calls and code more quickly, while high memory bandwidth ensures CPU cores can efficiently handle data-intensive and concurrent tasks. These capabilities help minimise GPU idle time and maintain consistent performance as workloads expand.

To support these workloads, OCI Kubernetes Engine allows customers to scale CPU and GPU resources independently, while Oracle Acceleron provides hardware-based isolation for workloads involving untrusted, AI-generated code.

Customers participating in the early access programme for NVIDIA Vera CPUs on OCI, including OpenAI, are exploring their capabilities across various AI workloads.

“NVIDIA Vera CPUs on OCI deliver the world-class loaded performance needed to execute code and run tools across our rapidly growing agentic sandbox workloads,” said Uday Ruddarraju, chief technology officer of compute at OpenAI. “We are bringing Vera into production on OCI to support these workloads at scale.”

“Oracle Cloud Infrastructure gives customers the performance, security and scale to advance frontier AI,” said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure at Oracle. “Our collaboration with NVIDIA extends those capabilities across training, inference and agentic AI. By combining NVIDIA Vera Rubin GPUs and Vera CPUs with Oracle Acceleron networking and AI-optimized data centers, we’re helping customers keep GPUs productive, secure and scale efficiently as their needs grow.”

“AI factories must be productive, durable and fungible — delivering efficient computing, remaining useful through continued software advances and serving different models and workloads as demand changes,” said Ian Buck, vice president of hyperscale and high-performance computing at NVIDIA. “With Oracle, we’re bringing Vera Rubin and Vera CPUs to OCI to help customers put more of their infrastructure to work today and sustain its value as AI evolves.”