Oracle and NVIDIA have expanded their AI infrastructure partnership, broadening the availability of NVIDIA Vera Rubin systems and Vera CPUs on Oracle Cloud Infrastructure.
Oracle plans to scale OCI Superclusters with NVIDIA Vera Rubin NVL72 systems to the multi-gigawatt level, while making Vera Rubin NVL72 and Vera CPUs generally available on its cloud platform. The companies said the setup is intended to support AI model training, inference and software agents that rely on both GPUs and CPUs.
The move builds on an existing collaboration around NVIDIA-based systems on OCI, including GB200 and GB300 NVL72 deployments. Oracle said its supercluster design combines NVIDIA rack-scale systems with Oracle Acceleron networking, NVIDIA Spectrum-X Ethernet components and BlueField-4 data processing units.
In the Vera Rubin NVL72 setup, each rack links 72 GPUs through NVIDIA NVLink. Oracle said its network architecture can scale across hundreds of thousands of GPUs. It also plans to introduce NVIDIA HGX Rubin NVL8, an eight-GPU system in an x86 server configuration with two 64-core Intel Xeon 6 processors, 4 TB of DDR5 memory and 6.4 Tbps of RDMA networking.
The planned HGX Rubin NVL8 option is aimed at customers that want next-generation Rubin GPUs in a smaller configuration while retaining compatibility with x86-based applications. Oracle said this would give organisations another route for training, fine-tuning and inference workloads.
OpenAI use case
OpenAI was cited as an early user of the infrastructure. In Abilene, it has trained and served GPT-6 Astra on an OCI Supercluster with more than 100,000 NVIDIA Blackwell GPUs, according to Oracle and NVIDIA.
"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."
Alongside the GPU rollout, Oracle and NVIDIA said they are addressing the processor demands created by AI agents. These systems do more than generate text or responses and often need to call tools, execute code and retrieve information between model interactions.
Oracle said those tasks can leave GPUs underused if CPUs become a bottleneck. Vera CPUs on OCI are designed to address that by supporting single-threaded work and memory-heavy operations linked to code execution and concurrent task handling.
The companies also pointed to OCI Kubernetes Engine as a way for customers to scale CPU and GPU resources separately. Oracle said its Acceleron networking hardware can also isolate workloads running untrusted code generated by AI systems.
OpenAI is also testing Vera CPUs on OCI for what Oracle described as agentic sandbox workloads. That reflects a broader effort among cloud providers and chipmakers to adapt infrastructure not only for large model training, but also for the operational work surrounding automated software agents.
"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."
Broader build-out
The announcement highlights how cloud providers are expanding AI infrastructure beyond GPU clusters alone. As businesses build applications around large models and autonomous agents, providers are trying to offer more tailored mixes of processors, networking and memory configurations.
For Oracle, the partnership also strengthens its position in a market where access to advanced NVIDIA systems remains a key selling point for cloud operators. Rival hyperscalers are racing to secure AI hardware, expand data centre power capacity and demonstrate large-scale customer deployments.
Mahesh Thiagarajan, President of Oracle Cloud Infrastructure, said the latest expansion covers both model development and operational workloads. "Oracle Cloud Infrastructure gives customers the performance, security and scale to advance frontier AI," he said. "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 scaling efficiently as their needs grow."
NVIDIA presented the move as part of a longer-term effort to support changing AI workloads across hardware generations and software updates. "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."