NVIDIA, Oracle Scale AI Infrastructure

NVIDIA and Oracle have announced a major expansion of their artificial intelligence infrastructure partnership, bringing NVIDIA Vera Rubin systems and Vera CPUs to Oracle Cloud Infrastructure (OCI). The collaboration aims to support multi-gigawatt AI superclusters capable of powering frontier model training, large-scale inference and increasingly complex autonomous AI applications.

The announcement marks another milestone in the global race to build AI factories. OCI Superclusters will integrate NVIDIA Vera Rubin NVL72 systems, advanced networking and high-capacity cooling infrastructure, enabling enterprises to deploy thousands of interconnected GPUs while addressing escalating computational demands.

Each NVL72 rack connects 72 GPUs through NVIDIA NVLink technology, while Oracle Acceleron and NVIDIA Spectrum-X networking enable large-scale connectivity. Oracle also plans to introduce HGX Rubin NVL8 systems, offering eight Rubin GPUs in a smaller configuration for flexible enterprise deployment.

A significant development is the general availability of NVIDIA Vera CPUs on OCI. Unlike conventional AI workloads dominated by GPU processing, agentic AI requires substantial CPU resources for executing code, retrieving information, coordinating tools and managing workflows between model interactions.

By allowing CPU and GPU resources to scale independently, Oracle aims to improve infrastructure utilisation and reduce expensive GPU idle time. Hardware-based isolation through Oracle Acceleron also addresses security concerns surrounding untrusted AI-generated code and autonomous agent execution.

OpenAI is among the early adopters exploring this infrastructure. According to the announcement, its GPT-6 Astra workloads have operated on an OCI Supercluster containing more than 100,000 NVIDIA Blackwell GPUs, highlighting the extraordinary computing requirements of frontier AI development.

The larger industry implication is a fundamental shift from GPU-centric computing toward integrated AI infrastructure. As autonomous agents become commercially important, processing efficiency, cybersecurity, energy consumption and infrastructure economics will determine competitive advantage. NVIDIA and Oracle are positioning themselves to capture this transition, although multi-gigawatt expansion also raises questions about power availability, capital expenditure and long-term investment returns.