NVIDIA and Oracle expand AI infrastructure partnership
NVIDIA and Oracle announced an expansion of their AI infrastructure partnership. Oracle plans to scale OCI Superclusters using NVIDIA Vera Rubin NVL72 systems to multiple gigawatts and is also making Vera CPUs available for cloud customers.
The announcement matters because the partnership now covers more than GPU capacity. It includes CPUs, networking and the infrastructure needed for AI agents to run tools, test code and retrieve information between model calls. Oracle also plans to introduce the smaller HGX Rubin NVL8 configuration for customers that do not need a full rack-scale system.
CPU performance can affect GPU utilization. If tool execution or retrieval is slow, expensive GPUs may sit idle between model calls. Separately scaling CPU and GPU resources could improve utilization across the system.
Investors should not treat the phrase “multi-gigawatt scale” as booked revenue. The companies did not disclose a contract value, final unit count or complete delivery schedule. The key issues are actual deployment progress, the timing of Rubin revenue recognition and whether agent workloads create sustained demand for coordinated CPU and GPU capacity.