NVIDIAAnalysisNVIDIA and Oracle expand Vera Rubin NVL72 superclusters to multi-gigawatt scale on OCI
What happened.NVIDIA and Oracle launched multi-gigawatt superclusters on Oracle Cloud Infrastructure, bringing Vera Rubin NVL72 systems and standalone Vera central processors to market.
Why it matters.Complex agentic workflows spend considerable execution time executing peripheral tasks including database queries and code linting between generation steps. Decoupling high-performance accelerators from auxiliary tool processors prevents costly graphics clusters from sitting idle during intermediate agent operations, substantially improving operational economics for enterprise software teams.
The technical specifications:
- Independent compute scaling: OCI allows customers to scale GPU capacity for model serving separately from Vera CPU instances used for intermediate agent tool execution.
- Rack networking fabric: Each NVL72 cluster integrates Spectrum-X Ethernet, high-bandwidth Spectrum-6 switches, and BlueField-4 data processing units inside liquid-cooled server racks.
- System expansion: Oracle committed to adding an eight-GPU Vera Rubin configuration running on Intel Xeon server processors later in the deployment cycle.
STORYLINE PROGRESSION & CONTEXT
This report represents a verified evolution of prior newsroom intelligence:
- ↳Nvidia Releases Tensorrt Llm 0 22 With Native Speculative Decoding For Bla