NVIDIA and Oracle have announced the availability of NVIDIA Vera CPUs and Vera Rubin NVL72 systems on Oracle Cloud Infrastructure (OCI), alongside plans to expand Vera Rubin Superclusters to multi-gigawatt scale and introduce NVIDIA HGX Rubin NVL8.
OpenAI is bringing NVIDIA Vera CPUs into production on OCI for agentic sandbox workloads. Initial testing, presented at Oracle AI World and shared by Tae Kim, showed 51% higher Python throughput and 2.2 times the memory bandwidth per thread compared with a leading x86 CPU. The results point to potential performance gains in code execution and data access for AI agents.
The expanded infrastructure portfolio combines CPUs, GPUs, networking and CUDA software to support AI training, inference and agent workloads. Independently scaling CPU and GPU resources could help organisations allocate computing capacity according to workload requirements and reduce idle capacity.
NVIDIA and Oracle are also positioning the infrastructure to support evolving AI models and workloads across hardware generations. A common software foundation and continued optimisation are intended to help organisations adapt their infrastructure as computing requirements change.