Available now, with future plans stated separately
NVIDIA and Oracle announced general availability of Vera Rubin NVL72 systems and Vera CPUs on Oracle Cloud Infrastructure. An NVL72 rack links 72 GPUs through NVLink for large-model training and inference. Vera CPUs handle work between model calls, including tool execution, code, testing and data retrieval. OCI lets customers scale CPU and GPU resources independently so that one class of processor does not remain idle while waiting for the other.
Multi-gigawatt scale is still future tense
The partners plan to expand OCI Superclusters to multi-gigawatt scale and add HGX Rubin NVL8, an eight-GPU Rubin server with two 64-core Intel Xeon 6 CPUs, 4TB of DDR5 memory and 6.4Tbps RDMA networking. Those specifications describe a proposed configuration, not a service already deployed. The release also cites OpenAI, which says it used an OCI cluster with more than 100,000 Blackwell GPUs to train and serve GPT-6 Astra; that is a participant’s account rather than an independent performance test.