Oracle says Vera Rubin NVL72 racks and Vera CPUs are now generally available on its cloud, and the two companies plan to grow its superclusters to multi-gigawatt scale.
Nvidia and Oracle said on Tuesday that Vera Rubin NVL72 systems and Vera CPUs are now generally available on Oracle Cloud Infrastructure, and announced plans to expand Oracle’s AI superclusters built on Vera Rubin to multi-gigawatt scale. The companies say customers can scale CPU and GPU resources separately, pairing GPUs for training and serving models with CPUs for the work agents do between model calls, such as running tools, testing code and retrieving data. Oracle says Vera CPU instances will reach select regions this quarter. Oracle also plans to offer HGX Rubin NVL8, an eight-GPU server with two 64-core Intel Xeon 6 processors for customers who need x86 compatibility. The release says OpenAI trained and serves GPT-6 Astra on an Oracle supercluster in Abilene, Texas, with more than 100,000 Nvidia Blackwell GPUs. The multi-gigawatt expansion is a plan; no timeline or capacity figure was given.
Sources: Nvidia Newsroom · Oracle · Oracle Cloud Infrastructure blog · DatacenterDynamics