Nvidia's GB200 NVL72 system is now available via Oracle Cloud Infrastructure (OCI).
Revealed in a Nvidia blog post, Oracle has reportedly deployed thousands of Nvidia Blackwell GPUs in its data centers, and is now offering access to them via its cloud platform offering.
The Nvidia GB200 NVL72 is a system made up of 72 Nvidia Blackwell GPUs and 36 Arm-based Nvidia Grace CPUs, connected via fifth-generation Nvidia NVLink. Each GB200 NVL72 offers more than one exaflops of training performance.
Ultimately, Oracle aims to build a cluster of more than 100,000 Blackwell GPUs which will form one of its "OCI Superclusters."
Previous reports suggest this could be as many as 131,072 Nvidia Blackwell GPUs, and it was planned to launch in the first half of 2025. That would give the system 2.4 zettaflops of peak performance, Oracle said, although it was referring to FP4 format, rather than FP64 or even FP8.
The company also has an OCI Supercluster with 65,536 Nvidia H200 GPUs that offers up to 260 exaflops of peak FP8 performance, which went live in November 2024. The OCI H100 Supercluster can scale to 16,384 GPUs.
According to the blog post, the Blackwell GPUs are available via Oracle's public, government, and sovereign clouds, as well as in customer-owned data centers through its OCI Dedicated Region and Alloy offerings.
Other cloud providers to have made the GB200 NVL72 system available include Google and CoreWeave. Lambda has also deployed at least two NVL72 racks. Microsoft offers the GB200 GPUs, though they are not deployed as an NVL72 machine.
Amazon Web Services is working on "Project Ceiba," an AI cluster that will offer 20,736 B200 GPUs and 10,368 Grace CPUs. Ceiba will be built using the new liquid-cooled GB200 NVL72 platform with fifth-generation NVLink.
