Oracle has proceeded with the deployment and optimization of its first wave of liquid-cooled NVIDIA GB200 NVL72 racks across its data centers, bringing thousands of NVIDIA Blackwell GPUs online for immediate customer use. These racks are now powering NVIDIA DGX Cloud and Oracle Cloud Infrastructure (OCI) services, enabling the development and scaling of next-generation reasoning models and AI agents. Oracle’s deployment features high-speed NVIDIA Quantum-2 InfiniBand and Spectrum-X Ethernet networking, combined with an integrated software and database stack from NVIDIA and OCI to deliver scalable, low-latency AI performance.
OCI plans to build one of the world’s largest Blackwell GPU clusters, with OCI Superclusters scaling beyond 100,000 GPUs to meet soaring demand for inference and accelerated computing workloads. These installations mark a new era for cloud-based AI factories, utilizing the GB200 NVL72 rack-scale system — combining 36 Grace CPUs and 72 Blackwell GPUs per rack — for exceptional energy efficiency and agentic AI workloads. Customers including major technology firms, enterprises, government agencies, and regional clouds are preparing to deploy immediately on OCI’s GB200 systems. The new racks are available now via NVIDIA DGX Cloud and Oracle Cloud Infrastructure, with flexible deployment options across public, government, sovereign, and dedicated environments.