NVIDIA, a leader in AI computing technology, is collaborating with Oracle to accelerate innovation in AI agents and inference models by supporting NVIDIA Blackwell GPUs.

Oracle has deployed and optimized NVIDIA's state-of-the-art liquid-cooled GB200 NVL72 racks in its data centers.

This brings thousands of NVIDIA Blackwell GPUs to NVIDIA DGX Cloud and Oracle Cloud Infrastructure (OCI). The system delivers powerful performance for developing next-generation AI inference models and AI agents.

Oracle’s new GB200 NVL72 includes NVIDIA Quantum-2 InfiniBand and NVIDIA Spectrum-X Ethernet networking to deliver scalable, low-latency performance. It is designed as a full-stack system that supports NVIDIA and OCI software and database integrations.

OCI aims to build the world’s largest Blackwell cluster. The OCI Supercluster will scale to over 100,000 NVIDIA Blackwell GPUs to meet the growing demand for AI computing. AI innovation is accelerating further as major technology companies such as OpenAI recently announced new AI inference models.

OCI's latest move is considered the latest case of NVIDIA's Grace Blackwell system being put into full operation. The system, which is transforming cloud data centers into large-scale AI manufacturing facilities (AI factories), is based on the NVIDIA GB200 NVL72 platform. The platform combines 36 NVIDIA Grace CPUs and 72 NVIDIA Blackwell GPUs to provide an optimized environment for high-performance AI inference models and agentic AI.

OCI will offer Blackwell GPUs with flexible deployment options for public sector, government, and enterprise customers. The technology can also be leveraged in customer-owned data centers through OCI Dedicated Regions and OCI Alloy.

The new GB200 NVL72 rack is the first system available on NVIDIA DGX Cloud, which provides an optimized platform for developing and deploying AI workloads across major cloud environments, including OCI. NVIDIA plans to use it to pursue a variety of projects, including training AI inference models, developing autonomous vehicles, accelerating chip design and manufacturing, and developing AI-based tools.

The GB200 NVL72 rack is currently available on DGX Cloud and OCI, delivering powerful performance to customers looking for large-scale AI infrastructure.