Microsoft Azure Announces World’s First NVIDIA GB300 NVL72 Supercomputer Cluster for OpenAI's AI Workloads

Microsoft Azure has unveiled the world’s first NVIDIA GB300 NVL72 supercomputing cluster, a major step in building next-generation AI infrastructure. The system is part of Azure’s new NDv6 GB300 VM series, designed specifically for OpenAI’s large-scale inference and training needs.

The cluster houses more than 4,600 NVIDIA Blackwell Ultra GPUs, all linked through NVIDIA’s high-speed Quantum-X800 InfiniBand network. This setup delivers unprecedented computing power to handle advanced reasoning models, agentic AI systems, and multimodal generative AI workloads.

According to Microsoft, the achievement stems from years of collaboration with NVIDIA to create infrastructure that supports massive AI scale. Nidhi Chappell, Microsoft’s VP of Azure AI Infrastructure, said the milestone shows both companies’ commitment to optimizing every layer of the modern AI data center.

What makes GB300 NVL72 unique

Each liquid-cooled GB300 NVL72 rack combines 72 Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs. It delivers 37 terabytes of fast memory and up to 1.44 exaflops of FP4 Tensor Core performance per VM, which is ideal for training trillion-parameter models.

The system also benefits from NVIDIA’s NVLink Switch fabric for ultra-fast GPU communication and Quantum-X800 InfiniBand for large-scale performance. Together, they transform the entire setup into a unified, high-speed AI engine.