Microsoft Azure Becomes First Cloud Provider to Validate Nvidia's Vera Rubin NVL72 AI Supercomputer

Azure Takes the Lead in Next-Gen AI Hardware

Microsoft has officially become the first cloud provider to validate Nvidia's Vera Rubin NVL72 system, a next-generation rack-scale AI supercomputer that promises to reshape the economics of artificial intelligence. CEO Satya Nadella confirmed the milestone on March 13, calling it "another big step in building the next generation of AI infrastructure with Nvidia."

A Five-Fold Leap in Raw Power

The Vera Rubin NVL72 packs 72 Rubin GPUs and 36 custom Arm-based Vera CPUs into a single rack, connected through sixth-generation NVLink fabric delivering 260 terabytes per second of internal bandwidth. The result is 3.6 exaflops of inference performance, a five-fold improvement over the previous Blackwell-based GB200 NVL72 systems, alongside a claimed ten-fold reduction in cost per token for inference workloads.

Years of Planning Behind the Milestone

Microsoft's first-mover advantage stems from years of co-design work with Nvidia. The company's Fairwater AI superfactory sites in Wisconsin and Atlanta were purpose-built from the ground up to accommodate Rubin's demanding power, cooling, and bandwidth requirements. Azure's hardware lead Rani Borkar confirmed these facilities were specifically engineered to slot the new systems in without architectural rework, thanks to a multi-year redesign of power delivery and liquid-cooling infrastructure.

The Competition is Close Behind

While Microsoft leads the validation race, rivals are preparing their own deployments. Amazon Web Services, Google Cloud, Oracle Cloud Infrastructure, and Nvidia cloud partners CoreWeave, Lambda, Nebius, and Nscale are all expected to offer Vera Rubin-based instances in the second half of 2026. Nvidia began shipping initial samples to key partners in late February, with full production ramping through the year.