Microsoft (MSFT) Becomes the First Cloud Provider to Begin Validating Nvidia’s Most Powerful AI Chip

Posted on by

TLDR

  • Microsoft Azure has become the first cloud provider to start validating Nvidia’s Vera Rubin NVL72 system
  • CEO Satya Nadella shared this milestone via a post on X (previously Twitter) last Friday
  • Every NVL72 rack offers up to 3.6 exaflops of performance — five times higher than earlier GB200-based systems
  • The rack integrates 72 GPUs and 36 CPUs linked by sixth-generation NVLink with 260TB/s bandwidth
  • Amazon, Google, CoreWeave, Nebius, and Oracle are also projected to roll out Rubin systems in 2026

Microsoft Azure has outpaced its cloud competitors to become the first provider to initiate validation of Nvidia’s Vera Rubin NVL72 system. CEO Satya Nadella revealed this news in an X post on Friday afternoon, describing it as “another major step forward in developing the next generation of AI infrastructure.”

The Vera Rubin NVL72 is a rack-level platform that houses 72 Rubin GPUs and 36 custom Arm-based Vera CPUs within one system. The GPUs are connected using sixth-gen NVLink fabric, providing a bandwidth of 260 terabytes per second.

Microsoft Corporation (MSFT)
MSFT Stock Card

This represents a significant leap in raw computing power. Each NVL72 rack can achieve up to 3.6 exaflops of performance — about five times greater than the prior GB200-based systems it will replace.

According to Rani Borkar, Microsoft’s President of Azure Hardware Systems, this validation was years in the making. “Microsoft has years of market-proven experience in designing and deploying scalable AI infrastructure that evolves with every major advancement of AI technology,” Borkar said.

The critical term here is “co-design.” Microsoft states it has collaborated with Nvidia for years on interconnects, memory systems, thermal management, packaging, and rack-scale architecture. As a result, the Rubin hardware fits seamlessly into Azure’s current infrastructure without requiring any modifications.

Years of Planning Behind the Move

Microsoft’s Azure data centers — including those in Wisconsin and Atlanta — were purpose-built to accommodate the power density and liquid-cooling requirements of NVL72 racks. This type of infrastructure planning takes time and can’t be done quickly.

Borkar confirmed that Azure’s “superfactories” were already built to absorb these systems. “Rubin integrates directly into Azure’s platform without rework,” she said, pointing to years of preparation behind what looks like a clean first-mover announcement.

The company needed to revamp power and liquid-cooling systems at several locations to manage the higher watt density of these new racks. This investment is now yielding results, as Microsoft has validated hardware that others are still preparing to launch.

A consortium led by BlackRock, with support from Microsoft and Nvidia, recently initiated the acquisition of Aligned Data Centers in a $40 billion transaction aimed at boosting global capacity ahead of this upcoming hardware wave.

Rivals Still Waiting on Rubin

Microsoft’s first-mover status doesn’t mean it will remain the only one for long. Amazon Web Services, Google, CoreWeave, Nebius, and Oracle are all set to deploy Vera Rubin systems — most of them in the latter half of 2026.

Bernstein analysts have noted that Microsoft’s “first-to-validate” standing mirrors its wider efficiency improvements in cloud and SaaS services, which they quantify using a “Rule of 37.3%” outperformance measure.

On the day the news was released, MSFT shares dropped 1.57% and NVDA fell 1.58% — this movement aligned with overall market weakness rather than a negative response to the announcement.

Rubin Ultra, the next version of the platform, is scheduled to launch in 2027.