Microsoft's data centers have taken delivery of the first production Vera Rubin systems from Nvidia, CEO Satya Nadella confirmed on August 21, 2026. The new NVL72 architecture pairs Vera CPUs with Rubin GPUs carrying up to 288 GB of HBM4 memory, and Nvidia says it delivers up to 5x faster inference than the previous Blackwell platform.
Microsoft's data centers have received the first production Vera Rubin systems from Nvidia, and CEO Satya Nadella confirmed the milestone on August 21, 2026. The delivery puts Microsoft among the earliest companies to deploy what Nvidia calls a generational leap in AI computing.
Inside the Vera Rubin NVL72
Each Vera Rubin NVL72 system pairs 88-core Vera CPUs with Rubin GPUs carrying up to 288 GB of HBM4 memory per chip, a sharp jump that gives models more room during training runs that previously needed aggressive optimization just to fit into GPU memory. The systems also add NVLink networking and liquid cooling, tackling data movement between chips and the heat generated by running thousands of GPUs at once.
Nvidia says Vera Rubin delivers up to 5x faster inference and 3.5x improved training compared with Blackwell. If those figures hold in production workloads, the cost-per-token economics of running large language models could shift, making some AI applications viable at a scale that was previously too costly. CoreWeave, another early adopter, has reported a 10x increase in AI throughput per megawatt using the new systems.
Microsoft's build-out and the wider field
Microsoft is targeting the new hardware at next-generation superfactory sites planned for Wisconsin and Atlanta, facilities built around Vera Rubin's power and cooling needs. The company joins a short list of early adopters that also includes Google Cloud, CoreWeave, and OpenAI.
Nvidia moved Vera Rubin into full production by June 2026, so the gap between production start and first customer deployment ran roughly two months. Amazon Web Services, however, was conspicuously absent from early Vera Rubin announcements and may be leaning more heavily on its own Trainium chips instead.
The company has also been describing a vision of trillion-GPU architectures for what it calls "agentic AI factories" — autonomous systems that can reason, plan, and execute complex tasks without constant human oversight. Vera Rubin is positioned as a building block toward that vision.
Source: Crypto Briefing
Trading involves risk.
Risk Disclaimer
