Skip to content
Source: NVIDIA Blog

☁️NVIDIA and Microsoft Put Vera Rubin NVL72 Systems Live in Azure Data Centers

Your Azure AI workloads just got a hardware upgrade option

TL;DR

NVIDIA and Microsoft are running Vera Rubin NVL72 systems in Azure data centers and adding Nemotron 3.5 models to Foundry. If you build agents or run inference on Azure, you now have a faster path to production-grade reasoning hardware and open models without leaving the platform.

NVIDIA and Microsoft just widened their full-stack partnership, and the headline hardware is real: Vera Rubin NVL72 systems are now running in Azure data centers, with preview access expanding. The pitch is infrastructure built for reasoning, agentic apps, and large-scale training and inference, all in one stack. If you're building enterprise agents, this matters because the software side got just as much attention. Foundry now offers Nemotron 3.5 in four flavors: Super for long-running agents, Lightning for fast tool use, Bolt for high-throughput reasoning, and Transcribe for speech. Programmed Training lets you control post-training workflows, and Foundry Managed Compute now accepts custom model weights. EY already used it to adapt Nemotron for classifying corporate purchasing records. The numbers are where the claims get concrete. Azure scaled to 10,240 NVIDIA GPUs on GB200 NVL72 for DeepSeek-V3 in MLPerf Training v6.1. Its GB300 NVL72 systems led cloud submissions on DeepSeek-R1 interactive throughput by over 10% in MLPerf Inference v6.1. Confidential computing is also being extended to bare-metal and rack-scale AI infrastructure, and Azure Local is adding RTX PRO 6000 Blackwell GPUs and Jetson Thor and Orin for edge work. Watch the Windows 365 Direct Stream preview: sign-ups are open now, but public preview isn't planned until early 2027.

NVIDIA and Microsoft Put Vera Rubin NVL72 Systems Live in Azure Data Centers — NVIDIA Blog

Key Points

1

Azure scaled to 10,240 NVIDIA GPUs using GB200 NVL72 systems for DeepSeek-V3 in MLPerf Training v6.1, a serious data point for teams planning large training runs on cloud infrastructure.

Keep reading — it's free

Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.

Also get
or

Already a member? Sign in

NVIDIAMicrosoft AzureVera RubinNemotronFoundry

Comments

Subscribe to join the conversation...

Be the first to comment