NVIDIA and Microsoft expand Azure Vera Rubin access and Nemotron availability in Foundry
NVIDIA and Microsoft widened Azure Vera Rubin NVL72 VM access and added Nemotron 3.5 models plus post-training and custom-weight options to Foundry.
Named counterparties: Microsoft Corporation
NVIDIA and Microsoft expand Azure Vera Rubin access and Nemotron availability in Foundry
At Microsoft Ignite on November 19, 2026, NVIDIA and Microsoft said they are expanding access to Azure virtual machines powered by NVIDIA Vera Rubin NVL72 systems. NVIDIA says these systems are running in Azure data centers. On Microsoft Foundry, Nemotron 3.5 Super, aimed at long-running agents, and Nemotron Lightning, aimed at fast responses and tool use, are available on Foundry Managed Compute. Nemotron Embed and Rerank, Parse and Content Safety models are listed alongside them.
The update joins next-generation rack-scale hardware, open models and managed customization inside one cloud. Foundry's serverless Programmed Training service gives developers control over post-training workflows while Foundry runs the infrastructure. Managed Compute now supports deploying custom model weights. For NVIDIA, this ties demand for Vera Rubin capacity to its Nemotron 3.5 model line, which it also used in its Nemotron for Science release on November 17.
EY used Programmed Training to adapt Nemotron open models for classifying corporate purchasing records. NVIDIA reports that accuracy rose 55% on EY's evaluation dataset and cost fell 15x compared with proprietary frontier models. These are company-reported figures from one customer. Builders with narrow classification or extraction tasks can test tuned open models on Foundry against a proprietary baseline using their own evaluation data. Teams that plan around Vera Rubin should treat the Azure offering as a preview to evaluate, not as production capacity.



