NVIDIA and Microsoft enhance AI capabilities with Azure NVL72 systems, Foundry tools, and RTX Spark PCs, enabling enterprises to deploy advanced applications.
NVIDIA and Microsoft are deepening their collaboration to advance artificial intelligence (AI) capabilities across cloud, local, and edge environments. Announced at Microsoft Ignite, the partnership introduces expanded access to NVIDIA technologies on Azure, enhanced customization tools for enterprise AI, and new deployment options for Windows PCs and industrial systems.
Microsoft Azure now supports NVIDIA Vera Rubin NVL72 systems, which are optimized for large-scale model training, inference, and agentic applications. These systems enable demanding AI workloads, including reasoning and data processing, while incorporating NVIDIA Confidential Computing to protect sensitive data and models. Azure’s infrastructure achieved notable performance benchmarks, scaling to 10,240 NVIDIA GPUs in MLPerf Training v6.1 and leading cloud submissions in MLPerf Inference v6.1 with over 10% higher throughput on the DeepSeek-R1 model, according to Signal65.
The companies are expanding Microsoft Foundry’s capabilities to help enterprises customize and deploy AI models. Foundry’s Programmed Training offering allows developers to refine post-training workflows, while Foundry Managed Compute supports deploying custom model weights. For example, EY used these tools to adapt NVIDIA Nemotron open models for corporate purchasing classification, achieving a 55% accuracy improvement and reducing costs by 15x compared to proprietary models.
5 lineup now includes specialized models for diverse applications. Nemotron Super supports long-running tasks, Lightning focuses on fast responses, and Embed and Rerank assist in information retrieval. Additional models, such as Nemotron Bolt for low-latency reasoning and Transcribe for speech recognition, are set to launch soon.
Source link







