NVIDIA and Microsoft Extend Accelerated Computing from Azure Cloud to Windows PCs and Edge Devices

Image: NVIDIA · Source
NVIDIA and Microsoft expand their collaboration to offer accelerated AI computing from Azure data centers to on-premises and edge environments, enhancing enterprise customization, deployment, and AI-powered applications.
NVIDIA and Microsoft have deepened their longstanding partnership to help organizations build, customize, and deploy intelligent AI applications spanning cloud, on-premises, and edge environments. Announced at Microsoft Ignite, the expanded collaboration emphasizes delivering full-stack accelerated computing — from Azure’s next-generation infrastructure to Windows PCs and industrial edge systems.
Central to this expansion is the deployment of Azure virtual machines powered by NVIDIA’s Vera Rubin NVL72 systems, designed for demanding AI model training, inference, and reasoning tasks. Microsoft and NVIDIA are also enhancing confidential computing capabilities to protect sensitive models and data during processing by enabling attestation and security at bare-metal and rack scale AI infrastructure. This development addresses organizations’ needs to handle proprietary or regulated information securely.
The partnership also advances efficient AI computing through joint work on power optimization with NVIDIA MaxLPS technology. Azure demonstrated significant performance leadership in recent MLPerf training and inference benchmarks, scaling to over 10,000 GPUs using NVIDIA Blackwell architecture.
On the enterprise AI front, NVIDIA and Microsoft are broadening access to NVIDIA open models and customization tools via Microsoft Foundry, enabling businesses to tailor open-source AI models for specific tasks. For example, Ernst & Young utilized Microsoft Foundry's Programmed Training to customize NVIDIA Nemotron models to significantly improve accuracy in classifying corporate purchasing data while dramatically reducing costs compared to proprietary alternatives.
To bring AI closer to data sources, NVIDIA and Microsoft are expanding model deployment options outside the cloud. This includes adding NVIDIA RTX PRO 6000 Blackwell GPU support and Nemotron 3.5 models to Azure Local environments, and extending edge AI support to NVIDIA Jetson Thor and Jetson Orin platforms for robotics, machine vision, and autonomous operation use cases.
Furthermore, Windows PCs powered by NVIDIA RTX Spark will support running large language model AI locally, showcased at Ignite, demonstrating new possibilities for AI functionality directly on user devices. For Windows 365 Cloud PCs, Microsoft and NVIDIA plan to introduce GPU-accelerated Direct Stream capabilities enabling demanding graphics workloads with multiple 4K streams and low latency, currently in limited preview, with a public rollout anticipated in early 2027.
For scientific research communities, NVIDIA BioNeMo and ALCHEMI toolkits are integrating into Microsoft Discovery, facilitating agent-driven computational biology workflows and protein design through AI models developed with academic partners.
Microsoft is also internally leveraging NVIDIA technologies such as Dynamo inference software and RTX PRO 6000 GPUs to accelerate reinforcement learning and Bing search AI models.
This comprehensive expansion reinforces the joint commitment to delivering powerful, secure, and versatile AI computing across cloud, edge, and endpoint hardware, enabling enterprises to innovate with tailored AI applications that span from data centers to local devices.
Sources and original reporting
Read the original source ↗

Comments (0)
No comments yet. Start the discussion.
Write a comment
Comments are published after moderation. Your name and comment will be visible publicly. Account