NVIDIA and Microsoft Expand Nemotron 3.5 Support in Foundry

NVIDIA and Microsoft are expanding enterprise AI options through Microsoft Foundry, Azure and Windows PCs, including more Nemotron 3.5 models and a limited preview of graphics streaming for Windows…

Nov 19, 2026
•
4 min read
Technobezz
NVIDIA and Microsoft Expand Nemotron 3.5 Support in Foundry

Don't Miss the Good Stuff

Get tech news that matters delivered to your inbox.

NVIDIA and Microsoft are expanding Nemotron 3.5 support through Microsoft Foundry, giving businesses more options for customizing AI models and running agents on their own infrastructure. The companies are also widening access to Azure virtual machines using NVIDIA Vera Rubin NVL72 systems through a preview, alongside work on local AI for Windows PCs.

For Windows 365 Cloud PCs, a limited preview accepting sign-ups brings Direct Stream to machines with GPU acceleration. NVIDIA says the technology can deliver up to two simultaneous 4K streams at 60 frames per second for graphics applications, including over difficult network connections. A public preview is planned for early 2027.

Read more: Anthropic Brings Claude AI Models to Microsoft Foundry on NVIDIA Blackwell Hardware

Within Foundry Managed Compute, Nemotron 3.5 Super targets agents that handle extended coding, research and business tasks, while Lightning targets applications needing quick responses and tool use. Developers can also deploy their own customized model weights through Foundry Managed Compute. Nemotron 3.5 Bolt, intended for reasoning with high throughput and low latency, and the speech recognition model Transcribe are due later.

Microsoft’s serverless Programmed Training lets developers direct post-training work while Foundry handles the infrastructure. NVIDIA says EY used it to customize Nemotron models for sorting corporate purchasing records, achieving a 55% accuracy improvement on EY’s evaluation dataset and a 15x reduction in cost against proprietary frontier models. That reported improvement concerns a specialized purchasing task measured on EY’s dataset.

For deployments on company premises, broader Nemotron 3.5 support in Foundry Local brings reasoning, coding and agents that use tools to Azure Local environments. Azure Local will support NVIDIA RTX PRO 6000 Blackwell GPUs supplied by hardware partners and is expanding to Jetson Thor and Jetson Orin for industrial uses, including robotics and machine vision.

The companies will demonstrate AI applications running directly on Windows PCs powered by NVIDIA RTX Spark at Ignite. That follows Wednesday’s launch of NVIDIA RTX Spark compact desktops from ASUS, Dell and MSI, with up to 128GB of unified memory for workloads including local agents and model fine-tuning.

Microsoft is also using NVIDIA Dynamo during post-training to generate experiences for reinforcement learning, with the software coordinating inference across a fleet of GPUs. Separately, the companies are working to extend confidential computing protections to rack-scale and bare-metal AI infrastructure, aiming to protect sensitive models and data during processing.

Share