In a nutshell: NVIDIA and Microsoft combine specialized hardware (RTX Spark, DGX Station for Windows), secure runtimes (OpenShell), and open-source models (Nemotron, Cosmos) into an end-to-end stack for agentic AI deployment from local Windows devices to Azure Cloud.
NVIDIA and Microsoft extend their partnership with a comprehensive infrastructure for agent-based AI systems: from local execution on Windows devices through GPU-accelerated enterprise workstations to Azure cloud deployments. The new architecture addresses the specific requirements of long-running agents with dedicated hardware, runtime, and model components.
Agentic AI deployment requires more than just powerful models: it needs specialized hardware, secure runtime environments, high-performance data layers, and models optimized for extended operations. NVIDIA and Microsoft bring this complete stack together across three tiers — Windows devices, Azure cloud, and local infrastructure.
On the Windows side, NVIDIA is introducing two new product lines: RTX Spark are the first Windows PCs designed specifically for personal agents, with 1 petaflop AI performance, up to 128 GB unified memory, and all-day battery life. Multiple manufacturers (Microsoft Surface, ASUS, Dell, HP, Lenovo, MSI) are bringing such devices to market in the fall. For enterprise applications, DGX Station for Windows follows as a deskside AI supercomputer featuring the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip — up to 748 GB coherent memory and 20 petaflops FP4 performance enable execution of frontier models with up to 1 billion parameters. These systems will be available from Q4 onwards from ASUS, Dell, GIGABYTE, HP, MSI, and Supermicro. Both product lines leverage NVIDIA OpenShell, a secure runtime tailored for autonomous agents.
For Azure deployments, NVIDIA complements its open model portfolio on Microsoft Foundry: Nemotron 3 Ultra is a new open frontier reasoning model for long-term agent workflows in coding, research, and enterprise applications (available this month on Foundry Managed Compute). Anthropic Claude runs natively on Azure systems with GB300 Blackwell Ultra. The portfolio expands with NVIDIA Cosmos 3, a fully open multimodal model for Physical AI featuring vision reasoning, world simulation, and action generation, as well as NVIDIA Earth-2 AI weather models for enterprise forecasting via Microsoft Planetary Computer Pro.
For the data layer, NVIDIA is now accelerating Microsoft Fabric Data Warehouse through integrated accelerated computing — a step that is critical for agent systems that rely on fast data access. With this, the partnership addresses a common bottleneck in productive agentic AI systems in the enterprise.
Source: blogs.nvidia.com · Published June 2, 2026
Lumi AI News — AI-assisted curation in accordance with Article 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.2.9.