NVIDIA and Microsoft Unify the AI Stack — The Agentic AI Era From Windows to Cloud

Introduction: The Agentic AI Moment Has Arrived

At Microsoft Build 2026, NVIDIA founder and CEO Jensen Huang and Microsoft chairman and CEO Satya Nadella announced a massive expansion of their partnership focused on the next phase of artificial intelligence — agentic AI. This is not merely a technology alliance; it represents an attempt to build a unified infrastructure that enables AI agents to operate seamlessly across Windows devices, cloud services, and local deployments.

Agentic AI refers to systems that don't just respond to questions but independently execute complex tasks — writing code, analyzing data, making decisions, and acting in the physical world. NVIDIA and Microsoft are building the full stack to realize this vision, from hardware to models, from security to the data layer. The implications for developers, enterprises, and the broader technology ecosystem are profound.

Reinventing Windows: RTX Spark and DGX Station

NVIDIA and Microsoft are fundamentally reimagining the Windows ecosystem for the age of agentic AI. Two new products — RTX Spark and DGX Station for Windows — enable developers to build, tune, and run agents natively on Windows.

RTX Spark represents a new beginning — the world's first Windows PCs purpose-built for personal agents. With 1 petaflop of AI performance, up to 128GB of unified memory, all-day battery life, and full AI and graphics performance unplugged, it embodies over 30 years of NVIDIA innovation including CUDA, RTX, DLSS, and TensorRT technologies. Systems arrive in fall 2026 from Microsoft Surface, ASUS, Dell, HP, Lenovo, and MSI.

DGX Station for Windows is the most powerful deskside AI supercomputer for building and running agents in Windows enterprise applications and workflows. Powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip with up to 748GB of coherent memory and 20 petaflops of FP4 performance, it runs frontier models of up to 1 trillion parameters for always-on enterprise agents. Systems from ASUS, Dell, GIGABYTE, HP, MSI, and Supermicro are expected in Q4 2026. Both products run NVIDIA OpenShell, a secure-by-design runtime built specifically for autonomous agents.

These products represent a fundamental shift in how we think about personal computing. The PC is no longer just a tool for human operators — it's becoming a platform where AI agents can autonomously build, test, and deploy solutions. For developers, this means the ability to prototype and run sophisticated agent workflows locally before scaling to the cloud.

Open Models on Microsoft Foundry: Agentic Workflows at Enterprise Scale

Agentic AI runs on a system of models. NVIDIA, Anthropic, and OpenAI models — plus Hermes special agents — are now available on Foundry Agent Service, enabling enterprises to bring agentic systems to life on Azure with built-in identity and governance. Anthropic's Claude models now run natively on NVIDIA GB300 Blackwell Ultra systems on Azure.

NVIDIA Nemotron 3 Ultra, a new open frontier reasoning model designed for long-running agents across coding, research, and enterprise workflows, is available this month on Foundry managed compute. Nemotron 3.5 ASR for speech recognition and Nemotron 3.5 Content Safety round out the offering. Developers can compose Nemotron alongside frontier and local models, optimizing cost and quality for each specific workflow — a critical capability for enterprises that need to balance performance with economics.

NVIDIA's open model portfolio on Foundry now spans agentic, physical, and scientific AI. NVIDIA Cosmos 3, the first fully open omnimodel for physical AI, brings vision reasoning, world simulation, and action generation together in a single architecture. NVIDIA Agent Toolkit and NemoClaw blueprints give developers an open-source platform for building production agents on Foundry. NVIDIA CUDA-X libraries including cuDF, cuOpt, AI-Q, and NeMo are now accessible to agents as domain-specific skills — essentially giving AI agents the ability to leverage GPU-accelerated computing for specialized tasks.

Accelerating the Data Layer: Fabric Data Warehouse

Data fuels agentic AI, and fast access to it is critical for agents that continuously query and reason over enterprise information. NVIDIA accelerated computing is now built into Microsoft Fabric Data Warehouse, with Microsoft's internal benchmarking delivering SQL execution up to 6x faster than the CPU-powered baseline and up to 7x faster than three other leading cloud data warehouse providers for high-concurrency workloads.

This represents years of deep engineering collaboration between NVIDIA and Microsoft, from research to production. The enterprise data layer can now keep pace with AI agents that need to continuously query, analyze, and act on data in real time. For enterprises building agentic systems, this eliminates one of the key bottlenecks: the ability to provide agents with fast, reliable access to the data they need to make informed decisions.

Physical AI and Autonomous Systems

Physical AI represents the next frontier for agents — systems that can perceive, reason, plan, and act in the physical world. Microsoft is integrating NVIDIA's open-source physical AI skills and tools with Azure and its Physical AI Toolchain. Developers get a unified platform, powered by Cosmos 3's mixture-of-transformers architecture, to simulate, train, and deploy autonomous systems including robots, autonomous vehicles, and industrial systems.

Cosmos 3 ranks first among open models on key benchmarks for vision reasoning, world generation, and action generation. This is significant because it means developers can build physical AI systems using a single, unified model architecture rather than stitching together multiple specialized models. For industries like manufacturing, logistics, and agriculture — sectors that are increasingly relevant in developing technology ecosystems — this opens new possibilities for automation and intelligence.

Security and Local AI: OpenShell and Azure Local

One of the biggest challenges in agentic AI is security. When AI agents independently execute tasks — writing code, accessing files, making network requests — traditional security models are insufficient. NVIDIA OpenShell, now integrated into GitHub Copilot, addresses this directly: each agent runs in its own sandboxed container, and every outbound call is evaluated against policy before it can reach files, networks, or credentials. Policies are written as code, versioned in the repository, and updatable on the fly. OpenShell is open source under Apache 2.0, model-agnostic, and spans on-premises, hybrid, and cloud environments.

For organizations with strict data sovereignty requirements — including government agencies, financial institutions, and healthcare providers — Microsoft is bringing Foundry Local on Azure Local to the NVIDIA RTX PRO 6000 Blackwell Server Edition platform. Paired with NVIDIA Nemotron open models, enterprises can run high-performance AI workloads where their data resides, whether in on-premises, hybrid, or sovereign environments. Foundry Local on Azure Local now supports multinode deployments and the vLLM runtime, scaling inference for manufacturing, energy, sovereign data centers, and other latency-sensitive scenarios.

The AI Factory Era: Fairwater Wisconsin and Vera Rubin

The most infrastructure-intensive aspect of the partnership involves building the physical backbone for agentic AI at scale. Microsoft's Fairwater Wisconsin AI factory is now live ahead of schedule, running hundreds of thousands of NVIDIA Grace Blackwell systems as a single AI factory. Notably, it's connected with a similar AI factory in Georgia to deliver a scalable and distributed AI system for the most demanding frontier models.

Most significantly, Microsoft has validated the NVIDIA Vera Rubin platform — now in full production — for deployment across Azure data centers. Vera Rubin slots in alongside Blackwell with no retrofits, delivering up to 10x inference throughput per megawatt and reducing cost per agentic token by an order of magnitude. Built-in NVIDIA Confidential Computing protects models and data as agents reason at scale. The NVIDIA Dynamo inference framework extends those gains into software, accelerating model cold starts on AKS and bringing Kubernetes-native distributed inference orchestration via NVIDIA Grove.

This infrastructure innovation means that agentic AI is no longer the exclusive domain of large technology companies. The existence of a unified stack — from hardware to models, from security to the data layer — gives organizations of any size the ability to begin integrating agentic AI into their workflows. The era of autonomous AI agents working alongside humans is not approaching — it has arrived, and the infrastructure to support it is being built today.

🔗 წყარო: NVIDIA