Enter your email address below and subscribe to our newsletter

Microsoft and Nvidia unveil unified AI agent stack at Build 2026

Share your love

  • Microsoft 0.13% and Nvidia 2.95% announced a joint agentic AI stack spanning Windows, Azure, and edge systems at Build 2026 in San Francisco.microsoft
  • The Surface RTX Spark Dev Box delivers up to one petaflop of AI compute locally, capable of running models with up to 120 billion parameters.microsoft
  • New security features include Nvidia’s OpenShell secure runtime and Microsoft’s Agent 365, enforcing OS-level isolation and governance for autonomous agents.windowsforum

NVIDIA and Microsoft Unveil Unified Agentic AI Stack at Build 2026

NVIDIA and Microsoft used the opening of Microsoft Build 2026 on Monday to announce a unified computing stack for agentic AI that spans Windows PCs, deskside workstations, Azure cloud infrastructure, and edge environments. The partnership, presented onstage by CEOs Satya Nadella and Jensen Huang in San Francisco, aims to make Windows and Azure function as a single continuous execution environment for AI agents.microsoft

From Device to Data Center

The expanded collaboration integrates NVIDIA’s physical AI tools, Nemotron models, Cosmos simulation software, and secure runtime components with Microsoft Foundry and Azure’s Physical AI Toolchain. Microsoft is also bringing Foundry Local on Azure Local to NVIDIA RTX PRO 6000 Blackwell Server Edition systems, enabling enterprises to deploy AI and visual computing workloads in distributed and edge environments with cloud-style orchestration.microsoft

On the model front, Microsoft Foundry now hosts NVIDIA Nemotron reasoning models alongside offerings from OpenAI and Anthropic, giving developers a multi-model catalog governed through a single enterprise platform. NVIDIA’s Hermes agent framework, which enables self-improving AI agents optimized for always-on local use, is designed to run on NVIDIA RTX PCs and DGX Spark hardware.windowsforum

The hardware centerpiece for developers is the Surface RTX Spark Dev Box, powered by NVIDIA RTX Spark, delivering up to one petaflop of AI compute and 128 GB of unified memory — enough to run models with up to 120 billion parameters locally.microsoft

Security and Governance at the Core

With agents expected to take actions, invoke tools, and operate over long sessions, both companies emphasized security. NVIDIA’s OpenShell secure runtime for autonomous agents uses Microsoft’s new Execution Containers to enforce OS-level isolation, policy management, and PII obfuscation. Microsoft also introduced Agent 365 for local agents, extending its Entra, Defender, and Purview security tools into a unified control plane for governing agents regardless of where they run.windowsforum

The announcement positions the two companies at the center of a rapidly evolving market where the boundaries between local and cloud AI are dissolving — though enterprises will still need to determine which data can travel where and which agents can act without human approval.windowsforum

Leave a Reply

Your email address will not be published. Required fields are marked *

Stay informed and not overwhelmed, subscribe now!