markdown

NVIDIA Unveils DGX Station for Windows: Desktop‑Scale AI in the Office

NVIDIA has announced the DGX Station for Windows, a compact machine that brings data‑center‑level AI power straight to office desks. The device runs on Windows, enabling compatibility with the majority of corporate workstations that traditionally rely on Linux for heavy AI workloads.

Powerhouse Inside

• Core: NVIDIA’s GB300 Grace Blackwell Ultra Desktop Superchip
• 72‑core CPU + powerful GPU
• Up to 748 GB unified memory
• Peak performance: 20 petaflops (FP4 precision) – enough to train or fine‑tune models with up to a trillion parameters.

Connectivity & Extensibility

• ConnectX‑8 SuperNIC: 800 Gb/s data transfer
• Allows multiple DGX Stations to network without bottlenecks.
• Optional RTX PRO 6000 Blackwell GPU for demanding visual or simulation tasks.

Agentic Inference

The primary use case is agentic inference: AI agents run continuously in the background, automating tasks and providing real‑time reasoning.

• Supports hundreds of agents per station.
• Direct integration with existing Windows applications for engineering, design, and everyday office use.

Security

• NVIDIA OpenShell: open‑source runtime that isolates each AI agent in its own sandbox using Windows security features.
• Prevents agents from accessing system policies or leaking sensitive data.
• Seamless integration with standard Windows security, update, and management tools.
• Linux‑based workloads can run via Windows Subsystem for Linux (WSL), easing transition without a full rewrite.

Scalability

• Serves as a single developer’s workstation or a shared node for entire teams.
• Option to migrate workloads to larger data‑center or cloud resources as projects grow.

Production & Partnerships

ASUS has already demonstrated the GB300 chip in a desktop form factor, underscoring the industry’s push to bring high‑performance AI hardware from racks onto desks.
Production for the Windows version is slated to begin in Q4 2026, with partners such as ASUS, Dell Technologies, GIGABYTE, HP, MSI, and Supermicro handling manufacturing and distribution.