The new DGX Station for Windows brings deskside supercomputing capabilities directly to enterprise developers and researchers. NVIDIA announced this hardware platform at the GTC Taipei event. Notably, the system runs frontier artificial intelligence models of up to 1 trillion parameters locally.
This hardware bridges the gap between heavy-duty Linux data centers and everyday Windows productivity environments. Consequently, developers can now build and deploy autonomous agents within their existing workflows. The system is scheduled for release in the fourth quarter of 2026.
Features of the DGX Station
The system is powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. Specifically, this chip connects a Blackwell Ultra GPU to a 72-core Grace CPU. This connection uses the high-speed NVLink-C2C interconnect to maximize communication speeds.
Furthermore, the hardware features up to 748GB of coherent memory and 20 petaflops of FP4 performance. Users can also add an NVIDIA RTX PRO 6000 Blackwell Workstation GPU. Meanwhile, the integrated ConnectX-8 SuperNIC supports networking speeds up to 800Gb/s.
Agent Infrastructure and Capabilities
Enterprise AI is shifting from simple chatbots to autonomous agents that operate continuously. Therefore, this system serves as dedicated infrastructure for running hundreds of these agents simultaneously. Developers can connect these agents directly to 3D design and engineering applications.
“As enterprises scale AI agents across their organizations, they need AI infrastructure that can connect directly to the applications and workflows that power their business.”
Chris Marriott, vice president of enterprise platforms at NVIDIA
Security and System Management
To secure these autonomous agents, NVIDIA introduced the OpenShell runtime environment. This open-source runtime uses new Windows security and containment primitives. As a result, it creates an isolated sandbox for each agent to protect sensitive cybersecurity data.
In addition, IT teams can manage the DGX Station using existing Windows compliance and fleet management tools. Linux workloads also receive support through the Windows Subsystem for Linux. This integration ensures that organizations maintain operational readiness across their hardware fleets.
Availability and Ecosystem Partners
Several global manufacturers will distribute the DGX Station starting in the fourth quarter of this year. Specifically, these partners include ASUS, Dell Technologies, GIGABYTE, HP, MSI, and Supermicro. This widespread availability will allow enterprises to scale their local AI workloads.
Ultimately, this collaboration between NVIDIA and Microsoft represents a shift in local enterprise computing. By bringing data-center-class power to the desktop, the companies aim to accelerate productivity. Organizations can now transition from basic digital tools to advanced autonomous systems.
