NVIDIA has revealed the DGX Station for Windows, bringing its GB300 Grace Blackwell hardware to the desktop environment. The machine delivers up to 20 petaFLOPS of compute power to run models with 1 trillion parameters directly on a local workstation. The system bridges enterprise productivity on Windows with heavy local artificial intelligence processing, eliminating the requirement to switch operating systems.
Previous iterations of the DGX Station operated exclusively on Linux. This forced developers across Fortune 500 companies to maintain 2 separate environments, dividing their time between productivity software and model training. The new workstation operates directly on the GB300 Grace Blackwell Ultra Desktop Superchip. It provides 748GB of coherent memory and 20 petaFLOPS of FP4 compute capability, while preserving access to Linux toolchains through the Windows Subsystem for Linux.
Capabilities that once required renting a cluster, now in a deskside supercomputer.
Microsoft and NVIDIA are co engineering system software to support autonomous software agents. Microsoft introduced the general availability of Microsoft Execution Containers, an operating system foundation that permits agents to run in the background under direct administrative control. Pavan Davuluri, executive vice president of Windows and Devices, stated that these tools allow agents to remain secured, observed, and governed alongside Agent 365.
Just as Windows and DirectX revolutionized how applications were built, MXC is going to revolutionize how agents are built and deployed.
NVIDIA founder Jensen Huang and Microsoft chief executive Satya Nadella detailed the collaborative effort during an industry event in San Francisco. Huang pointed to the historical connection between both firms, explaining that the personal computer remains the primary workstation for developers building autonomous tools. Nadella described the local execution of large models as a permanent industry shift that establishes the desktop as a secure environment for enterprise deployment.
