At GTC Taipei, NVIDIA officially launched the DGX Station for Windows, the world's first desktop supercomputer bringing datacenter-grade AI compute power into the Windows ecosystem. Built on the GB300 Grace Blackwell Ultra desktop superchip, the DGX Station delivers 20 petaflops of FP4 performance and features 748GB of unified memory, enabling local execution of cutting-edge AI models with up to one trillion parameters while supporting workstation-level scalability via expansion slots equipped with RTX PRO 6000 Blackwell GPUs. Traditionally, large-model training and multi-agent development have relied on Linux-based datacenters, whereas most enterprise daily operations occur within Windows environments. The DGX Station bridges this gap, allowing corporate developers and data scientists to build and run always-on AI agents in their familiar Windows environment while directly integrating with existing enterprise applications and workflows. In terms of security, the DGX Station incorporates NVIDIA's open-source runtime, OpenShell, which leverages Windows security primitives to create isolated sandboxes for each agent, enforcing system-level policy controls to prevent credential leakage or unauthorized access. Regarding networking, it is equipped with a ConnectX-8 SuperNIC that supports 800Gb/s ultra-high-speed interconnectivity, enabling multiple DGX Station systems to cluster together for handling larger-scale workloads. The DGX Station is scheduled to launch globally in Q4 of this year, with initial partners including ASUS, Dell, Gigabyte, HP, MSI, and Supermicro. It also supports running Linux AI toolchains through WSL. Microsoft stated that this collaboration will "extend Windows' capabilities from ultrabooks to datacenter-class workstations."