NVIDIA, Microsoft Unveil RTX Spark and DGX Station for Windows AI Agents

NVIDIA and Microsoft launched a comprehensive initiative to integrate advanced AI agents into the Windows ecosystem, marking a strategic convergence of hardware and software development. During the Windows AI and Surface event in San Francisco, NVIDIA CEO Jensen Huang and Microsoft CEO Satya Nadella underscored a shared vision for the personal computer as the definitive platform for the agentic era. Huang reflected on the historical synergy between the two firms, noting Windows' foundational role in NVIDIA's success, while Nadella praised Huang's consistent foresight in positioning AI as a secular trend. The executives emphasized that running capable agents locally on a PC will transform the device into a secure, always-on personal assistant. Microsoft announced the general availability of Microsoft Execution Containers (MXC), an OS-level infrastructure designed to host AI agents securely within the Windows environment. Pavan Davuluri, Microsoft's Executive Vice President of Windows and Devices, described MXC as a core primitive that enables agents to run persistently in the background while remaining fully governed and observable. Integrated with Agent 365 and Microsoft Security, MXC establishes the desktop as the most reliable execution environment for autonomous agents, ensuring enterprises and users can deploy AI workloads without compromising system integrity. To support these workloads, NVIDIA unveiled RTX Spark, a unified superchip delivering the full NVIDIA AI stack to Windows laptops and compact desktops. RTX Spark combines a Blackwell RTX GPU with up to 6,144 cores and a Grace CPU featuring up to 20 cores, linked via a high-speed interface. The architecture supports up to 128GB of unified memory and delivers one petaflop of FP4 AI performance, enabling the execution of large models such as the 125-billion parameter Qwen 3.8 Flash Next entirely locally. RTX Spark runs the full CUDA platform, allowing developers to leverage industry-standard tooling for local inference and model fine-tuning. Surface Laptop Ultra has been engineered around RTX Spark technology, offering configurations with up to 128GB of memory and petaflop-scale compute. Preorders for RTX Spark-powered systems are open immediately, with laptops shipping on October 16 and compact desktops following in November. The ecosystem spans major vendors including Acer, ASUS, Dell, HP, Lenovo, Microsoft, MSI, and Gigabyte, covering form factors from slim notebooks to compact desktops optimized for 24/7 operation. For enterprise workloads, NVIDIA previewed the DGX Station for Windows, the first deskside AI supercomputer to bring GB300 Grace Blackwell-class infrastructure directly into the Windows ecosystem. Previously restricted to Linux, this new iteration utilizes the GB300 Grace Blackwell Ultra Desktop Superchip to provide 748GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute. This capability allows organizations to run frontier models up to a trillion parameters locally, bridging the gap between Windows-based productivity workflows and high-performance AI development. The system retains compatibility with Linux toolchains via Windows Subsystem for Linux, enabling teams to fine-tune and deploy multi-agent systems without leaving their primary operating environment.