Microsoft has opened pre-orders for two new Surface devices built around Nvidia’s RTX Spark platform, bringing its latest push for running demanding AI workloads locally closer to release.
The Surface Laptop Ultra and Surface RTX Spark Dev Box start at $2,599 and $5,999 respectively. The laptop will be available from October 16, while the Dev Box is scheduled to begin shipping in the U.S. in November.
Neither device is being announced for the first time. Microsoft introduced the Surface Laptop Ultra on May 31 and followed with the Dev Box on June 2. The October 7 update primarily adds pricing, pre-orders and release timing, while reinforcing Microsoft’s broader strategy of moving more AI development and inference onto local Windows hardware.
Microsoft Is Pushing More AI Work Onto the PC
Both systems use Nvidia’s RTX Spark architecture, combining a Blackwell RTX GPU with a Grace CPU and up to 128GB of unified memory shared across the system.
Microsoft says the hardware can deliver up to one petaflop of theoretical FP4 AI performance and run models exceeding 120 billion parameters locally. That figure reflects a specific low-precision AI workload using sparsity, however, and should not be read as a general measure of application performance.
The Surface Laptop Ultra packages that platform into a 15-inch Windows laptop aimed at developers, creators and other users with demanding GPU workloads. It weighs under 5 pounds, is less than 18mm thick and includes three USB-C ports, HDMI, USB-A, an SD card reader and a removable SSD. It is touchscreen but does not support Surface Pen input.
Microsoft is also introducing Magnetic Connect, a USB-C charging cable that attaches magnetically and disconnects when pulled. Unlike a separate proprietary charging port, the USB-C connection remains available for data, displays and accessories.
The Dev Box uses the same underlying platform in a compact desktop designed more specifically for AI development. It ships with Windows 11 Pro and a development environment that includes tools such as Visual Studio Code, Git, Python, Node.js and Windows Subsystem for Linux. Microsoft says it is intended for workloads such as local model inference, evaluation and experimentation that might otherwise consume cloud GPU resources.
The company says both systems can run agent workloads on-device, while future planned Windows capabilities are expected to add identity and management controls that distinguish agent activity from human users.
The new Surface systems sit at the high end of Microsoft’s PC lineup and is an early test of whether developers and professional users see enough value in running larger AI workloads locally to justify workstation-level hardware and pricing.
Leave a Reply