Microsoft unveils Surface RTX Spark Dev Box for sustained AI workloads

Microsoft has announced the Surface RTX Spark Dev Box, a desktop device engineered specifically for sustained artificial intelligence workloads. Powered by NVIDIA’s Arm-based RTX Spark chip, the unit is designed to handle long-running training jobs, agentic AI pipelines, and local model fine-tuning. The announcement was made at the company’s Build conference, positioning the device as a dedicated tool for developers rather than general consumers.

The hardware features up to 128GB of unified memory and delivers one petaflop of AI computing power. Unlike mobile equivalents, the Dev Box utilises a 100W thermal envelope, allowing it to manage higher heat output and maintain performance during intensive tasks. It also includes an embedded NVIDIA RTX Blackwell GPU, which Microsoft notes provides gaming performance comparable to the laptop version of the RTX 5070.

Microsoft has not yet disclosed official pricing for the Surface RTX Spark Dev Box. However, the device competes in a niche market where similar systems, such as AMD’s Ryzen AI Halo PC and NVIDIA’s DGX Spark mini PC, are priced at approximately $3,999. The Dev Box is positioned as a desktop alternative to the Surface Laptop Ultra, which also utilises the RTX Spark chip but is constrained by mobile thermal limits.

Distribution for the new hardware will be exclusive to Microsoft’s online store, with availability scheduled for later this year. The company confirmed that the device will not be sold through major retail outlets such as Best Buy, reinforcing its focus on professional developers who require specific hardware configurations for local AI development.

Performance metrics for local model fine-tuning may vary depending on specific model architectures and quantisation methods. While the device offers significant computing power for its form factor, Microsoft has not provided detailed benchmarks beyond the general comparison to the RTX 5070 laptop variant.