Surface RTX Spark Dev Box Opens at $5,999, Ships in November

Microsoft opened US preorders for the Surface RTX Spark Dev Box on October 7, 2026, at a listed $5,999.99. Microsoft says shipments will begin in November, without specifying a particular day. Orders are exclusive to Microsoft's US online store; the announcement does not establish wider international availability.

The compact Windows developer machine was introduced in June, so the new development is not the existence of the product or its RTX Spark architecture. It is the confirmed retail configuration, price, preorder availability and November delivery window. The distinction matters to developers weighing an on-premises machine against hosted GPU capacity.

Microsoft's current US Store lists a single Surface RTX Spark Dev Box configuration: NVIDIA RTX Spark N1X with a 20-core CPU, a 6,144-core Blackwell RTX GPU, 128GB LPDDR5x unified memory, a removable 2TB PCIe Gen 5 SSD and Windows 11 Pro. Microsoft includes a 165W power supply; that rating should not be mistaken for measured system power consumption.

The enclosure measures approximately 326.5 × 172.0 × 42.9mm including its foot and weighs about 2.0kg. Microsoft's specification page lists Wi-Fi 7, two USB-C ports, HDMI 2.1b, USB-A 5Gbps and a 3.5mm audio jack. It does not establish an Ethernet link speed, so network-throughput assumptions should wait for a verified specification.

Microsoft positions the Dev Box as a local-first AI development workstation rather than a general-purpose gaming mini PC. Its developer-oriented Windows 11 Pro image is described as including Visual Studio Code, Git, GitHub Copilot, GitHub CLI, Python, Node.js and PowerShell 7. Microsoft also says WSL 2 is configured with GPU passthrough and CUDA support, making Linux-based AI development part of the intended workflow.

This software setup is a meaningful difference from buying an ordinary Windows desktop and configuring every runtime manually. It does not guarantee that every existing x86 Windows application, Linux GPU library, CUDA package or AI model will run unchanged on the Arm-based platform. Compatibility still depends on the specific binary, framework, driver and workload.

Microsoft and NVIDIA describe RTX Spark as delivering up to one petaflop of AI compute and supporting models exceeding 120 billion parameters locally. Microsoft's earlier developer announcement identifies the petaflop figure as theoretical FP4 throughput using sparsity. It is not a claim of one petaflop for ordinary FP32 work, nor a measured rate for a particular model.

Unified memory can help accommodate large model weights and context data in one addressable pool, but fitting a model does not prove interactive generation speed. Quantization, available working memory, context length, kernel support, memory bandwidth and software optimization all affect real-world results. Core Tech Tips has not independently benchmarked the shipping Dev Box.

Surface Laptop Ultra and Surface RTX Spark Dev Box use the same broad NVIDIA Grace/Blackwell RTX Spark platform, but target different operating conditions. The laptop starts at $2,599.99 and is due October 16, while the desktop Dev Box fixes the configuration at 128GB unified memory and 2TB storage, adds the developer-ready Windows environment and ships later. Those differences are relevant to sustained desk-based development, but no independent head-to-head sustained-performance result is available yet.

For developers considering the $5,999.99 desktop, the next meaningful evidence will be shipping-unit testing in November: sustained local-inference throughput, power and thermals, framework compatibility under Windows and WSL, and the cost of maintaining local hardware versus paying for cloud GPU time. Neither Microsoft specifications nor launch-stage demonstrations settle those questions.