Is Microsoft's Local AI Desktop Worth It?

Microsoft's Surface RTX Spark Dev Box costs $6,000 for local LLM inference on Windows, but is it worth it versus NVIDIA's DGX OS alternative? The Surface RTX Spark Dev Box costs $5,999.99 for 128GB of unified memory and a 2TB SSD, and that is not a Microsoft tax. Dell's GB10 desktop with the same memory and storage runs DGX OS for $5,903, about $100 less, and NVIDIA's own 4TB DGX Spark Founders Edition is $6,950 (US prices as of October 16, 2026). The price is a wash, so the real choice is the operating system. If your models run in Windows-native tools such as llama.cpp, Ollama, LM Studio or Windows ML, the Surface is a fair buy: CUDA Toolkit 13.4 added Windows on Arm support for RTX Spark devices, Ollama added Windows on Arm CUDA support in July, and LM Studio followed in October. If you depend on Linux serving engines like vLLM, SGLang or TensorRT-LLM in containers, a DGX OS box is the documented route. On the Surface they would have to run through WSL2, and as of October 16 nobody had published that setup serving models. Nobody has published an independent Dev Box benchmark yet. The yardstick is the DGX Spark: in the llama.cpp maintainers' February benchmarks it ran gpt-oss-120B at 58.7 tokens per second from a fresh prompt and 42.8 with 32K tokens of context. Nothing was run for this video; every figure comes from the makers' and projects' own pages. Microsoft's US preorder terms make testing low-risk: your card is charged when the box ships in November (a temporary authorization hold is possible), you can cancel before release, and holiday orders placed October 1 to December 2, 2026 can be returned through January 31, 2027, so you can test your own workload before deciding. Chapters: 0:00 Intro 1:52 The Price 4:52 Your Tools 9:11 Speed 11:46 Buying It 13:41 Verdict Tools & resources mentioned: - llama.cpp - Ollama - LM Studio - VS Code - GitHub Copilot CLI - WSL2 - CUDA Toolkit - Windows ML - TensorRT-RTX - PyTorch - vLLM - SGLang - TensorRT-LLM - NVIDIA Container Toolkit - Docker About The Stack The Stack helps you build with AI. Each video takes one tool, model, or workflow and shows how it works in a few focused minutes, with the real benchmarks and real costs. We go deep on Claude Code and Cursor for AI coding, AI agents and MCP servers, the open-source AI tools and GitHub repos most people miss, RAG and vector search, fine-tuning, and running local LLMs on your own machine with Ollama and LM Studio. We compare models like ChatGPT and Claude, test AI automation with Zapier, Make, and n8n, and flag the tools that actually ship. Subscribe for new breakdowns: https://www.youtube.com/@the-stack-ai?sub_confirmation=1 #localai #RTXSpark #DGXSpark #localLLM #WindowsOnArm