Announced 7 Oct 2026 · Sources checked
What did Microsoft and NVIDIA announce today?
Davuluri’s Windows Experience Blog post, dated 7 October 2026, is the written Microsoft account of the Windows and Surface event. Microsoft Source’s same-day media page lists Surface Laptop Ultra, Surface RTX Spark Dev Box, Microsoft Execution Containers and hybrid intelligence among the assets. NVIDIA’s blog, signed Gerardo Delgado and also dated 7 October, places Jensen Huang and Satya Nadella at Dogpatch Studios in San Francisco with a fireside chat hosted by Sriram Krishnan.
The hardware headline is RTX Spark on Windows. NVIDIA describes a Blackwell RTX GPU with up to 6,144 cores and an up-to-20-core Grace CPU linked at 600 GB/s, with one petaflop of FP4 AI performance and up to 128 GB of unified memory. Davuluri, quoted in NVIDIA’s post, said: “We built Surface Laptop Ultra around NVIDIA RTX Spark. With up to 128 gigs of unified memory and up to a petaflop of AI compute, you can run models on this laptop that simply don’t fit on a traditional machine.” Those are vendor specifications, not measurements we took. For how vendors present chip numbers, see AI chip metrics.
What is hybrid intelligence on Windows?
Davuluri defines hybrid intelligence as work that runs locally when that is enough and in the cloud when it is not. The post names three Copilot additions for Copilot+ PCs: local context from files and recent activity with permission, local actions such as organising files or troubleshooting, and local models used together with cloud models. Those Copilot features are expected to begin rolling out on Copilot+ PCs in the coming months, not today.
On the model side, Microsoft says it is bringing MAI Code 1.1 Flash to the device at 3-bit precision, with a 256K local context window, and names an upcoming NVIDIA Nemotron model of more than 70 billion parameters at 2-bit (just over 20 GB) and DeepSeek V4 Flash (284 billion parameters) for RTX Spark. GitHub HydraFusion, which already routes cloud tasks, is being extended to local models. Hybrid intelligence powered by HydraFusion is coming to the GitHub Copilot app, Copilot CLI and Visual Studio Code in experimental preview later in October. Windows ML also gains llama.cpp support.
A local model on a laptop is still a hosted-vs-local decision. Our managed versus self-hosted AI guide is the right frame: you trade cloud tokens and tenancy for device memory, thermals, update burden and a smaller model than the frontier cloud endpoint.
What does MXC general availability change?
Microsoft Execution Containers were previewed at Build 2026. Today they become generally available on Windows 11. Davuluri says MXC lets organisations define which files and networks agents can access, with those policies enforced at runtime, and that personal-PC users get the same class of safeguard as part of the agent experience. Isolation options named in the post run from process and session isolation to WSLc, virtual machines and Windows 365 for Agents. That is OS containment for computer-use agents, not a claim that every agent is now safe.
Agents already listed as supporting MXC are Codex from OpenAI, GitHub Copilot, OpenClaw, Replit, LM Studio, OpenShell from NVIDIA and Unsloth AI. Named as coming later: Anthropic Claude Code, Box, Egnyte, Heidi Health, Hermes Agent by Nous Research, Manus, Perplexity, Raycast and Simular. Meta’s Muse for Windows is “coming soon” as a native app with MXC. Governance at scale, footnote 1 says, may require additional services such as Agent 365 and Intune.
Which RTX Spark PCs can be preordered?
The US Microsoft Store page for Surface Laptop Ultra, opened on 7 October, lists $2,599.99. The page describes an NVIDIA RTX Spark Superchip, up to a 20-core Grace CPU, up to 128 GB of unified memory, a 15-inch touchscreen, peak HDR of 2,000 nits on a 10 percent window, three USB-C ports plus USB-A, HDMI, an SD reader and a headphone jack, and a laptop under 4.5 lb and 18 mm. Footnote 3 on that page says the 120-billion-parameter local-model line is theoretical FP4 TOPS with sparsity. Davuluri’s post says Surface Laptop Ultra is available beginning 16 October and that Surface RTX Spark Dev Box ships to US customers in November.
Partner preorders named today are ASUS ProArt P16 and P14, Dell XPS 16 Creator Edition, HP OmniBook Ultra 16, Lenovo Yoga 9n 2-in-1, MSI Prestige N16 Flip AI+ and Surface Laptop Ultra. NVIDIA’s blog adds Acer and Gigabyte to the wider Spark list and says compact Spark desktops go on sale in November. RTX Spark mini desktops and other Spark-based dev boxes are “later this year” in Davuluri’s wording.
| Product | What the posts say | When |
|---|---|---|
| RTX Spark Windows laptops | Preorder opened; partner list includes Surface, ASUS, Dell, HP, Lenovo, MSI | Shipments begin 16 October |
| Surface Laptop Ultra | Most powerful Surface; US Store starting price $2,599.99 | Available beginning 16 October |
| Surface RTX Spark Dev Box | Desk-side local AI box for developers | US shipments in November |
| Spark compact desktops / mini PCs | Always-on agent machines; NVIDIA says compact desktops go on sale in November | Later this year / November |
| DGX Station for Windows | GB300 Grace Blackwell Ultra; up to 748 GB coherent memory and 20 petaFLOPS FP4 | Later this year |
What remains later, experimental, or vendor-claimed?
DGX Station for Windows is a preview, not a preorder. Davuluri and NVIDIA say it uses the GB300 Grace Blackwell Ultra Desktop Superchip, with up to 748 GB of coherent memory and 20 petaFLOPS of FP4, enough in their telling to run models up to about a trillion parameters locally. Named later-this-year OEM systems are Dell Pro Precision with GB300 and HP ZGX Fury AI Station. Linux toolchains remain available through WSL.
Windows Search actions—phrases such as “switch to dark mode” or “arrange my windows”—begin rolling out to Windows Insiders in the experimental channel today. Inline Copilot in Search is opt-in and starts in select markets later this year. Call of Duty on RTX Spark is announced for 2027.
The 2.1× time-to-first-token, 4.3× image-generation and 6.2× video-generation comparisons are against an Apple MacBook Pro 16-inch with M5 Pro and 64 GB, run in September 2026 on a selection of preproduction Spark PCs. Footnotes 6–8 name llama.cpp with Qwen3.5 27B, ComfyUI with FLUX.2 Klein 4B, and ComfyUI with LTX 2.3 22B. Those are Microsoft-commissioned or NVIDIA tests, not a bake-off we ran. Features require Windows Update and vary by device, market and silicon.
Common questions
Is Surface Laptop Ultra shipping today?
No. Microsoft opened preorders on 7 October and says the laptop is available beginning 16 October. The US Store page we opened listed $2,599.99. We did not place an order.
Does MXC general availability mean every Windows agent is sandboxed?
No. Microsoft lists specific agents that already support MXC and a longer list that will add support. The post also says Windows Update is required and that timing varies by device and silicon.
Did Microsoft publish independent proof that Spark beats an M5 Pro MacBook?
No. The multipliers come from Microsoft-commissioned and NVIDIA tests on preproduction 64 GB Spark PCs against one M5 Pro configuration, with the models and tools named in footnotes 6–8.
What to remember
Buy an RTX Spark PC for the ship date and the configuration on the Store page, not for a general-availability Copilot that already routes every task locally. MXC is the OS change that can be used now by the agents Microsoft named. HydraFusion, Copilot hybrid features and DGX Station for Windows are later.
Sources & further reading
How this story was made
Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. Cover photography is illustrative and does not depict the specific announcement or product.
Our editorial standards
