Rumors abound about what Microsoft will show at its big launch event this Wednesday. I’ll be there to tell you firsthand, but before I go, I want to give you a heads-up on what I think is coming and why it’s a very big deal.
First things first: Microsoft is billing its Oct. 7 event as a Windows- and Surface-focused affair, but it’s certain to be anything but a conventional Surface product launch. Instead, I believe this will mark Microsoft’s leap into local AI powered by Nvidia RTX Spark.
However, I believe the event will focus a great deal on the systems we’ve already had early looks at, such as the Surface Laptop Ultra and the Surface RTX Spark Dev Box mini PC, both built around Nvidia’s RTX Spark hardware.
These products have been teased throughout the year, with early looks at Computex and a trickle of rumors since then. So while it’s always possible Microsoft could throw a curveball, it’s highly likely that these products will be the stars of the show this week. Here are some of my best guesses about details Microsoft will unveil.
Core Silicon and Architecture: Nvidia RTX Spark
Let’s start with the RTX Spark itself. It’s an evolution of the Grace Blackwell architecture used in the enterprise-focused Nvidia DGX Spark, starting with the RTX Spark N1X chip. Built on TSMC’s 3nm node, it pairs a 20-core Grace CPU and a Blackwell GPU featuring 6,144 CUDA cores, all on a single system-on-a-chip (SoC), with up to 128GB of LPDDR5X unified coherent memory dynamically shared across the CPU, GPU, and NPU.
Apple has already popularized this sort of super-SoC design, and AMD has taken a similar approach with both the AMD Ryzen AI Halo (AMD’s answer to the DGX Spark) seen in laptops like the HP ZBook Ultra G1a 14, which is pitched as a high-end mobile workstation.
Nvidia RTX Spark takes that same super-SoC design and AI processing muscle—delivering up to 1 petaflop of AI compute power—and packages it into a portable laptop form factor.
Jensen Huang talks about RTX Spark at Computex 2026 (Credit: Nvidia)
The result is a workstation-like system that supports the full RTX technology stack and scales its dynamic power envelope from single-digit wattage for daily tasks up to 80W for full local AI capability and heavy workloads.
Adopting the RTX Spark suggests that Redmond is going all-in on agentic AI, pushing Windows as a premier platform for AI development, local AI processing, and autonomous agent tools.
The high-powered processing and elimination of traditional VRAM bottlenecks allow this hardware to run AI models with up to 120 billion parameters—all offline on the device, without involving a data center, an internet connection, or a loss of privacy.
Oh, and a side effect of this AI-ready power is that the Nvidia RTX Spark also supports 3D rendering, 12K video editing, gaming at 1440p/100 frames per second, as well as DLSS Frame Generation, DLSS Ray Reconstruction, and OptiX acceleration. It’s not being sold strictly as gaming hardware, since it’s oriented toward professionals and AI enthusiasts, but that level of graphics power (on par with an RTX 5070 laptop GPU) also makes it an exceptionally capable gaming and content creation machine.
Flagship Hardware: Surface Laptop Ultra, Dev Box, and OEM Partners
We already know that RTX Spark laptops will include models from Acer, Asus, Dell, Gigabyte, HP, Lenovo, and MSI—all of which were announced at Computex, with Nvidia saying that more than 30 premium laptops and 10 desktops will be announced this fall.
(Credit: John Burek)
But to me, the most intriguing manufacturer to adopt RTX Spark is Microsoft itself. Built around this same Nvidia silicon, Microsoft has already shown us that the Surface Laptop Ultra will weigh less than 4.5 pounds and feature a 15-inch mini-LED PixelSense Ultra touch screen (2,000 nits peak HDR brightness, 120Hz VRR, 3:2 aspect ratio), a 30% larger haptic touchpad, and a full selection of ports (HDMI, USB-C, USB-A, SD card slot, headphone jack). Inside, it houses a dual-fan thermal system delivering over twice the cooling capacity of the Surface Laptop 7.
Microsoft has also previously announced that the RTX Spark will power a mini PC aimed at developers. The Surface RTX Spark Dev Box isn’t aimed at general consumers; Microsoft pitched it as a desktop tool for software developers built for prototyping, fine-tuning, and running their own AI models locally. It’s optimized with a 100W sustained thermal envelope and a cooling-first design that uses an all-aluminum body as a heatsink, with a grid-like ventilation pattern across the chassis surfaces for maximum airflow.
Both devices will run Windows 11, with the developer box shipping preinstalled with Windows 11 Pro, along with pre-configured defaults, preinstalled tools, and settings tuned specifically for an AI development environment from first sign-in.
That signals one major change: a lot more support for AI in Windows. If Microsoft is arming developers with tools specifically built for AI development and agentic AI on Windows, it’s because Microsoft is investing in what it sees as the future of Windows for everyone.
Windows on Arm and OS Ecosystem Enhancements
One of the big changes coming with Nvidia RTX Spark is a suite of improvements for Arm-based CPUs running Windows OS. Microsoft has been expanding the footprint of Windows on Arm for several years now, most specifically in the Surface line with Qualcomm Snapdragon processors. Bringing Nvidia into the mix bolsters those efforts and boosts the entire Windows on Arm ecosystem.
For starters, the Nvidia hardware integrates with the existing Microsoft AI PC platform with an integrated NPU—a specific hardware addition made for RTX Spark as it leaps onto Windows. (Just don’t call it Copilot+. More on that in a bit.)
Windows is getting kernel-level OS execution support for agentic AI tools, enabling autonomous local software agents (such as OpenClaw and Hermes Agent) to run in the background without disrupting the Windows experience. Nvidia has also announced its own Open Agent Safety Platform, a security and containment framework to make agents safer to use, which will also be coming to RTX Spark laptops.
Microsoft’s AI Problem, Explained by Someone Who Loves Windows
Smarter memory management will also be needed to dynamically allocate the massive unified RAM pool (up to 128GB) between graphics processing and the memory required for LLMs and agentic AI. It’s a balancing act that impacts both the user experience and the efficiency of local AI, so Nvidia’s experience in this area will be useful for Microsoft as the company refines Windows 11 to handle a new class of hybrid desktop workloads.
Prism, Microsoft’s x86/x64-to-Arm translation layer, is receiving further RTX Spark optimizations. Prism already added support for emulating x86 AVX and AVX2 instruction-set extensions, improving compatibility with games and applications that rely on them.
Windows on Arm is also gaining native Arm64 support from anti-cheat providers, including Epic’s Easy Anti-Cheat and BattlEye, addressing a major compatibility hurdle for multiplayer games. Riot Games has confirmed League of Legends and Valorant for the platform, while Krafton is bringing PUBG: Battlegrounds. Together, these changes expand Windows-on-Arm gaming support on RTX Spark systems, though many games will still run through Prism rather than as native Arm builds, and compatibility remains title- and anti-cheat-specific.
The Strategic Shift: A New Era of Local and Agentic Computing
This also heralds what Microsoft and Nvidia are calling “a new era of PC.” With Surface and Copilot+, Microsoft laid claim to the AI PC. In the years since then, AI PCs have expanded to include far more than consumer systems featuring NPU hardware. Gaming and workstation systems with high-powered graphics and VRAM for running small local models have expanded the AI landscape, and systems with large pools of unified memory have become the vanguard of agentic AI.
The biggest example of this is Apple’s Mac lineup, which began using unified memory well before the AI boom and has since seen widespread adoption among developers and AI enthusiasts precisely because it supports local AI functions that most Windows machines lack.
Satya Nadella talking about Windows as a platform for agents at Build 2026 (Credit: Chris Hoffman)
And finally, we have developer boxes like the Nvidia DGX Spark and the AMD Ryzen AI Halo, which are purpose-built for running large AI models locally, both for agentic capabilities and for building and fine-tuning AI tools.
RTX Spark is an evolution of this growth, taking the high-powered capabilities of the DGX Spark, packaging it into a more portable design on a more accessible platform (Windows), and adding the sorts of commodity features that shoppers are looking for.
This first wave of RTX Spark systems will not be aimed at general consumers. But I don’t think that will be the case for very long. Instead, this is the first wave of tools being given to developers specifically to build out similar AI capabilities for individual users.
Recommended by Our Editors
What We Still Don’t Know: Pricing, Demos, and Possible Surprises
Even with all of the products we’ve seen, the specs that have been announced, and the rumors we feel confident about, there are still a lot of unknowns around RTX Spark, the Surface Laptop Ultra, and more.
Price: DRAM Volatility and Memory Tiers Will Affect Costs
The biggest unknown is pricing. Computex announcements of specific products were intentionally vague about cost, due in part to early reveals coming so far ahead of retail availability and in part to volatile memory markets.
Jensen Huang announcing RTX Spark Laptops at Computex 2026 (Credit: Nvidia)
We’ve seen several companies raise prices on existing products due to increasing DRAM and component costs. Price tiers for this new class of RTX Spark laptops will rely heavily on memory configurations—whether we’re talking about entry-level models with 16GB or 32GB of unified memory or maxed-out 128GB flagships built for running massive local LLMs. (Even Nvidia has succumbed to memory pricing pressures with a new DGX Spark 64GB model.)
The Surface Laptop Ultra is being positioned as a competitor to the Apple MacBook Pro and similar creator and workstation-adjacent systems, which usually start around $2,000 for a genuinely capable premium mobile workstation or creator laptop, land around $3,000 to $4,500 for the mainstream high-end configurations, and reach $5,000 or higher for maxed-out GPU workstations or unusually high-memory local-AI systems. And for a system built around 128GB of memory, it could even be higher. The actual price could determine whether it competes with premium creator laptops or becomes a niche local-AI developer system.
While the primary announcements at the event will likely focus on the Surface Laptop Ultra and the Surface RTX Spark Dev Box, I expect third-party OEM pricing for partner laptops to be revealed as well.
Live Demos and the Next Chapter of Windows AI
We also expect to see live demos of Windows 11 agentic features. Autonomous agents have been the headline AI story this year, and with Satya Nadella, Jensen Huang, and Pavan Davuluri sharing the stage, we expect the keynote to feature live demos showing background agents like OpenClaw or Hermes Agent (or possibly a branded Microsoft equivalent) executing multi-step workflows in real time. Use cases in automated coding, video editing, and local file manipulation all seem incredibly likely—all running locally and offline on the Surface Laptop Ultra.
What makes this demo push so interesting is how dramatically Microsoft’s overall AI messaging has shifted. In response to widespread user pushback, Microsoft is retiring the “Copilot+ PC” branding and unbundling Copilot shortcuts from core Windows features—removing Copilot buttons from apps like Notepad, stripping AI feeds from Widgets, and letting users disable or remap the dedicated Copilot key to the standard Right Control key.
Microsoft conspicuously avoided calling the RTX Spark-powered Surface Laptop Ultra a Copilot+ PC at Build, and Surface vice president Brett Ostrum confirmed that new Surface devices won’t carry the label.
Microsoft is now pitching Copilot as an operating system? (Credit: Microsoft)
So, where does that leave Copilot? Nadella recently pitched Copilot as a standalone “OS for work,” signaling a major strategic divergence. Copilot is evolving into a cross-platform, cloud-centric productivity layer, while Windows 11 itself is refocusing on raw power, developer utilities, and local OS functionality. Instead of cluttering the desktop with cloud chat shortcuts, Microsoft’s new hardware narrative will likely emphasize raw local agentic compute powered by silicon like RTX Spark.
At this event, I expect Microsoft to articulate this new chapter, unveiling dedicated agentic features in Windows 11, and leaning into safety and transparency with sandboxing for local agents and adding guardrails and safety features, to make Windows the new home for autonomous AI tools as agents become mainstream and user-friendly.
Potential Hardware Surprises
Finally, a major onstage event like this is always ripe for a surprise hardware teaser. Microsoft could tease an RTX Spark desktop workstation along the lines of the Surface Studio, or even a gaming handheld concept built around Nvidia’s Arm chip to showcase native anti-cheat support and AVX2 Prism emulation. The company could even make a play for the dedicated AI hardware space with a novel personal AI device.
Whatever happens this week, one thing is already clear: The two biggest stories will be the launch of RTX Spark Surface hardware and the arrival of true, localized agentic AI on Windows.
About Our Expert
Brian Westover
Principal Writer, Hardware
Experience
From the laptops on your desk to satellites in space and AI that seems to be everywhere, I cover many topics at PCMag. I’ve covered PCs and technology products for over 15 years at PCMag and other publications, among them Tom’s Guide, Laptop Mag, and TWICE. As a hardware reviewer, I’ve handled dozens of MacBooks, 2-in-1 laptops, Chromebooks, and the latest AI PCs. As the resident Starlink expert, I’ve done years of hands-on testing with the satellite service. I also explore the most valuable ways to use the latest AI tools and features in our Try AI column.