For the better part of two years, AI PCs have been a category in search of a compelling reason to exist. Qualcomm launched its Snapdragon X chips with bold promises about battery life and on-device AI. Intel followed with efficiency-focused Lunar Lake processors. The result, for most consumers, has been a quiet shrug. The apps they rely on still run better on x86. The AI features bundled with these machines have felt more like marketing than utility.
NVIDIA and Microsoft intend to change the conversation entirely.
On May 31, 2026, at Nvidia’s GTC Taipei event during Computex, Jensen Huang and Microsoft CEO Satya Nadella took the stage together to announce the Nvidia RTX Spark, a new superchip that the companies describe as a fundamental reinvention of the Windows PC. Huang framed it bluntly: “For forty years, you launched apps. Click. Type. With RTX Spark and Microsoft Windows, you ask, and the PC does the work.”
The RTX Spark will arrive this fall in up to 30 laptops and 10 compact desktop PCs from brands including ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI. Whether it delivers on that promise, and who it is actually built for, are the more complicated questions.
What Is the RTX Spark, Exactly?
RTX Spark is an Arm-based system-on-chip (SoC) developed by Nvidia in collaboration with MediaTek and manufactured on TSMC’s 3nm process node. It combines two distinct computing architectures on a single package:
- A 20-core Nvidia Grace CPU, based on Arm v9 architecture
- A Blackwell RTX GPU with 6,144 CUDA cores and fifth-generation Tensor Cores with FP4 precision
- Up to 128 gigabytes of unified LPDDR5X memory, shared across both the CPU and GPU
- NVIDIA’s NVLink-C2C interconnect, which links the CPU and GPU with extremely high bandwidth
The result is a chip rated at up to one petaflop of FP4 AI performance, capable of running 120-billion-parameter large language models entirely on-device, rendering 3D scenes larger than 90 gigabytes, editing 12K video, and playing AAA games at 1440p resolution above 100 frames per second (aided by DLSS 4.5 and frame generation).
For context, Nvidia’s entire CUDA software stack, the same foundation that powers the company’s data center chips and the world’s most widely used AI frameworks, runs natively on RTX Spark. That is not a trivial claim.
This Chip Has a History
RTX Spark did not appear without warning. The project has been in development since at least 2023, when Reuters first reported that Nvidia was building Arm-based CPUs capable of running Windows. The timing was strategic: Microsoft’s exclusivity arrangement with Qualcomm for Windows-on-Arm chips was set to expire in 2024, opening the door for other chip designers to enter the space.
The chip’s closest existing relative is the GB10 Grace Blackwell superchip already shipping inside Nvidia’s DGX Spark, a Linux-based compact AI workstation aimed at researchers and developers. The DGX Spark and RTX Spark share essentially the same core silicon and unified memory architecture. The critical difference is the operating system: DGX Spark runs a customized version of Ubuntu, while RTX Spark is built specifically for Windows 11 on Arm.
Microsoft’s role here goes beyond simply supplying the operating system. The two companies have co-developed new Windows security primitives designed to run on-device AI agents, alongside Nvidia’s OpenShell runtime, a framework for executing automated tasks locally without sending data to the cloud.
What Can It Actually Do?
The headline capability is local AI inference. Because RTX Spark keeps all computation on-device, users can run large AI models, generate outputs, and execute automated workflows entirely without an internet connection or a cloud subscription.
NVIDIA and Microsoft specifically highlight:
- Personal AI agents that can run continuously in the background, executing tasks and generating content on demand through integration with the Windows taskbar
- 120-billion-parameter LLM inference with up to one million tokens of context, which covers very large documents or extended conversations
- Creative production at scales that have historically required dedicated workstations: 12K video editing, 90GB-plus 3D scene rendering, and 4K AI video generation
- Gaming with full RTX support including ray tracing, DLSS 4.5, Nvidia Reflex, and G-SYNC
Adobe is restructuring both Photoshop and Premiere from the ground up to target RTX Spark, with the company claiming this will deliver twice the AI and graphics performance compared to current hardware. Over 100 software developers have announced support for the platform at launch, including Blackmagic Design, Blender, CapCut, ComfyUI, OTOY, and game developers such as Riot Games, Remedy Entertainment, and KRAFTON.
Who Is Building RTX Spark Machines?
Eight specific models were announced at Computex 2026. NVIDIA says the lineup will expand to approximately 30 laptops and 10 desktop PCs by the fall release window.
Announced models at launch:
- Microsoft Surface Laptop Ultra
- ASUS ProArt P16 (H7607)
- ASUS ProArt P14 (H7407)
- Dell XPS 16 Creator Edition (with tandem OLED display)
- HP OmniBook Ultra 16
- HP OmniBook X 14
- Lenovo Yoga Pro 9n
- MSI Prestige N16 Flip AI+ (a 16-inch convertible with UHD+ tandem OLED display)
Additional models from Acer and Gigabyte are expected to follow. HP has positioned its OmniBook X 14 as the thinnest RTX Spark laptop announced so far, underlining Nvidia’s emphasis on thin-and-light form factors rather than traditional high-performance chunky chassis.
One important note: RTX Spark is not available as a standalone processor that you can purchase and install yourself. It will only be available in prebuilt OEM designs from these manufacturers.
How Much Will These Machines Cost?
Nvidia has not released official pricing. However, analyst estimates provide a realistic range.
According to Morgan Stanley research cited by industry sources, machines using the full N1X chip (the RTX Spark’s internal codename) are expected to start at around $2,899, while devices using the more power-efficient N1 variant could start closer to $1,799.
For comparison, Nvidia’s existing DGX Spark, which shares the same core silicon but runs Linux and targets enterprise workloads, currently retails between approximately $3,999 and $5,000, with prices elevated partly due to memory component constraints.
The RTX Spark laptops are being positioned to rival the Apple MacBook Pro in both price and target audience. Whether the value proposition holds up against Apple Silicon, which has had years to mature its software ecosystem on Arm, remains one of the central questions heading into the fall release.
The Competitive Picture: Why This Move Matters
This is Nvidia’s first serious entry into the Windows PC chip market, and it carries implications well beyond a product launch.
It challenges Intel and AMD directly. For over four decades, Windows PCs have run on x86 processors designed by Intel or AMD. NVIDIA is now introducing an Arm-based alternative with data center-class AI credentials and a fully integrated GPU, targeting exactly the professionals who currently buy premium Windows laptops powered by Intel Core Ultra or AMD Ryzen AI chips.
It capitalizes on Qualcomm’s stumbles. Qualcomm has been the primary ARM chip supplier for Windows PCs since its exclusivity arrangement with Microsoft began, but the Snapdragon X lineup has faced persistent software compatibility problems. Professional tools like AutoCAD have run poorly or not at all under x86 emulation, and games have crashed under certain conditions. Intel’s Lunar Lake effectively neutralised Qualcomm’s battery life advantage while retaining full native x86 compatibility. NVIDIA enters this space with a stronger software story: CUDA runs natively, meaning no emulation layer for any part of NVIDIA’s AI stack.
It completes Nvidia’s hardware presence across the computing stack. NVIDIA already dominates AI workloads in data centers and cloud infrastructure. With RTX Spark, the company now has a premium consumer-facing hardware platform for local AI computation. Competitors like Intel, AMD, and Qualcomm have long owned both the consumer and enterprise ends of the computing market. NVIDIA is closing that gap.
The Risks Are Real
The RTX Spark announcement is significant, but it is not without meaningful risk.
Software compatibility remains the ARM problem on Windows. Qualcomm’s experience has demonstrated that moving professional Windows users to ARM is harder than it looks. Many enterprise applications are built for x86, and even with Microsoft’s Prism emulation layer, compatibility gaps appear. NVIDIA and MediaTek are newer to this specific consumer space than Qualcomm. Early adopters should expect some software edge cases and driver maturity issues, particularly in the first months after launch.
The addressable market is still small. Even a successful RTX Spark launch will represent a relatively minor contribution to Nvidia’s overall revenue. The company generates the overwhelming majority of its income from enterprise data center hardware and cloud AI infrastructure. Consumer AI PCs, even premium ones at $2,899 or above, are a niche segment by comparison. This product is strategically important for Nvidia’s long-term positioning, not a near-term revenue driver.
The price point narrows the audience significantly. At $2,899 and above for initial models, RTX Spark machines are targeting a narrow slice of the market: AI developers, serious creative professionals, and technically sophisticated users who specifically need local AI inference at scale. Mainstream consumers who want a capable everyday laptop will find far more affordable options elsewhere, including capable x86 machines with long battery life.
Apple Silicon is a mature, well-optimized competitor. The MacBook Pro with M4 Pro or M4 Max has years of software optimization, a refined developer ecosystem, and a proven track record for professional workloads. RTX Spark’s CUDA advantage is real, but it matters primarily to users already working within Nvidia’s AI development stack. For everyone else, the comparison is more nuanced.
What This Means for Buyers
If you are a developer building and testing AI models locally, the RTX Spark’s native CUDA support and 128GB of unified memory represent a genuinely compelling combination in a laptop form factor. Previously, running 120-billion-parameter models locally required either an enterprise-grade workstation or a DGX Spark running Linux. RTX Spark brings that capability to a thin Windows laptop.
If you are a creative professional working with large video or 3D projects, Adobe’s specific optimization work for Premiere and Photoshop is worth watching closely. The performance claims around 12K video and 90GB scene rendering are meaningful if the software support delivers.
If you are a gamer or a general productivity user, the value proposition is harder to justify at this price point. The gaming performance is real, but competing RTX 5070-class gaming laptops are available at significantly lower prices on conventional x86 hardware. And for productivity work, $2,899 buys a great deal of capable laptop from established manufacturers without the software risk of a first-generation Windows-on-Arm platform from a new chip vendor.
The honest summary: RTX Spark is a first-generation product targeting a specific and technically demanding audience. It is not designed for everyone, and its pricing reflects that clearly.
Game Changer:
The AI PC category needed a more compelling argument than better battery life and marginally faster app launches. NVIDIA’s entry into the Windows laptop chip market with RTX Spark is the most technically substantial attempt yet to make that argument. Bringing data center-class AI performance, CUDA compatibility, and 128GB of unified memory to a thin laptop chassis is a meaningful engineering achievement.
Whether it translates into a mass-market product category is a separate question entirely. The pricing, the software maturity risks, and the narrow initial audience all suggest this is a first-generation move designed to establish Nvidia’s presence and protect its long-term strategic position, not a product that will redefine computing for everyday users this year.
What it does definitively change is the competitive landscape for premium Windows laptops. Intel, AMD, and Qualcomm now have Nvidia to contend with at the high end. That competition, regardless of how RTX Spark itself sells, is likely to push better hardware and better AI capabilities across the entire Windows ecosystem over the next several years. In that sense, the most important effects of this announcement may not be visible at launch at all.
