NVIDIA’s RTX Spark Puts Six PC Makers Behind an Arm-Based Windows Bet

NVIDIA unveiled a complete system on a chip for Windows laptops at Computex 2026, a first for a company that has spent three decades selling graphics cards to somebody else’s machine. The chip, called RTX Spark, pairs a 20-core Arm processor with a Blackwell graphics core and up to 128GB of shared memory built for running AI agents on the device itself.

Microsoft, ASUS, Dell, HP, Lenovo and MSI all agreed to build RTX Spark laptops before NVIDIA named a single price. NVIDIA has run this exact play on this exact operating system once before, and it did not go well the first time.

NVIDIA Stops Selling Parts and Starts Selling the Whole PC

At his Computex 2026 keynote in Taipei, NVIDIA founder and chief executive Jensen Huang introduced RTX Spark as a new class of computer built for what the company calls personal AI agents. The official announcement describes a superchip delivering up to 1 petaflop of AI performance alongside full support for CUDA, the software layer that has anchored NVIDIA’s data center business for years.

The PC is being reinvented. For forty years, you launched apps. Click. Type.

Huang said the difference now is that a user simply asks, and the machine carries out the work on its own.

Microsoft chief executive Satya Nadella framed the partnership in similarly sweeping terms, saying the goal is to deliver unmetered intelligence to every home and every desk with Windows. The centerpiece is OpenShell, a new runtime NVIDIA and Microsoft built so agents can act inside Windows without constant cloud calls, a theme that carried straight into Microsoft’s own Build 2026 conference days later, where the company leaned into agents rather than a new version of Windows itself. The same keynote also produced a far bigger machine, the DGX Station for Windows, a deskside supercomputer built around NVIDIA’s separate GB300 chip that the company says can run models up to a trillion parameters, aimed at enterprise desks rather than laptop bags.

For three decades NVIDIA supplied graphics chips that other companies plugged into machines they designed, priced and sold themselves. RTX Spark folds the CPU, GPU, memory and AI software into one package NVIDIA controls end to end, leaving PC makers to build the chassis around it.

The Specs Behind the One-Petaflop Claim

The flagship configuration, code-named N1X inside NVIDIA, pairs the 20-core Grace CPU with a Blackwell GPU carrying 6,144 CUDA cores and fifth-generation Tensor Cores tuned for FP4 math, the low-precision format that makes the petaflop figure possible. NVIDIA connected the two halves with NVLink-C2C, a fast internal link, inside a 45 to 80 watt power budget thin enough for a laptop.

MediaTek, the Taiwanese chipmaker better known for smartphone processors, designed the custom memory controller behind RTX Spark’s unified memory pool, the detail that lets a laptop hold an entire AI model in memory instead of streaming it from a server rack.

That memory pool is doing real work rather than padding a spec sheet. NVIDIA says the platform can:

• Render 3D scenes larger than 90GB without swapping to a hard drive
• Edit 12K resolution video at 4:2:2 color depth in real time
• Run local AI models as large as 120 billion parameters with a million-token context window
• Play AAA games at 1440p above 100 frames per second using DLSS 4.5

Adobe is rebuilding Photoshop and Premiere around that same memory pool, with NVIDIA claiming roughly double the AI and graphics performance of earlier RTX laptop chips once the rewrite ships.

Six Laptop Makers Already Have Skin in the Game

NVIDIA named Microsoft Surface, ASUS, Dell, HP, Lenovo and MSI as launch partners, with Acer and Gigabyte confirmed to join later without named models yet. The company says more than 30 laptops and roughly 10 desktops carrying RTX Spark are already in development for a fall 2026 launch.

ASUS says its ProArt P16 and P14 are the first laptops built specifically around RTX Spark for creators and developers. HP, meanwhile, is pitching its OmniBook Ultra 16 and OmniBook X 14 as the thinnest RTX Spark machines on the market, according to Samuel Chang, HP’s senior vice president for consumer personal systems, who said developers are “moving from experimenting with AI to shipping agentic applications.” NVIDIA’s Jeff Fisher, senior vice president of personal computing, added that RTX Spark is “purpose-built for creators, gamers and AI developers in the personal AI era.”

The Price Nobody Will Confirm

NVIDIA has not published a price for a single RTX Spark machine. The closest thing to a number comes from Morgan Stanley’s channel checks with PC brands at the show, which point to flagship N1X configurations landing around $2,899, with the lighter N1 variant closer to $1,799. Those are analyst estimates, not NVIDIA guidance, and could shift before anything reaches a shelf.

• What we know: RTX Spark’s specs, its six launch partners and the fall 2026 window all come directly from NVIDIA’s own announcement, alongside more than 30 laptops and 10 desktops already confirmed in development.

• What’s unconfirmed: Official retail pricing, exact ship dates by region, and how completely the PC gaming library, anti-cheat systems included, will run on Arm silicon at launch.

Markets treated the announcement like a threat rather than a routine product reveal. Shares of AMD, Intel and Qualcomm all slipped after Huang’s keynote, and market-data firm TradingKey estimated Qualcomm alone lost more than $10 billion in value within hours, a figure not independently verified elsewhere. Qualcomm had the Windows on Arm field largely to itself since 2024, when its exclusivity agreement with Microsoft expired.

Windows Bet on Arm Before, and It Flopped

This is not NVIDIA’s first attempt at Windows on Arm. The company supplied the Tegra processor inside Microsoft’s Surface RT, the tablet that introduced Windows RT and is still widely remembered as a commercial failure once app support never fully arrived.

Reuters reported in October 2023 that NVIDIA was developing Arm-based chips for Windows as part of a Microsoft push to widen its Arm hardware options beyond a single supplier, Qualcomm, whose exclusive hold on Windows-compatible Arm chips was set to expire the following year. That reporting is now playing out on a stage in Taipei, more than a decade after the first attempt collapsed.

Will Your Old Software Actually Run on It?

Most everyday Windows software already runs natively on Arm or through Microsoft’s Prism emulator, which has matured considerably since Qualcomm’s first Snapdragon laptops arrived. The open questions for RTX Spark are narrower: specialized drivers, anti-cheat systems in older games, and whether NVIDIA’s software team can outrun the compatibility problems that shadowed Qualcomm’s launch.

Reporters who got early hands-on time with RTX Spark systems in Taipei found Prism handling demanding titles reasonably well, though NVIDIA acknowledged a large share of the PC gaming catalog will still lean on emulation rather than native Arm builds for now, according to TechPowerUp’s hands-on coverage from the show floor. NVIDIA does carry one advantage Qualcomm never had: decades of experience shipping its own GPU drivers, an area where Qualcomm’s Adreno graphics drivers have drawn persistent complaints from Snapdragon laptop owners.

Microsoft also appears to be treating RTX Spark differently than it treated Qualcomm’s chips. Windows 11 reportedly received kernel-level optimizations built specifically for RTX Spark, changes that were never made for Qualcomm’s Snapdragon platforms, per that same show-floor reporting.

The argument over Arm compatibility predates RTX Spark by years. Windows Central has argued that much of the criticism aimed at Snapdragon laptops recycles complaints Microsoft already fixed, noting that users of Windows on Arm PCs now spend the large majority of their time in apps that run natively. RTX Spark inherits that same compatibility layer on day one, for better and worse.

Intel and Qualcomm Are Fighting Over What’s Left

RTX Spark is aimed at the premium end of the laptop market. NVIDIA was not the only company redrawing lines at Computex 2026.

Intel pushed into handheld gaming for the first time, launching Arc G3 and Arc G3 Extreme chips built on the same architecture as its Core Ultra Series 3 laptop processors. Early handhelds carrying the new chips include Acer’s Predator Atlas 8 and a OneXPlayer model, alongside MSI’s Claw 8 EX AI+, a handheld that ends AMD’s long run as the default chip in Windows handheld gaming. Dan Rogers, an Intel vice president and general manager overseeing PC products, said the new chips prove that “while others make tradeoffs, gamers don’t have to.”

At the opposite end of the price scale, Qualcomm introduced the Snapdragon C Platform for students, families and small businesses rather than power users. Acer became the first PC maker to build around it with the Aspire Go 15, while ASUS paired Qualcomm’s higher-end Snapdragon X2 Elite chip with an 80 TOPS neural processor inside the Ascent QN10, calling it the first AI mini PC built on that platform. Kedar Kondap, Qualcomm’s senior vice president and general manager of compute and gaming, said Snapdragon C brings “value-oriented computing, all-day battery life, AI capabilities and responsive performance” to buyers previously priced out of AI features. ASUS also kept pushing its OLED ROG Xbox Ally X20 further into the same handheld race, a reminder that Computex’s cheapest and most expensive new Windows machines were announced on the very same stage.

Frequently Asked Questions

What is NVIDIA RTX Spark?

RTX Spark is NVIDIA’s first complete system on a chip for Windows PCs, combining a 20-core Arm-based Grace CPU with a Blackwell RTX GPU and up to 128GB of shared memory on a single TSMC 3-nanometer package. NVIDIA calls the flagship version N1X internally, with a smaller N1 variant planned for lighter, cheaper machines.

When can I buy an RTX Spark laptop?

NVIDIA says RTX Spark laptops and small desktops arrive in fall 2026 from Microsoft, ASUS, Dell, HP, Lenovo and MSI, with Acer and Gigabyte following afterward. A separate, larger machine NVIDIA calls the DGX Station for Windows, built around its GB300 chip for trillion-parameter models, is expected around the fourth quarter of 2026.

How much will an RTX Spark laptop cost?

NVIDIA has not announced pricing. Analyst estimates from Morgan Stanley point to roughly $2,899 for flagship N1X machines and about $1,799 for the lighter N1 tier, well below NVIDIA’s separate DGX Spark developer desktop, which was reportedly repriced to around $4,699 earlier this year.

Is RTX Spark the same thing as NVIDIA’s DGX Spark?

No. DGX Spark is a developer-focused desktop NVIDIA began shipping in late 2025 that handles AI models up to 200 billion parameters. RTX Spark is a newer, laptop-oriented platform built for consumers and creators, positioned well below DGX Spark despite sharing related Grace and Blackwell engineering.

Will RTX Spark laptops run my current games and software?

Most mainstream Windows software already runs natively on Arm or through Microsoft’s Prism emulator. NVIDIA demoed titles including Forza Horizon 6 and 007 First Light running on early RTX Spark hardware, though anti-cheat systems and some specialized drivers remain the biggest unresolved risk for gamers moving from x86 machines.

Which PC makers have not confirmed RTX Spark laptops yet?

Acer and Gigabyte are both named as future RTX Spark partners without specific models announced yet. Acer’s Computex 2026 lineup so far runs on Intel Core Ultra Series 3 and Snapdragon X2 chips instead, including its Aspire and Swift Spin 14 AI laptops.