Nvidia's RTX Spark Superchip Brings Arm to Windows Laptops

For two decades, the Windows laptop has been a settled affair. Intel and AMD supplied the brains, the architecture was x86, and the only real drama was the annual contest over clock speeds and battery life. On the Computex stage in Taipei this week, Nvidia decided that settlement was over. Jensen Huang, the company's leather-jacketed founder and chief executive, held up a slim slab of silicon called the RTX Spark and declared the personal computer reinvented for the first time in 40 years. Coming from the man whose chips already underpin most of the world's artificial intelligence, it was less a product launch than a land grab.

The RTX Spark, built on an internal design known as N1X, is what Nvidia calls a superchip: an Arm-based central processor and a Blackwell graphics engine fused onto a single package. It will arrive this fall inside a fresh line of Windows machines from Microsoft, Dell, HP, Asus, Lenovo and MSI, with Nvidia promising more than 30 laptop models and 10 desktops over time. For a company that built its empire selling discrete graphics cards and data-center accelerators, putting a complete brain inside thin-and-light consumer laptops is a genuine departure, and a direct shot at rivals who thought the PC was their turf.

What Is Actually Inside the RTX Spark

The specifications explain why Nvidia is confident enough to call this a new category. The N1X processor pairs 20 Arm CPU cores with a Blackwell GPU carrying 6,144 CUDA cores, the same core count found in a desktop RTX 5070. It draws on a unified memory pool that scales to 128GB of LPDDR5X, delivers roughly one petaflop of AI compute, and links its CPU and GPU over an NVLink interconnect rated at 300GB/s. Built alongside MediaTek on TSMC's 3-nanometer process, the chip is efficient enough to fit in laptops as thin as 14 millimeters, just over half an inch.

What sets it apart from previous Arm-on-Windows attempts is the software underneath. The RTX Spark ships the full CUDA stack, the proprietary platform that has made Nvidia indispensable to AI developers. That means a laptop can run a 120-billion-parameter language model locally, without a round trip to the cloud, while still handling the games and creative tools that buyers expect. Nvidia is pitching the platform as the foundation for an "agentic" version of Windows, where an on-device assistant can read files, see the screen, and carry out multi-step tasks on a user's behalf.

A Coordinated Strike With Microsoft

The most striking feature of the announcement was the company standing next to Nvidia. "Microsoft and Nvidia are going to reinvent the PC," Huang said, framing the effort as a joint venture rather than a chip vendor courting an operating system. Microsoft's own Surface line is among the first wave of RTX Spark machines, and the two companies teased the partnership with coordinated social posts about "a new era of PC" in the days before the keynote.

That alignment matters because Microsoft has spent years trying to make Windows on Arm credible, with mixed results. The barriers were never the silicon alone but the ecosystem: anti-cheat software that refused to run, applications compiled only for x86, drivers that lagged. Nvidia is attacking those gaps head-on, citing native support for popular anti-cheat systems and leaning on its decades of relationships with game and software developers. "This is the first across the lineup of PC reinvention for 40 years," Huang told the audience, a claim that only holds if the software finally cooperates.

The Battle Over the AI PC

Nvidia is walking into a market long carved up by Intel, AMD, Qualcomm and Apple, and each has reason to be nervous for a different reason. Qualcomm spent the past few years as the standard-bearer for Arm-based Windows machines under its Snapdragon brand, but its exclusivity arrangement with Microsoft has lapsed, leaving an opening Nvidia is happy to fill. Intel and AMD, the incumbents, suddenly face a competitor with the strongest GPU and AI credentials in the industry attacking from outside the x86 world.

Apple is the more philosophical target. Its M-series chips proved that a tightly integrated Arm processor could deliver desktop-class performance with laptop battery life, and they reset expectations for what a premium notebook should be. Nvidia's answer is to match that integration while adding something Apple deliberately lacks: a desktop-grade discrete-class GPU and the CUDA software that AI developers actually use. The RTX Spark is expected to land in the premium tier, with early estimates placing it at or above the price of Apple's high-end MacBook Pro, though Nvidia did not confirm pricing on stage.

Investors grasped the stakes immediately. Nvidia shares rose nearly 4 percent in early trading after the keynote, while Intel and AMD each fell more than 3 percent, a clean illustration of who Wall Street thinks stands to gain and lose.

Vera Moves From Roadmap to Reality

The consumer story shared the stage with a data-center milestone. Nvidia's Vera CPU, the in-house Arm processor that anchors its next-generation server systems, has entered full production. Huang said the company is now manufacturing millions of the chips for "a market that never existed before," naming early customers including Anthropic, OpenAI, xAI, Oracle, Dell and CoreWeave.

Vera and the RTX Spark are two ends of the same strategy. "AI agents will be the largest users of computing," Huang said, casting Vera as the first CPU purpose-built to run agentic workloads at hyperscale. The same logic flows down to the laptop: if software agents are about to become the heaviest consumers of compute, Nvidia wants to own the silicon at every layer, from the AI factory to the machine on a writer's desk. "Nvidia has really become an infrastructure company," Huang said, "not just a GPU company."

The Roadmap Nvidia Wants You to See

Nvidia made clear the RTX Spark is a beginning, not a one-off. The company outlined a multi-generation roadmap in which every future architecture ships a companion Spark chip, with a Vera Rubin generation using faster LPDDR6 memory and a later Rosa Feynman generation beyond that. The message to OEMs and developers is that this platform is a long-term commitment, not an experiment to be abandoned if the first wave underperforms.

Whether buyers agree will depend on the things keynotes never settle: real battery life, real app compatibility, and whether an agentic Windows feels useful or gimmicky in daily use. Nvidia has the hardware lead and, for the first time, a willing Microsoft beside it. The fall lineup from Dell, HP, Asus, Lenovo, MSI and Microsoft will be the first honest test of whether the most powerful company in AI can also reinvent the device most people actually touch. For an industry that had grown comfortable with the status quo, the more uncomfortable possibility is that Huang is right.