Nvidia RTX Spark superchip targets AI agents on Windows PCs

Nvidia is moving beyond its dominant position in data center chips to challenge established players in personal computing. At the Computex conference in Taipei, CEO Jensen Huang introduced the RTX Spark, an Arm-based superchip developed with Microsoft that combines a central processing unit and a high-end graphics processor on a single package. The chip will appear first in laptops from Microsoft, Dell, HP, Asus, Lenovo and MSI beginning this fall.

The flagship version pairs a Blackwell graphics unit with 20 Arm CPU cores from a design by MediaTek and includes up to 128 gigabytes of unified memory. Nvidia says the device can deliver roughly the graphics performance of an RTX 5070 laptop GPU while consuming far less power, and it claims the full system reaches one petaflop of AI throughput. Lower-cost versions with as little as 16 gigabytes of memory are planned later. Because the architecture is Arm-based, existing Windows software written for Intel or AMD processors will run through an emulation layer, an approach Microsoft has refined over several years.

Huang framed the launch as a fundamental reset for the personal computer. He compared the shift to the move from basic mobile phones to smartphones, arguing that agentic AI systems capable of acting across applications will require new hardware foundations. Microsoft’s Surface Laptop Ultra will be among the first devices, featuring a 15-inch mini-LED display with 2,000 nits peak brightness, a large haptic trackpad and multiple ports including HDMI, USB-C and full-size SD. Other partners have shown early designs that range from thin-and-light notebooks to compact desktops.

The move places Nvidia in direct competition with Qualcomm’s Snapdragon X series, AMD’s Ryzen AI processors and Apple’s M-series chips, all of which already use Arm cores. Intel, long the default supplier for Windows machines, faces another entrant that bundles strong graphics and AI acceleration. Industry analysts note that the same silicon family already powers Nvidia’s DGX Spark workstation for AI developers, suggesting the company is repurposing high-performance designs for broader consumer use.

Early performance claims include the ability to render large 3D scenes, edit high-resolution video and run demanding games at high frame rates without being plugged in. The chip also meets Microsoft’s threshold for Copilot+ features, though Nvidia is emphasizing its tensor cores over a dedicated neural processing unit. Battery life is described in general terms as “all day,” consistent with other Arm-based Windows devices.

The announcement arrives as the PC market continues to absorb AI features into mainstream products. Previous attempts to introduce Arm Windows machines encountered software compatibility hurdles, but Microsoft’s investment in its Prism emulator and broader ecosystem support may reduce those frictions. Whether the RTX Spark can capture meaningful share will depend on pricing, software optimization and the willingness of major manufacturers to promote the new platform alongside their existing Intel and AMD lines.

For now, the launch signals that Nvidia intends to extend its influence from cloud training clusters into the devices people use every day, potentially reshaping both hardware supply chains and the kinds of AI applications that run locally rather than in distant data centers.