Nvidia launches RTX Spark chip to power AI agent PCs, reshaping Windows ecosystem

Nvidia has entered the Windows PC market with a new chip designed to run artificial intelligence agents directly on laptops and desktops, marking the company's most ambitious push into consumer computing since its failed attempt with the Surface RT in 2012. The RTX Spark, an ARM-based processor unveiled at Computex in Taipei, represents a fundamental shift in how the company sees personal computers working—not as machines you command, but as systems that anticipate your needs and handle routine tasks without intervention.

During his keynote address, Nvidia CEO Jensen Huang positioned the RTX Spark as the foundation of what he called a reinvention of the PC, a collaboration between Microsoft and Nvidia that he claimed was the first major redesign in four decades. The chip packs substantial computing power: up to 1 petaflop of AI performance, 128 gigabytes of unified memory, 6,144 GPU cores, and 20 CPU cores. Huang emphasized that the processor "runs everything the world has ever created, plus it runs agents"—a reference to AI systems that can automate tasks like scheduling, email management, or data analysis without requiring manual intervention at each step. The company will offer less powerful versions with as little as 16 gigabytes of RAM for different price points and use cases.

The practical applications Nvidia outlined suggest the company is thinking beyond gaming or creative professionals. On the creative side, the RTX Spark can render 3D scenes larger than 90 gigabytes, edit video at 12K resolution, and generate AI video in 4K. Adobe is optimizing Photoshop and Premiere for the chip from the ground up, with tools like Firefly-powered generative fill accelerated for faster performance. For gamers, the chip delivers AAA titles at 1440p resolution with over 100 frames per second, ray tracing, and Nvidia's DLSS upscaling technology. But the core pitch is about agents—local AI systems that run privately on your device, using frameworks like OpenShell and models such as OpenClaw or Hermes to handle work without sending data to the cloud.

The hardware itself reflects Nvidia's ambitions for a slim, portable form factor. RTX Spark laptops will measure as thin as 14 millimeters and weigh as little as three pounds, with screen sizes ranging from 14 to 16 inches. Microsoft is leading the charge with the Surface Laptop Ultra, which will ship with up to 128 gigabytes of RAM and a 15-inch mini-LED display capable of 2,000 nits of peak brightness. The device will arrive before the end of 2026. Asus, Dell, HP, Lenovo, and MSI have all committed to shipping RTX Spark laptops this fall, suggesting broad industry backing for the platform.

Beyond consumer devices, Huang used the keynote to articulate a broader vision of compute as the new engine of profit in enterprise AI. He unveiled Vera Rubin, a multi-rack supercomputer designed specifically for agentic AI workloads, and described it as "the most ambitious endeavour in Nvidia history." The system is twice the size of Nvidia's previous Grace Blackwell platform and now in full production. Huang stated that assembly time for a Vera Rubin stack has dropped from two hours to five minutes, and both Microsoft and Dell already have working systems. He also introduced new Vera CPUs built with a different architecture than traditional processors, optimized for the constant decision-making that AI agents require. OpenAI, Anthropic, and SpaceX are early adopters.

Nvidia also announced Nemotron 3 Ultra, an open-source AI model the company claims is five times faster and 30 percent cheaper than competing efficient models, with full transparency about its training data. The company introduced Cosmos 3, an open platform for physical AI, and Isaac Groot for humanoid robotics. For autonomous vehicles, Nvidia unveiled Alpamayo, described as the world's first reasoning autonomous vehicle model capable of explaining its decisions aloud as it drives.

The timing of the RTX Spark launch carries strategic weight. Qualcomm has held an exclusive license to produce ARM-based chips for Windows 11, but that arrangement appears to have ended, opening the door for Nvidia to compete directly in the PC processor market. Huang's repeated emphasis that "compute is revenue now" and "the absence of compute is now loss" signals how thoroughly Nvidia has repositioned itself—no longer primarily a gaming GPU maker, but a company betting that every layer of computing, from personal devices to enterprise infrastructure, will be built around AI agents that require constant, local processing power. The RTX Spark is the consumer-facing expression of that bet.