Nvidia Pushes Into the PC With a New AI Superchip

Nvidia built a trillion-dollar business selling the chips that train AI models in distant data centers. At Computex in Taipei, the company made its clearest move yet to put that silicon under the keyboard of an everyday laptop.

The centerpiece was RTX Spark, an Arm-based processor for Windows machines that Nvidia is calling a superchip. It pairs a Blackwell graphics processor carrying 6,144 CUDA cores with a 20-core Arm CPU co-designed with MediaTek, and it can address as much as 128GB of unified memory. According to CNBC, the first RTX Spark laptops will come from Dell, HP, Lenovo, Microsoft, Asus and MSI later this year, in chassis as thin as 14 millimetres and aimed at creators, AI developers and gamers.

The significance is competitive as much as technical. For the first time Nvidia is selling a main PC processor, putting it head to head with Intel, AMD, Apple and Qualcomm in a market it has mostly left alone. Chief executive Jensen Huang framed the chip as a way to run AI models on the machine itself rather than shipping every request to the cloud.

A roadmap that runs to the end of the decade

Nvidia did not present RTX Spark as a one-off. Tom's Hardware reported a three-generation plan: the current Grace and Blackwell design this year, a Vera Rubin Spark generation with faster LPDDR6 memory in 2027 and 2028, and a Rosa Feynman Spark part toward the end of the decade.

The data center, still Nvidia's core business, got its own headline. Huang said the Vera CPU, built specifically to run AI agents, is now in full production, with millions of units being made. Early customers named on stage include Anthropic, OpenAI, Oracle, CoreWeave, Dell and Elon Musk's xAI. Nvidia claims Vera can complete agent-style tasks about 80 percent faster than the hardware it replaces, a pitch aimed squarely at companies trying to run swarms of automated software around the clock.

Whether RTX Spark reshapes the laptop market will depend on price and battery life, neither of which Nvidia detailed beyond calling the first machines premium. The ambition is plain enough. Nvidia wants the phrase AI PC to mean a computer with its name on the processor, not just its graphics card. Our earlier coverage of Dell's record AI server sales shows how much money is riding on the data-center side of that bet.