Nvidia on Sunday unveiled the first personal laptops designed to run artificial-intelligence “agents,” introducing a new chip called RTX Spark and a broader hardware suite known as Vera Rubin at the Computex conference in Taipei.
The company described RTX Spark as “the most efficient PC chip ever built.” The new PCs will be as slim as 14 millimeters thick, and the lightest will weigh less than 3 pounds. Mark Aevermann, Nvidia’s senior director of product development, said the computers will be “targeted at creators, AI developers and gamers” and priced at the premium end of the market. Eventually, Nvidia said, there will be 30 laptop models and about 10 desktop models using the new chips, which are developed from Nvidia’s graphics-processing unit architecture.
Nvidia will work with six manufacturers to produce the initial laptops: Dell Technologies, Lenovo Group, Microsoft, HP, Asus and MSI.
The announcement underscored how AI agents — autonomous bots capable of performing many tasks — have begun reshaping the tech industry in a matter of months. Nvidia couched the PC launch as part of a broader transformation in AI computing: away from the era in which consumers hold conversations with chatbots such as OpenAI’s ChatGPT, Google’s Gemini and Anthropic’s Claude.
“That era is ending,” Kari Briski, Nvidia’s vice president for generative AI software, said. “Agents are the new workload. They will run everywhere, from the data center to the edge.”
Although demand for GPUs to train AI models made Nvidia the world’s most valuable company, agentic computing leans heavily on central-processing units, or CPUs, Nvidia said. The shift has led the company to redesign its silicon products and marketing around autonomous-agent workloads rather than conversational AI.
Ian Buck, Nvidia’s vice president for hyperscale and high-performance computing, said the rise of agentic AI requires more than powerful chips alone. It requires advanced networking hardware, software libraries for developers, and large data-center clusters that knit together tens of thousands of processors.
“AI is moving from answering questions to doing real work,” Buck said.
Nvidia also said it is expanding its partnership with Chinese robotics company Unitree. The companies introduced a robotics design template that other companies can replicate, with Nvidia providing the brains and software and Unitree manufacturing the physical components. A company official said the arrangement is the only way to accelerate domestic U.S. robotics activity on a large scale given China’s dominance in producing physical robot parts.
The official said data would remain with the user and that other precautions would address concerns about Chinese government influence, adding that U.S. companies are among Unitree’s biggest customers.
The Vera Rubin suite includes Nvidia’s most powerful GPU to date, the Rubin, along with servers consisting only of Vera CPUs and a system incorporating chips customized for inference and designed by Groq. Nvidia paid $20 billion last year to license Groq’s technology and hire its leadership.