NVIDIA has introduced Jetson Thor, a new line of onboard computers built to push robots and autonomous machines out of research labs and into everyday, mass-market use.
According to the NVIDIA Blog, the company unveiled two modules, the T3000 and T2000, aimed at what it calls mainstream robotics and edge AI. The pitch is that general-purpose robots and autonomous machines are shifting from experimental projects toward real-world deployment, and that transition is driving demand for compact, power-efficient "AI supercomputers" small enough to sit inside a machine.
The key idea is running advanced AI where the action happens. NVIDIA says Jetson Thor is designed to run foundation models — the same broad, powerful type of AI behind today's chatbots and image tools — directly at the edge, meaning on the device itself rather than in a distant data center. That matters for robots, which need to sense and react in real time without waiting on a network connection.
Coverage from outlets including GamesBeat and Wccftech frames the launch around "physical AI" and humanoid robots, positioning the chips as the brains for a coming wave of machines that move and act in the physical world.
One notable engineering claim comes from Crypto Briefing, which reports that NVIDIA has shrunk the robotics chip roughly in half in size while maintaining performance with the new Jetson AGX Thor. Smaller, equally capable hardware makes it easier to fit serious computing power into robots, drones and other autonomous machines.
NVIDIA did not, in these sources, disclose pricing or specific customer deployments.
Why it matters: the hardware that lets AI leave the screen and operate in the real world is a bottleneck for the entire robotics industry, and NVIDIA is betting that cheaper, smaller, more powerful onboard computers will help turn long-promised general-purpose robots into products people actually buy.