NVIDIA announces powerful modules for robotics and physical AI

NVIDIA is introducing a new generation of Jetson system-on-module, the Jetson Thor, which breaks new ground in robotics and physical AI. The company’s CEO, Jensen Huang, believes that robotics will be one of the biggest opportunities for chipmakers beyond artificial intelligence in general.

NVIDIA’s CEO, Jensen Huang, said that robotics will be one of the biggest opportunities for chipmakers beyond artificial intelligence.

The company has been developing “robotic brains” for several years, and each new generation surpasses the previous one in power and efficiency. The new model features a Blackwell GPU architecture that delivers 7.5 times more computational power for AI and 3.5 times more energy efficiency than previous solutions. The Jetson Thor modules are capable of running generative AI models – large language and visual models – allowing robots to better interpret and interact with the world around them.

“We created Jetson Thor for the millions of developers building robotic systems that interact with and increasingly shape the physical world,” Huang said. The module runs on NVIDIA’s full stack Jetson software platform, which is specifically designed for physical AI and robotics applications.

The customer base includes companies such as Amazon, Meta*, Agility Robotics and Boston Dynamics, demonstrating the high demand for this technology in the industry. The Jetson AGX Thor developer kit is on sale now, priced at $3499. NVIDIA also plans to sell Thor T5000 modules for integration into production-ready robots. The minimum order for bulk quantities will be 1,000 units at $2999 per module.

* Owned by Meta, it is recognized as an extremist organization in the Russian Federation and its activities are banned.