- Nvidia Jetson Thor is framed like the robot brain for physical AI
- Large robotics companies are already listed as the first adopters in the world
- Nvidia Jetson Thor offers 2,070 FP4 Teraflops in a power envelope of 130 Watts
NVIDIA has published the Developer Kit Jetson AGX Thor, calling the next step to robotic systems that can work in real time.
The system, built on the Blackwell GPU line, is formulated as a platform for “physical AI” and advanced robotic functions through manufacturing, logistics, health care, agriculture, retail and transport.
Nvidia says that it can provide up to 7.5 times more AI calculation and more than three times the energy efficiency of its Jetson Orin line, which has been widely used since 2022.
Offers a level of supercalculator level
Nvidia then described Jetson Thor as “the ultimate supercomputer to lead the age of physical AI and general robotics”.
“We have built Jetson Thor for the millions of developers working on robotic systems that interact and shape the physical world more and more,” said Jensen Huang, founder and CEO of Nvidia.
“With unrivaled performance and energy efficiency, and the ability to execute several generative AI models at the edge, Jetson Thor is the ultimate supercomputer to lead the age of physical and general robotics.”
With a figure cited of 2,070 FP4 Teraflops in an envelope of 130 Watts, it is positioned as powerful enough to execute several generative models at the same time.
It supports visual action models like Isaac Gr00T N1.5, as well as other LLM systems.
The device also incorporates 128 GB of memory, which should make it able to manage larger AI work flows at the edge.
Several robotics players are already listed as early adopters, notably Agility Robotics, Amazon Robotics, Boston Dynamics, Caterpillar, Hexagon and Medtronic.
Meta was also appointed early partner, while companies such as John Deere, Openai and Physical Intelligence would test the system.
“Nvidia Jetson Thor offers the power of calculation and the energy efficiency necessary to develop and set the next generation of robots fueled by AI which can work safely and effectively in dynamic and real environments, transforming our way of moving and managing goods on a global scale,” said Tye Brady, chief technologist of Amazon Robotics.
Nvidia notes more than two million developers already use its robotic battery, more than 7,000 customers who have deployed Jetson Orin equipment in Edge AI projects.
Jetson Thor works on the Nvidia Jetson software platform, which is designed to support several AI tools at a time.
The package fits into Nvidia Isaac for simulation, the Metropolis for Vision AI and Holoscan for the treatment of sensors in real time.
This arrangement is intended to allow a module system to take care of many models of writers and IA workflow, rather than requiring several distinct chips.
The Developer Kit is available now at $ 3,499 and production systems, including carrier cards, will be distributed worldwide through its partners.
