Nvidia launches $3,499 robotics development platform

Nvidia has launched the Jetson AGX Thor, a $3,499 robotics AI computer aimed at advancing real-time reasoning in robots and AI-driven physical systems. The announcement comes following CEO Jensen Huang's past hints about delivering a "new brain" for intelligent machines. Production modules, sold in bulk, start at $2,999 per unit, according to an official blog post by Nvidia.

Built on Nvidia's Blackwell architecture, Jetson Thor delivers 2,070 FP4 teraflops of compute performance. As per the company, it provides 7.5 times more AI compute, 3.1 times more CPU performance, and double the memory of its predecessor, Jetson Orin. This boost is intended to help robots process data from multiple sensors with low latency, enabling applications such as visual reasoning and multimodal AI in dynamic, real-world environments.

Early adopters include robot development companies Agility Robotics and Boston Dynamics, which plan to integrate Jetson Thor into the next generations of their humanoid robots. Research institutions such as Stanford University, Carnegie Mellon University, and the University of Zurich are also testing the platform for projects on navigation, perception, and planning models, says Nvidia.

Jetson Thor supports popular generative AI frameworks and models, including Llama, Gemini, and Cosmos Reason, as well as robotics-specific platforms like Nvidia Isaac. The platform is designed for developers working on applications ranging from humanoid robots and industrial automation to surgical assistants and smart agricultural machinery.

Nvidia said the hardware is available now through its partners, while the DRIVE AGX Thor developer kit for autonomous vehicles is open for preorder with deliveries expected to start in September.