NVIDIA has officially rolled out the Jetson AGX Thor developer kit and production modules to address the needs of millions of developers working on robotic systems that increasingly interact with and shape the physical world.
The future of industries will be dependent on robots, especially in fields of manufacturing, logistics, healthcare, agriculture, and retail. Therefore, the Blackwell-powered kit is kitted with 128GB of memory for up to 2,070 FP4 teraflops of AI compute within a 130-watt power envelope (Also 7.5x better AI compute and 3.5x better energy efficiency), which is capable of running multiple generative AI models at the edge for simultaneous workflows, allowing robots to interact with people and environment in real time while processing data on the fly.
Adoption is already picking up, with Agility Robotics, Amazon Robotics, Boston Dynamics, Caterpillar, Figure, Hexagon, Medtronic and Meta among early users, while John Deere, 1X, OpenAI and Physical Intelligence are evaluating it for their next-gen robots. Brett Adcock of Figure highlighted that Jetson Thor enables real-time learning and interaction for humanoids, Tye Brady of Amazon Robotics said it will transform logistics with more intelligent and autonomous systems, and Caterpillar CEO Joe Creed pointed out its role in developing AI-powered mining and construction equipment for greater precision, safety, and efficiency.
Jetson Thor runs on the full-stack NVIDIA Jetson software platform, compatible with any popular AI framework and NVIDIA’s own ecosystem like Isaac for robotics simulation, Metropolis for vision AI and Holoscan for sensor processing.
Price and availability
The NVIDIA Jetson AGX Thor developer kit is available now starting at $3,499, while Jetson T5000 production modules can be purchased through worldwide distributors. Production systems and carrier boards are also available from embedded partners.
