Nvidia’s Jetson AGX Thor Now Available, Brings Real-Time Smarts to Future Robots

Nvidia, on Monday, announced the general availability of its Jetson AGX Thor development kit and production modules. Jetson Thor is the successor to the Jetson Orin chipset, which arrived in September 2022. The latest robotics platform significantly improves on both processing power as well as energy efficiency, and the company claims that it can enable physical artificial intelligence (AI) and robots — in particular, humanoid robots — to have real-time, intelligent interactions with people and the physical world. Nvidia also announced the list of companies that have already adopted the new computer system.

Nvidia Jetson Thor to Allow Real-Time Reasoning Inference for Robots

In a newsroom post, the tech giant announced the general availability of the new chipset and detailed its features. Developers and enterprises can now purchase the Jetson AGX Thor developer kit at the starting price of $3,499 (roughly Rs. 3 lakh). Additionally, the Jetson T5000 production modules are available from the company's distribution partners globally. Production systems and carrier boards for the platform can be separately purchased from Nvidia's embedded partners.

The Jetson Thor platform is powered by Nvidia's Blackwell GPU paired with 128GB of RAM. As per the company, it delivers up to 2,070 FP4 teraflops of AI compute (processing power), a 7.5X increase compared to its predecessor. The system can run on a power supply of 130W, marking 3.5X improved energy efficiency when compared to Orin.

Nvidia says that the new robotics chip is capable of running any generative AI model, including the complex systems designed for physical AI and robots. Highlighting an example, the company said that Jetson Thor can comfortably run vision language models (VLAs) such as Isaac GR00T N1.5.

This fast processing speed, which occurs on the cloud, accompanied by the improved power efficiency, enables Jetson Thor computers to offer real-time reasoning inference for humanoid robots. Put simply, the chipset, acting as the brain of a robot, can see an object, understand what it is, and execute the action in a split second.

In theory, this means individuals can order a robot powered by the platform (and a sophisticated AI system) to “open a black box and bring the white ball inside it,” and it should be able to follow the command without taking extra time to process the location of the box or to identify the ball.

Nvidia highlighted that the early adopters of the new platform include industry players such as Agility Robotics, Amazon Robotics, Boston Dynamics, Caterpillar, Figure, Hexagon, Medtronic and Meta. Additionally, companies such as 1X, John Deere, OpenAI, and Physical Intelligence are currently evaluating Jetson Thor.

“The future of robotics in logistics depends on the ability to deploy increasingly intelligent and autonomous systems. Nvidia Jetson Thor offers the computational horsepower and energy efficiency necessary to develop and scale the next generation of AI-powered robots that can operate safely and effectively in dynamic, real-world environments, transforming how we move and manage goods globally,” said Tye Brady, chief technologist at Amazon Robotics.