TLDR: NVIDIA has officially launched Jetson Thor, a groundbreaking chip module designed as the ‘brain’ for advanced humanoid robots and physical AI systems. This new platform delivers an unprecedented 7.5 times increase in AI computing performance and boasts a staggering 2070 TFLOPS of computing power, marking a significant leap forward in real-time reasoning and autonomous capabilities for intelligent machines.
NVIDIA has announced the release of its most powerful ‘brain’ for humanoid robots, the Jetson Thor, also known as Jetson AGX Thor. This cutting-edge chip module is set to revolutionize the field of robotics by significantly enhancing AI performance and computing capabilities at the edge. Built on NVIDIA’s advanced Blackwell architecture, the Jetson Thor delivers a maximum AI computing power of 2070 FP4 teraflops and is equipped with a substantial 128GB of LPDDR5X memory. This represents a monumental leap in performance, offering more than 7.5 times the AI computing power, 3.5 times greater power efficiency, and 3.1 times the CPU performance compared to its predecessor, the NVIDIA Jetson Orin (or AGX Origin). The memory capacity has also been doubled, providing ample resources for complex AI models.
This immense power is designed to enable workflows previously deemed challenging, such as high-speed sensor data processing and inference processing from visual information, crucial for robots operating in dynamic, real-world environments. NVIDIA CEO Jensen Huang emphasized the platform’s significance, stating, “There are millions of developers around the world working on robot systems that interact with and have an increasing impact on the physical world, and Jetson Thor is a product designed for them.”
The Jetson Thor is engineered to support a wide array of applications, including humanoid robotics, industrial automation, surgical assistants, smart agricultural machinery, and autonomous driving systems through its integration with the DRIVE Thor platform. It is compatible with popular generative AI frameworks and models like Llama, Gemini, and Cosmos Reason, as well as robotics-specific platforms such as NVIDIA Isaac and NVIDIA Metropolis for video analysis AI agents.
Early adoption of the Jetson Thor is already underway by leading robotics companies. Agility Robotics plans to integrate Jetson Thor as its onboard computing platform, while Boston Dynamics is incorporating it into its ‘Atlas’ humanoid robot to handle on-device computational tasks that were previously server-dependent, accelerating AI workloads and processing high-bandwidth data. Automaker Xiaomi has also announced the integration of Jetson Thor into its ‘YU7’ SUV to enhance driver assistance functions. Furthermore, prominent research institutions including Stanford University, Carnegie Mellon University, and the University of Zurich are actively testing the platform for projects focused on navigation, perception, and planning models.
Industry experts are hailing the Jetson Thor as a potential ‘iPhone moment’ for humanoid robots, suggesting it could usher in a new era of physical intelligence. Brett Adcock, founder and CEO of Figure AI Inc., highlighted this, stating, “The key to developing powerful humanoid robots lies in our ability to run powerful AI models directly on the robot, enabling real-time learning and interaction.” This breakthrough allows humanoid robots to directly predict the consequences of actions through generative reasoning, execute complex tasks, and fundamentally understand the physical world and work processes. Wang He, CTO of Galaxy Universal, noted that the movement speed and fluidity of their G1 Premium robot have significantly improved after adopting Thor.
Also Read:
- Generative AI Propels Robotics Market Towards Explosive Growth, Led by NVIDIA and OpenAI
- Nvidia Pledges $500 Billion for US AI Infrastructure and Supercomputer Production in Texas
The Jetson Thor Developer Kit is currently available for $3,499, with production modules sold in bulk starting at $2,999 per unit. Shipping for the development kits is anticipated to begin in September 2025.
