GR00T, standing for Generalist Robot 00 Technology, will enable robots to understand natural language and mimic human movements, swiftly acquiring coordination, dexterity, and other skills to navigate, adapt, and interact with the real world. (Source: Image by RR)

Introducing Jetson Thor: NVIDIA’s Next-Gen Computing Platform for Robots

NVIDIA has introduced Project GR00T, a versatile foundation model tailored for humanoid robots, aimed at advancing robotics and embodied AI. Accompanying this initiative, the company has unveiled Jetson Thor, a new computer optimized for humanoid robots, and significant upgrades to the NVIDIA Isaac™ robotics platform, including generative AI foundation models and tools for simulation and AI workflow infrastructure. As noted in, Jensen Huang, NVIDIA’s founder and CEO, emphasized the significance of developing foundation models for general humanoid robots, highlighting the potential for these robots to understand natural language and mimic human movements, facilitating navigation, adaptation, and interaction in real-world scenarios.

Jetson Thor, designed as a dedicated computing platform for humanoid robots, boasts a modular architecture optimized for performance, power efficiency, and size. Equipped with the NVIDIA Blackwell architecture-based GPU, featuring a transformer engine delivering 800 teraflops of AI performance, Jetson Thor is geared to support multimodal generative AI models like GR00T. With integrated functional safety features, a high-performance CPU cluster, and ample ethernet bandwidth, Jetson Thor simplifies design and integration efforts, catering to the diverse needs of humanoid robotics applications.

NVIDIA’s collaboration with leading humanoid robot companies underscores its commitment to driving innovation in robotics. Partnerships with industry players such as Agility Robotics and Sanctuary AI aim to leverage AI advancements to revolutionize labor and address pressing societal challenges. Jonathan Hurst of Agility Robotics emphasized the transformative potential of human-centric robots like Digit, highlighting NVIDIA’s role in providing the essential computing infrastructure and machine learning environments necessary for realizing this vision.

The Isaac platform, integral to Project GR00T, introduces significant enhancements to support embodied AI models for diverse robot embodiments and environments. Isaac Lab, powered by GPU acceleration, enables parallel simulations crucial for robot learning, while OSMO facilitates data generation, model training, and software/hardware-in-the-loop workflows across distributed environments. Additionally, Isaac Manipulator and Isaac Perceptor offer state-of-the-art dexterity and perception capabilities for robotic arms and multi-camera 3D surround-vision functionality, respectively, catering to the evolving needs of autonomous mobile robots in various industries.

These advancements are poised to revolutionize robotics development, empowering developers to create more capable and intelligent humanoid robots capable of addressing real-world challenges and transforming industries. The comprehensive suite of tools and capabilities provided by NVIDIA’s Project GR00T and Isaac platform underscores the company’s commitment to driving breakthroughs in robotics and embodied AI, shaping the future of human-robot interaction and automation.