In a significant push to accelerate the future of autonomous machines, NVIDIA has unveiled a comprehensive suite of new models and hardware designed to empower robots with the ability to learn, reason, and interact with the physical world. This initiative is centered around Project GR00T, a general-purpose foundation model for humanoid robots, and is supported by a powerful new computer, Jetson Thor, alongside substantial upgrades to the NVIDIA Isaac robotics platform. The announcements signal a concerted effort to create a holistic ecosystem for developing and deploying intelligent robots, from industrial automation to humanoid assistants, marking what many in the industry see as an inflection point for artificial general robotics.

At the heart of this new era is Project GR00T, which stands for Generalist Robot 00 Technology.[1] This foundation model is designed to enable robots to understand natural language and learn skills by observing human actions.[1] The goal is for robots powered by GR00T to quickly learn coordination, dexterity, and other complex skills necessary to navigate and interact with the real world in a human-like manner.[1] This is achieved through a multimodal approach, where the AI can learn from a variety of inputs including text, video, and live demonstrations.[2] The GR00T model features a dual-system architecture, mirroring human cognition with a fast-thinking "System 1" for reflexive actions and a "System 2" for more deliberate, reasoned planning.[3] This allows the robot to not only perform tasks but also to understand its environment and plan its actions accordingly. Leading robotics companies such as Boston Dynamics, Figure AI, and Agility Robotics are already exploring the capabilities of this new technology.[4][1]

Powering these advanced AI models directly on the robot is the newly announced Jetson Thor robotics computer.[4][5] This powerful system is based on NVIDIA's Blackwell architecture and is specifically designed to run the complex, multimodal generative AI models like GR00T in real time at the edge.[6][1] Jetson Thor delivers a massive leap in performance, offering 2,070 FP4 teraflops of AI compute, which is over seven times the performance of its predecessor, the Jetson Orin.[6][7] This server-class performance in a compact, power-efficient module is critical for enabling robots to perceive, reason, and act in complex and unstructured environments without constant reliance on the cloud.[6][8] The new computer is already being adopted by a range of partners, including Figure AI, Google DeepMind, and Meta, for the development of next-generation robots.[4][5][9]

Supporting the development and deployment of these intelligent robots is the NVIDIA Isaac platform, which has received significant upgrades. This platform provides a comprehensive set of tools, including Isaac Manipulator for AI-enabled robotic arms and Isaac Perceptor for autonomous mobile robots.[4][10][11][12] Isaac Perceptor offers advanced 3D surround-vision capabilities, crucial for robots operating in dynamic environments like warehouses and factories.[4][11] Isaac Manipulator simplifies the development of arms that can perceive and interact with their surroundings.[4] A key component of the platform is Isaac Sim, a simulation environment built on NVIDIA Omniverse that allows for the testing and validation of robots in physically-based virtual worlds.[4] This allows developers to generate vast amounts of synthetic data for training and to safely test robotic skills before real-world deployment.[4][13] This simulation-first approach is crucial for accelerating development and ensuring the safety and reliability of autonomous machines. The platform also includes the open-source Newton Physics Engine, co-developed with Google DeepMind and Disney Research, which allows for more realistic simulation of complex interactions.[4][13][14][15]

Underpinning this entire robotics ecosystem is NVIDIA's powerful AI infrastructure. The company has unveiled new hardware to handle the demanding workloads of training and running these advanced robotic models. The NVIDIA GB200 NVL72 is a rack-scale system that integrates 72 Blackwell GPUs and 36 Grace CPUs, providing unprecedented performance for AI training and inference.[4][5][15][16][17] This system is being adopted by major cloud providers to accelerate the development of complex AI.[4][5][15] Additionally, new RTX PRO Servers offer a unified architecture for the entire robot development workflow, from synthetic data generation and simulation to robot learning.[4][5] This comprehensive hardware and software stack provides a clear pathway for developers to take their robotic creations from concept to reality, a move that NVIDIA believes will accelerate the arrival of an era where everything that moves will one day be autonomous.[4] The broad adoption of these tools by leaders in the robotics industry indicates a strong belief that this integrated platform will be instrumental in solving the complex challenges of creating truly intelligent and capable robots.[4][18]

In conclusion, NVIDIA's latest announcements represent a watershed moment for the field of robotics and physical AI. By providing an end-to-end platform that spans from foundational models like GR00T to on-robot computers like Jetson Thor and a comprehensive simulation and software stack with the Isaac platform, the company is laying the groundwork for a new generation of intelligent machines. The focus on general-purpose, human-like reasoning and learning capabilities, combined with the raw computational power to run these models in the real world, has the potential to unlock applications across numerous industries, from manufacturing and logistics to healthcare and personal assistance.[18][19] As these technologies mature and are adopted by the wider robotics community, the vision of truly autonomous robots that can learn, reason, and act safely and effectively alongside humans is moving ever closer to becoming a reality.