Confront Nvidia CEO Jensen Huang when he says, “Building a foundational model for a common humanoid robot is one of the most exciting problems to solve in AI today.” It's difficult. The humanoid form factor is currently one of the most hotly contested topics in the world of robotics, raising large amounts of venture capital but also generating significant skepticism in the process.
Naturally, Nvidia wants the product. The chip giant has become perhaps the most important hardware company in his AI space, and has recently made a convincing case as a powerhouse of robotics innovation through initiatives such as Isaac and Jetson. At the annual GTC developer conference this week, the company is planting the flag in the humanoid race with Project GR00T. This may or may not be a nod to Marvel's Illyrist Talking Space Tree.
The chipmaker calls the new platform a “universal foundation model for humanoid robots.” Essentially, the company is an AI platform for recent entrants into the space, including companies like 1X Technologies, Agility Robotics, Apptronik, Boston Dynamics, Figure AI, Fourier Intelligence, Sanctuary AI, Unitree Robotics, and XPENG Robotics. is building. This covers nearly every prominent humanoid robot manufacturer at the moment, with a few notable exceptions like Tesla.
Quoting co-founder and chief robotics officer Jonathan Hurst, the announcement adds FaceTime to Agility. “We are at a turning point in history, where human-centered robots like Digit are poised to change the workforce forever. Modern AI is accelerating development and robots like Digit are poised to touch every aspect of daily life. In partnership with NVIDIA, we are partnering with NVIDIA to bring computing, simulation tools, machine learning environments, and other We are excited to invest in the necessary infrastructure.”
Geordie Rose, co-founder and CEO of Sanctuary AI, agrees. “Embodied AI will not only help address some of humanity's biggest challenges, but it will also generate innovations that are currently beyond our reach or imagination. We prioritize long-term partners like NVIDIA because critical technology shouldn't be built in silos.”
GR00T also supports new hardware from Nvidia. Keeping the Marvel Cinematic Universe going is Jetson Thor, a new computer designed specifically to run simulation workflows, generative AI models, and more for humanoid form factors. I keep warning people against casually throwing around terms like “general purpose” when describing these machines, but Nvidia's strong interest is validation for a category that will almost certainly accelerate development.
Nvidia has this to say about the new silicon:
This SoC includes a next-generation GPU based on NVIDIA Blackwell architecture with Trans Engine that delivers 800 teraflops of 8-bit floating point AI performance to run multimodal generative AI models like GR00T. Masu. Integrating a functionally safe processor, high-performance CPU cluster, and 100 GB of Ethernet bandwidth greatly simplifies design and integration efforts.
Universality is still years away, but democratizing access for third-party developers will go a long way toward closing that gap.
This week's GTC robotics announcements included two more significant programs: Isaac Manipulator and Isaac Perceptor. Manipulation has been a fundamental aspect of robotics for decades. Leading the way are giant industrial robotic arms that have become a necessity in car manufacturing. The next generation will be even more dexterous and much more maneuverable. Naturally, Nvidia wants a piece of the action.
“Isaac Manipulator provides cutting-edge dexterity and modular AI capabilities for robotic arms with a robust collection of foundational models and GPU-accelerated libraries,” the company writes. “This speeds up path planning by up to 80 times, increases efficiency and throughput with zero-shot recognition, and enables developers to automate more new robotic tasks.”
Nvidia is already joined by major companies such as Franka Robotics, PickNik Robotics, READY Robotics, Solomon, Universal Robots, and Yaskawa.
AMR (Autonomous Mobile Robotics) is also gaining traction in the form of Perceptor. This program maintains Nvidia's long-standing focus on vision processing in robotics, especially in the form of 'multi-camera robotics'. , 3D surround vision capabilities. ArcBest, BYD, and KION groups have already registered.
The next few years will see a fascinating competition for market share between humanoids and mobile manipulators, and Nvidia wants to be in on every move.