NVIDIA Partners with Unitree on H2+ Humanoid Robot Reference Design

NVDA1.56%
MS-0.05%

On June 1, NVIDIA CEO Jensen Huang announced a collaboration with Unitree Robotics to launch the H2+ humanoid robot reference design, also known as the Isaac GR00T system, aimed at accelerating innovation in the global humanoid robotics industry. The partnership combines Unitree's 1.8-meter-tall H2 robot body with NVIDIA's AI computing platform and robotics foundation models. NVIDIA is expanding its presence in robotics development tools, building on its established position in AI computing through the CUDA software platform, as demand grows for humanoid robots and physical AI applications.

Unitree Provides Robot Body, NVIDIA Supplies AI Computing Platform

The system will be available to higher education institutions and university researchers, with customers including Stanford University and ETH Zurich. Unitree Robotics provides the approximately 1.8-meter-tall H2 humanoid robot body, while NVIDIA supplies the AI computing platform, robotics foundation models, and simulation system support. The robot is equipped with five-finger mechanical hands manufactured by Singapore-based Sharpa.

During his presentation, Huang stated that the system has completed full integration. The robot body has 31 degrees of freedom, with each mechanical hand featuring 25 degrees of freedom. The complete system stands approximately 1.8 meters tall and weighs about 68 kilograms. Huang joked that this height and weight are "about the same as mine."

The system's computing power comes from NVIDIA's Jetson AGX Thor T5000 platform. Built on NVIDIA's Blackwell graphics processing unit (GPU), the platform is configured with 128GB of memory and delivers up to 2070 trillion FP4 floating-point operations for artificial intelligence computing. NVIDIA's Isaac GR00T foundation model for humanoid robots provides advanced reasoning capabilities, functioning as the robot's "brain."

NVIDIA Targets Higher Education Institutions with Integrated System

Huang stated that the platform integrates NVIDIA's complete software ecosystem, data generation systems, data simulation systems, and runtime environment, resulting in a humanoid robot that is convenient for researchers to use. He explained that the system primarily targets higher education institutions and university researchers because building a complete humanoid robot testing platform independently is "really too difficult" for them.

In the robotics industry, a reference design serves as a "technical blueprint" that other companies, developers, and research institutions can adopt and customize. NVIDIA continues to release such reference designs to position itself as an indispensable software and hardware supplier in the rapidly growing robotics sector.

Huang also stated: "For agent systems, robotic systems, and physical AI, data is the most difficult problem to solve." He noted that when generating data for AI robots, the data must come from the robot's perspective; however, most video data globally is currently from a third-person perspective rather than a first-person perspective.

NVIDIA Releases Cosmos 3 World Foundation Model

In addition to H2+, NVIDIA released the Cosmos 3 world foundation model. Described as "the frontier of physical AI models," the model can understand the physical world from both third-person and first-person perspectives.

This new system further expands NVIDIA's footprint in robotics software development. Previously, NVIDIA established advantages in AI computing through its widely used CUDA software platform. As humanoid robots and "physical AI" gain momentum, NVIDIA is attempting to extend this advantage into the robotics industry chain.

Huang previously predicted that "physical AI" could develop into a market worth tens of trillions of dollars. He told investors last month that he expects the robotics business to grow rapidly over the next five years.

Morgan Stanley Projects China Humanoid Robot Sales Growth

Unitree Robotics is one of multiple robot manufacturers adopting NVIDIA technology. Besides Unitree, Chinese companies including Zhiyuan Robotics, Galaxy General, and UBTECH have also used NVIDIA's Jetson AGX Thor mass production modules to provide computing power support for their latest robot hardware products.

Morgan Stanley analysts noted in a report on May 14 that China's humanoid robot annual sales are expected to more than double this year to approximately 28,000 units, the highest level globally. The industry is poised to become "the next important industry supporting China's manufacturing and export system."

FAQ

What did NVIDIA and Unitree announce on June 1?

On June 1, NVIDIA CEO Jensen Huang announced a collaboration with Unitree Robotics to launch the H2+ humanoid robot reference design, also called the Isaac GR00T system. The system combines Unitree's 1.8-meter-tall H2 robot body with NVIDIA's Jetson AGX Thor T5000 AI computing platform and Isaac GR00T foundation models, targeting higher education institutions and university researchers.

What are the technical specifications of the H2+ humanoid robot system?

The H2+ system features a robot body with 31 degrees of freedom, standing approximately 1.8 meters tall and weighing about 68 kilograms. Each five-finger mechanical hand has 25 degrees of freedom. The NVIDIA Jetson AGX Thor T5000 platform provides computing power based on the Blackwell GPU, with 128GB memory and up to 2070 trillion FP4 floating-point operations for AI computing.

How many humanoid robots does Morgan Stanley project China will sell this year?

In a report dated May 14, Morgan Stanley analysts projected that China's humanoid robot annual sales will more than double this year to approximately 28,000 units, representing the highest level globally.

Disclaimer: The information on this page may come from third-party sources and is for reference only. It does not represent the views or opinions of Gate and does not constitute any financial, investment, or legal advice. Virtual asset trading involves high risk. Please do not rely solely on the information on this page when making decisions. For details, see the Disclaimer.
Comment
0/400
No comments