NVIDIA CEO Jensen Huang Releases Open Humanoid Robot Reference Design Based on Isaac GR00T Platform
NVIDIA CEO Jensen Huang announced the release of an open humanoid robot reference design based on the Isaac GR00T platform at GTC Taipei 2026.
The solution utilizes the Yushu Technology H2 Plus humanoid robot body, paired with the Sharpa dexterous hand (31 degrees of freedom), running on the Jetson Thor computer, and integrates NVIDIA's full software stack, including the Cosmos world model, Isaac Lab simulation training, Isaac Teleop remote control, and Isaac ROS.
The goal is to enable universities and research institutions to initiate cutting-edge research within hours, with the first batch of collaborating institutions including Ai2, ETH Zurich, Stanford Robotics Center, and UCSD.
In terms of market mechanisms, the development of humanoid robots is accelerating towards openness, with funding shifting towards standardized reference designs and full-stack software platforms. NVIDIA benefits from its ecological dominance, while traditional closed robot projects face pressure due to high R&D barriers and slower iteration speeds.
Source: Public Information
ABAB AI Insight
NVIDIA has previously established a presence in the field of robot simulation and deployment through the Isaac series. This open reference design continues its path of expanding from GPU computing to embodied intelligence full-stack, significantly reducing training costs through synthetic data generation with Cosmos and enabling edge deployment via Jetson Thor.
In terms of capital strategy, NVIDIA is concentrating resources on hardware reference designs and software toolchains, motivated by the goal of having more academic and research institutions rely on its platform, creating a positive feedback loop from data generation, simulation training to real-world deployment, thus expanding its long-term influence in the humanoid robot field.
Similar cases include NVIDIA's Drive platform strategy in the autonomous driving sector and the early reliance of companies like Figure and 1X on external simulation tools. Currently, humanoid robots are in the early stages of transitioning from high-cost closed R&D to open standardized platforms.
Essentially, this represents a technological shift: humanoid robot development is moving from fragmented hardware customization to open reference designs plus full-stack software platforms. The mechanism is that high R&D barriers hinder the diffusion of innovation, and NVIDIA is lowering entry barriers to concentrate capital and talent resources within its ecosystem, thereby restructuring the toolchain and pricing power distribution in the humanoid robot industry.
ABAB News · Cognitive Law
The more open the platform, the more explosive the innovation.
Hardware is the entry point, but the full-stack software is the real moat.
Excellent companies sell tools, while traditional companies sell individual machines.