Nvidia, Unitree & Sharpa unveil H2+ humanoid robot reference design for real-world work
Nvidia's Isaac GR00T brain + Unitree's H2 body + Sharpa's hands = faster humanoid development.
At Computex 2026, Nvidia CEO Jensen Huang announced a three-way partnership with Chinese robotics leader Unitree Robotics and Singapore's Sharpa to unveil the H2+, also branded as Isaac GR00T, a comprehensive humanoid robot reference design. The design integrates Unitree's human-sized H2 robot body with Sharpa's flagship Wave five-fingered robotic hands, while Nvidia's Isaac GR00T foundational models provide the advanced reasoning capabilities—effectively acting as the robot's 'brain'. The reference design is intended to serve as an open blueprint that researchers and companies can adopt and customize to build, fine-tune, and deploy skills faster, addressing the major bottleneck of data scarcity in physical AI.
Huang emphasized that for agentic systems, robotics, and physical AI, data remains the hardest problem. The H2+ reference design aims to solve this by streamlining the full development workflow: from automated data collection and policy training to safe real-world deployment. By offering a standardized yet customizable platform, Nvidia positions itself as an indispensable software and hardware supplier in the rapidly growing humanoid robotics industry. The collaboration marks a significant step toward creating humanoid robots capable of performing 'real work' in factories, logistics, and other commercial environments, with the reference design enabling faster iteration and lower development costs for the global robotics community.
- H2+ combines Unitree's H2 humanoid body, Sharpa's Wave five-fingered hands, and Nvidia's Isaac GR00T models as the robot's brain.
- The reference design streamlines the entire development pipeline—data collection, policy training, and deployment—to accelerate humanoid robotics research.
- Announced by Jensen Huang at Computex in Taipei, the design aims to solve the 'data is the hardest problem' in physical AI and agentic systems.
Why It Matters
This collaboration gives robotics researchers a standardized, modular blueprint to build and deploy humanoid robots faster, accelerating the path to real-world commercial applications.