Nvidia Unveils 6-Foot Humanoid Blueprint With Unitree as US-China Robotics Tensions Persist
Updated
Updated · WIRED · Jun 3
Nvidia Unveils 6-Foot Humanoid Blueprint With Unitree as US-China Robotics Tensions Persist
3 articles · Updated · WIRED · Jun 3
Summary
Jensen Huang presented a humanoid reference design built around Unitree’s 6-foot, 150-pound H2 Plus body, Nvidia’s Thor T5000 chip, a Sharpa hand and new training software.
The package is aimed at researchers and robot makers that want a faster way to build and train advanced humanoids, with Nvidia positioning its chips and software as the brains for multiple robot vendors.
Security concerns shadow the partnership: Unitree faced scrutiny last year over data-transmission risks, and Nvidia says the blueprint includes protections for users’ data and AI models.
The tie-up bridges a split supply chain in which the US leads in AI chips while China has a hardware-cost edge; Unitree’s G1 starts around $15,000, far below many rivals priced in the hundreds of thousands.
The launch lands amid export controls and calls in Washington for a stronger domestic robotics strategy, even as Huang argues humanoids could unlock a multitrillion-dollar industrial market.
With US AI powering Chinese robots, can America's new national strategy prevent another DJI-style market loss?
As humanoid robots become affordable, are we prepared for the massive societal and labor market shifts that will follow?
Humanoid Robotics 2026: Nvidia, China’s Surge, and the Geopolitical Race for 2.6 Million Robots by 2035
Overview
NVIDIA and Unitree Robotics have joined forces to launch the Isaac GR00T platform, combining Unitree’s robotics expertise with NVIDIA’s advanced computing power. This partnership has produced an integrated humanoid robot design that unifies hardware and software, aiming to accelerate development in the field. By leveraging Unitree’s proven track record in both commercial and academic robotics, and NVIDIA’s powerful onboard computing, the collaboration creates a flexible, modular environment for research teams. This unified approach marks a significant step forward, making advanced humanoid robotics more accessible and efficient for developers worldwide.