In a strategic announcement that caught many by surprise, Nvidia has revealed its inaugural open humanoid robotics platform—and the hardware comes from a Chinese startup that’s largely unknown outside robotics circles.
During the GTC Taipei event on June 1, 2026, Nvidia introduced the Isaac GR00T Reference Humanoid Robot to attendees. The physical platform utilizes the Unitree H2 Plus, a humanoid machine developed by Hangzhou-based Unitree Robotics. This robot measures six feet in height, carries a weight of 150 pounds, and features 31 degrees of freedom. The dexterous hands are supplied by Singapore’s Sharpa, contributing an additional 25 degrees of freedom for precise manipulation tasks. Processing power comes from Nvidia’s Jetson Thor chip equipped with a Blackwell GPU.
During his presentation, Nvidia CEO Jensen Huang highlighted the robot’s specifications—six feet tall and 150 pounds—then quipped, “Just like me.”
The Isaac GR00T platform delivers a comprehensive development pipeline for robotics applications. This encompasses data collection, simulation environments, training protocols, model assessment, and real-world deployment. Scientists and engineers can leverage Isaac Teleop for capturing demonstration data, utilize Isaac Sim along with Isaac Lab for virtual training scenarios, and implement Isaac ROS middleware to transfer trained models onto physical robotic systems.
Nvidia has also made open foundation models available that enable robots to acquire new skills and adjust to unfamiliar settings without requiring manual programming for every individual task. The reference implementation for Unitree’s G1 robot model will be accessible through GitHub and Hugging Face repositories.
Nvidia’s strategic objective is establishing a unified software infrastructure for humanoid robotics—mirroring how its CUDA platform became the industry standard for AI development in previous years.
Unitree’s selection wasn’t arbitrary. The manufacturer has established itself by producing humanoid robots at more competitive price points than Western rivals and has earned recognition for making hardware accessible to research communities.
Securing the role as Nvidia’s inaugural hardware partner positions Unitree alongside prestigious institutions like Stanford University, ETH Zurich, Ai2, and UC San Diego’s Advanced Robotics and Controls Laboratory among the platform’s early adopters.
The announcement’s timing carries particular significance for Unitree’s financial trajectory. The company has submitted documentation for an initial public offering on the Shanghai Stock Exchange’s STAR Market, targeting a capital raise between $610 million and $620 million. Qiming Venture Partners counts among its investors. This high-profile Nvidia partnership delivers substantial validation just as Unitree approaches its market debut.
Nvidia’s strategic initiative places it in an indirect rivalry with Tesla, which is simultaneously advancing its own humanoid robotics program. Tesla has recently halted manufacturing of its Model S and X vehicles, reallocating that factory space in Fremont, California, toward Optimus robot production.
Tesla’s stock price declined 4.6% on the day Nvidia made its announcement. Coincidentally, OpenAI revealed it was recruiting for a robotics division on that same date.
China’s manufacturing dominance in humanoid robotics currently exceeds 80% of worldwide production. Industry analysts anticipate substantial growth in U.S. domestic humanoid robot manufacturing capacity throughout the next twelve months.
While Nvidia’s data center operations continue generating the majority of company revenue, the June 1 announcement reveals where leadership is constructing its next major platform.
The post Nvidia (NVDA) Picks Chinese Startup Over Tesla for Humanoid Robot Partnership appeared first on Blockonomi.
