The Collaborative Future of Humanoid Robotics: Chinese Hardware, American AI Silicon

The next generation of humanoid robots is a cross-border build: its physical body hails from China, while its artificial brain runs on cutting-edge American silicon.

This week, Nvidia CEO Jensen Huang unveiled a new reference blueprint for a fully functional humanoid robot that integrates components from global innovators across the industry: the 6-foot, 150-pound H2 Plus base robot from Unitree, a fast-growing Chinese robotics startup; Nvidia’s high-performance Thor T5000 AI chip; an advanced dexterous humanoid hand; and a new suite of developer tools that simplify programming and AI training for the machine. The pre-integrated design makes it far easier for research teams, including U.S. academic labs, to assemble cutting-edge humanoid platforms and test their own custom AI algorithms.

Nvidia’s Thor chip powers the large AI models that let the robot interpret its environment and control its complex movements, while Unitree supplies the robot’s core hardware: its proprietary motors, actuators, and sensors. The flexible, human-like hand from Singaporean robotics firm Sharpa can pull off delicate tasks ranging from card tricks to peeling an apple — a notable milestone, as fine motor dexterity remains one of the most challenging unsolved problems in humanoid robotics today.

Spencer Huang, Nvidia’s director of product for robotics, told WIRED the company aims to provide its AI silicon expertise to as many humanoid robotics developers as possible. “Unitree is the first, but they're not going to be the last by a long shot,” said Huang (who is also Jensen Huang’s son). He added that the technology powering the H2 platform could also boost capabilities for a wide range of other robotic systems, including traditional industrial robotic arms.

In some respects, the partnership is surprising. Robotics has emerged as a critical new front in U.S.-China tech competition, with some U.S. politicians proposing full bans on Chinese-built humanoid robots. Last year, security researchers claimed Unitree’s robots are capable of capturing and transmitting user data off-platform, sparking widespread national security concerns.

But in other ways, the collaboration makes perfect sense, according to policy experts. “This is a fascinating development,” says Scott Singer, a fellow at the Carnegie Endowment for International Peace who studies AI governance and China. Singer notes that while the U.S. dominates the global market for the most advanced AI chips, Chinese robotics firms hold a major hardware advantage thanks to China’s mature, cost-effective manufacturing supply chain. “Both sides have key parts of the supply chain that they might be able to weaponize, but here they are working together,” he says.

Nvidia, for its part, has explicitly addressed existing security concerns. Alongside its upgraded AI brain and dexterous gripper, the new H2 Plus blueprint includes built-in security features designed to reassure users that their data and proprietary AI models remain secure.

Nvidia’s chips are currently the global gold standard for training large AI models, and the company has invested heavily to expand into advanced robotics, developing purpose-built hardware and software tools for the industry. While U.S. export rules bar Nvidia from selling its most capable chips to China, the government loosened restrictions late last year to allow the sale of a wider range of advanced chips to Chinese buyers.

Singer says robotics and AI are widely viewed as foundational to future manufacturing growth, economic productivity, next-generation military capabilities, and broader advances in AI research overall. He argues it will be critical for the U.S. to develop a clear strategy to nurture its own domestic robotics industry — a goal that may well require finding pathways to collaborate with Chinese manufacturers.

Unitree’s robots are already extremely popular among developers in China and across the globe, thanks to their accessible pricing and simple programmability. A base model of Unitree’s G1 humanoid costs roughly $15,000, while competing humanoid platforms often run to several hundred thousand dollars apiece. Viral social media videos regularly show Unitree’s robots pulling off parkour routines, kung fu forms, and other acrobatic feats, and the company’s hardware is already used for research published by dozens of Western academic labs.

Not all industry stakeholders are enthusiastic about the rapid growth of Chinese robot makers, however. Gavin Kenneally, co-founder and CEO of Ghost Robotics, which builds legged robots for defense and security use, argues that Unitree’s technology has drawn extensively on original innovations from Western academic research labs. He adds that it is critical for the U.S. to prevent China from gaining dominance over the humanoid robot market or the broader global robotics industry.

“Without a serious near-term policy response, including a national robotics strategy, the US risks ceding the commercial robotics market to Unitree and other Chinese companies, as we've already seen happen in the drone space with DJI,” he says.

But Nvidia’s CEO sees major upside in partnering with Chinese robot developers. “Humanoid robots will bring physical AI to the world’s largest industries, opening a multitrillion-dollar economic opportunity,” Huang said.

---

Update 6/04/26 1:40pm ET: This story has been updated to clarify how the new H2 technology could be adapted to other robot systems.

This is an edition of Will Knight’s AI Lab newsletter. Read previous editions here. (link placeholder per original text)