Nvidia humanoid training systems draw on China-linked robot hand suppliers

Traders Union
All News
Business & Wealth
Companies
Nvidia uses China robot hands

Highlights

Nvidia unveiled two humanoid robot training systems in 2024 using mechanical hands from Sharpa and Wuji, both with Shanghai-linked founders and production.
Agile Robots' Agile One SN system featuring Nvidia's Isaac software and Jetson Thor platform is slated to reach labs like Stanford and ETH Zurich by early 2027.
China's influence in advanced robot hand manufacturing is growing, with startups Wuji and Sharpa gaining international recognition for their roles in key Nvidia and Meta projects.

Nvidia’s latest humanoid platforms and suppliers

As reported by CNBC, Nvidia has showcased two humanoid robot training packages within a few months, one using mechanical hands from Sharpa and another using hands from Wuji. Both companies have ties to founders in Shanghai, underscoring China’s role in a key component of humanoid hardware development.

Pan Yunzhe founded Wuji in 2019 before moving much of the team to Shenzhen, and the startup has attracted backing from Ant Group, HSG and Future Capital, according to PitchBook. A demo video released last week says the Wuji Hand 2 can autonomously spin a pen despite interference.

The Wuji Hand 2 appeared on a slide on Oct. 21 behind Nvidia Chief Executive Jensen Huang as he presented in Berlin a new humanoid hardware system for training robots on Nvidia software. The robot body is supplied by Agile Robots, which has offices in Germany, China and the U.S.

Huang says the hardware system, called Agile One SN, will help accelerate robotics development. Agile Robots has announced that the system uses Nvidia’s Isaac robotics software and Jetson Thor computing platform.

According to Nvidia, research institutions including Stanford and ETH Zurich can likely start using Agile One SN in early 2027. Those two institutions were also among the groups Nvidia named when it announced its first reference humanoid design on May 31, a system using a body from Chinese humanoid company Unitree and mechanical hands from Sharpa, with deliveries scheduled from late 2026.

Nvidia did not immediately respond to requests for comment on whether deliveries have begun or whether researchers will test both reference robots or only one. Sharpa, which is headquartered in Singapore, was co-founded by David Li, chief executive of Shanghai-based lidar company Hesai, and the company has said its manufacturing research and development is in Shanghai.

China’s robot hand makers gain influence

Mechanical hands are becoming an increasingly important point of differentiation as humanoid developers try to overcome hardware limits. Their rising profile is evident enough that Google DeepMind research lead Keerthana Gopalakrishnan directly compared Sharpa and Wuji in a podcast last month.

She says Sharpa Wave is somewhat stronger than Wuji’s hand, while Wuji’s design is closer in size to a typical human hand but has the strength of a ten-year-old. It is not clear whether she was referring to Wuji Hand 2 or an earlier version.

In September, JoyIn founder Zhu Mingxuan says Wuji’s mechanical hands are very light and effective at completing tasks, although they can heat up in operation. She says Sharpa’s hands perform better but are far more expensive.

Zhu says hardware is constraining humanoid development and that hands remain the main challenge. She previously left U.S. humanoid company Figure last year, where she worked on models intended to help humanoid robots mimic human actions.

As China pushes its humanoid sector to develop more core technologies, robot hands are emerging as one area where suppliers can stand out. Xiaomi said last week that tests of its robotic hand in its car factory showed most tasks required only a specific range of movement.

Another sign of China’s progress in the field comes from Meta’s work with Chinese startup Fourier on a humanoid prototype, The Information reported on Oct. 31, citing sources. Neither company immediately responded to requests for comment.

In our earlier report on the growing AI safety debate, we highlighted Nvidia CEO Jensen Huang’s view that cooperation with Chinese AI companies is necessary to help monitor potentially dangerous uses as regulation tightens. We also noted discussions around creating U.S.-China communication channels for serious AI incidents and Huang’s suggestion that the UK could act as a neutral venue for broader international coordination on AI risks.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.