China leads global humanoid robot race with rapid production

China is rapidly asserting dominance in the emerging humanoid robotics sector, leveraging its vast manufacturing infrastructure to outpace international competitors. While U.S. firms focus on long-term development, Chinese manufacturers are already transitioning from flashy demonstrations to real-world deployment, driven by a unique convergence of supply chain maturity and aggressive state-backed industrial strategy.

The manufacturing advantage

Selina Xu, a China and AI policy lead at the office of Eric Schmidt, notes that Chinese companies are currently moving faster and shipping in higher volumes than their counterparts in the United States. This efficiency stems from a deeply integrated hardware supply chain—much of which was originally refined for the electric vehicle (EV) industry—covering everything from essential sensors to high-capacity batteries.

The scale of this output is significant. According to industry data, the leading Chinese firm Unitree shipped approximately 36 times more units last year than Tesla and Figure combined. This capability allows Chinese startups to iterate on hardware designs at a pace that Western competitors currently struggle to match.

Key players currently driving Beijing’s lead include:

• Agibot
• Unitree
• UBTech
• Leju Robotics
• Engine AI
• Fourier Intelligence

Shift toward practical application

The industry is moving beyond the “demo-driven” phase toward “operations-driven adoption.” Yuli Zhao, chief strategy officer at Galbot, emphasizes that the primary question from customers has shifted: they now demand proof that robots can operate stably in real-world environments to alleviate labor shortages.

This practical demand is supported by China’s broader “Made in China 2025” plan, which encourages automation upgrades. Consequently, startups are securing massive capital injections to fuel this transition. Galbot recently raised over $300 million, reaching a valuation of $3 billion, while Unitree is eyeing a potential IPO with valuations as high as $7 billion.

Software and data bottlenecks

Despite the hardware lead, the “brain” of these machines remains a work in progress. Most Chinese startups still rely on Nvidia’s Orin chips, though domestic alternatives are under development. A core technical hurdle is the scarcity of real-world training data; unlike Large Language Models (LLMs) that can scrape the internet, humanoid robots require physical state data from unpredictable environments.

Current limitations include:

• Data Scarcity: Difficulty in creating autonomous models without massive real-world datasets.
• Hardware vs. Software: Physical dexterity has improved, but internal software logic is still nascent.
• Safety Concerns: The industry must navigate potential public backlash from accidents, which may lead to tighter future regulations.

Global competition

The race is not confined to China and the U.S. Japan, a long-time pioneer in robotics, is targeting mass production by 2027, with a unique focus on eldercare solutions. Meanwhile, Hyundai’s Boston Dynamics unit is preparing its updated Atlas model for factory use by 2028, with plans to produce up to 30,000 units annually.

However, China’s ability to compress the entire development cycle—from R&D to customer deployment—into a tight loop remains its most significant competitive edge. As Zhao noted, this ecosystem allows for rapid learning and iteration, making it an formidable force in the global robotics landscape.