Chinaβs humanoid robots are flipping into the spotlight. Literally.
Human-like machines performed kung fu backflips on national television at this year’s Spring Festival Gala, signaling how far the countryβs robotics industry has come. Days later, Chinese smartphone maker Honor said it would unveil its first humanoid robot at Mobile World Congress in Spain β another sign that the sector is moving beyond research labs and into mainstream tech showcases.
Behind the spectacle is a serious industrial push.
Robotics has long been part of Beijingβs βMade in China 2025β blueprint. Originally, it centered on factory automation. But rapid advances in multimodal artificial intelligence are now accelerating what experts call βembodied AIβ β autonomous systems that operate in the physical world. Chinese officials see humanoids as a way to ease labor shortages and boost productivity in a slowing economy.
Speed and scale advantage
At this early stage, Chinese companies appear to be outpacing US rivals in both production speed and shipment volume.
Selina Xu, China and AI policy lead at the office of former Google CEO Eric Schmidt, said Chinaβs edge lies in its hardware ecosystem. The countryβs electric vehicle boom helped build a deep supply chain for sensors, batteries, and precision components β all critical to humanoid development β alongside the worldβs most expansive manufacturing base.
That combination allows Chinese firms to iterate quickly and release new models faster β and often cheaper β than Western competitors.
Xu noted that Unitree, one of Chinaβs leading humanoid startups, shipped roughly 36 times more units last year than US players Figure and Tesla.
Even so, the industry remains nascent. Global humanoid robot shipments totaled just 13,317 units last year, according to a recent Forbes report. Analysts expect annual shipments to nearly double in the coming years, reaching 2.6 million by 2035 β though the report cautioned that it is unclear how many of the current units reflect commercial sales versus demos or pilot deployments.
By projected 2025 shipments, Chinese firms dominate. Agibot and Unitree lead the pack, followed by UBTech, Leju Robotics, Engine AI, and Fourier Intelligence β underscoring Beijingβs early advantage.
From demos to deployment
The bigger shift, industry executives say, is moving beyond viral demonstrations to real-world use.
Yuli Zhao, chief strategy officer at Galbot, said the sector is transitioning from βdemo-driven excitementβ to βoperations-driven adoption.β Galbotβs G1 humanoid appeared at the Spring Festival Gala alongside robots from Unitree Robotics, Noetix, and MagicLab.
βCustomers are asking whether robots can run stably in real environments and actually take work off peopleβs plates,β Zhao said. Chinaβs policy incentives for automation, combined with fast manufacturing cycles, are accelerating practical testing and deployment.
Funding is also flowing. Unitree was valued at around $3 billion after its Series C round last year and is reportedly eyeing a higher valuation in a future IPO. Galbot, meanwhile, has raised more than $300 million in fresh funding, pushing its valuation to about $3 billion β one of the largest financings in Chinaβs humanoid robotics space to date.
US startups are also shifting toward real-world pilots. Foundation, a US humanoid robotics firm, says it aims to build 50,000 units by the end of 2027.
Still, a December TrendForce report suggests China is aggressively targeting both affordable mass-market models and high-end industrial applications, expanding humanoids across manufacturing, consumer, and rehabilitation sectors.
Hardware ahead of software
Despite Chinaβs hardware strength, questions remain about software maturity.
The industry is betting on vision-language-action models and so-called βworld modelsβ that allow robots to understand and anticipate physical environments. But these AI systems remain in early development.
Nvidia currently leads in end-to-end humanoid software stacks, Xu said, and many Chinese startups rely on Nvidiaβs Orin chips. Domestic chipmakers are working on alternatives, but the ecosystem remains heavily tied to US AI hardware.
A key challenge is data. Unlike large language models, which train on vast amounts of internet text, humanoid robots cannot simply scrape the web for physical-world experience. Most companies rely heavily on simulation-generated synthetic data, supplemented by costly real-world testing.
As a result, full autonomy remains far off. Hardware dexterity has improved markedly, but reliability issues persist. The βbodyβ of the robot is advancing faster than its βbrain,β analysts say.
Safety is another looming concern. A high-profile accident could trigger regulatory backlash. As humanoids move from factories to public-facing environments, governments β including Chinaβs β are expected to tighten oversight.
Not just a US-China race
Humanoid robotics is not confined to the worldβs two largest economies.
Japan is aiming for humanoid mass production by 2027, building on decades of robotics innovation from projects like Hondaβs Asimo and SoftBank Roboticsβ Pepper. The country has also pioneered the use of robots in eldercare β a response to its rapidly aging population.
Meanwhile, Hyundai Motorβs Boston Dynamics unit has introduced a new Atlas humanoid designed for factory use by 2028, with plans to produce up to 30,000 units annually in the US.
For now, however, Chinaβs combination of state policy, private capital, industrial depth, and labor pressures appears to be compressing the development cycle.
As Zhao put it: βThe ecosystem here tightens the loop β from R&D and supply chains to manufacturing and deployment. That allows companies to move from prototype to real-world operations and iterate at a pace thatβs difficult to match elsewhere.β
