China’s artificial intelligence industry is entering a new phase of competition with the United States, as billions of dollars flow into developing robots capable of learning, reasoning, and performing complex physical tasks without extensive human programming.
Chinese companies working on so-called embodied artificial intelligence raised approximately 93.5 billion yuan ($13 billion) across 322 financing deals during the first half of 2026, five times the amount recorded a year earlier, according to investment research firm ITJuzi.
The investment surge reflects a shift in China’s robotics industry, where manufacturers that previously concentrated on developing increasingly sophisticated humanoid machines are now racing to build the intelligence required to make them commercially useful.
A Nov. 1 Reuters investigation found that Chinese robotics companies are redirecting resources toward advanced AI systems capable of learning unfamiliar tasks and adapting to changing environments, opening another front in the technological rivalry between Washington and Beijing.
China’s robotics investment boom shifts toward artificial intelligence
China has spent years expanding its robotics manufacturing capabilities, producing humanoid machines capable of dancing, running, performing martial arts, and executing increasingly sophisticated movements. However, researchers and industry executives say the ability to perform carefully rehearsed demonstrations has not translated into the intelligence needed for widespread industrial deployment.
For the best of our weekly content！
You are now signed up for our newsletter
Check your email to complete sign up
“As competition drives down hardware prices,” Chinese robot manufacturers have “concluded that the margin won’t come from the hardware but from the brain,” Dermot McGrath, founder of Shanghai-based technology consultancy ZenGen Labs, told Reuters.
The investment boom is attracting technology giants, venture capital firms, and government-backed funds.
According to a July Yicai report published on the Shanghai Stock Exchange’s website, financing has increasingly concentrated on developing AI models, collecting large quantities of robot-training data, and creating commercially viable applications. Investors include Alibaba, Tencent, Meituan, and several state-supported investment organizations.
In August, Chinese electric vehicle manufacturer XPeng announced that its robotics business had raised more than $900 million at a valuation exceeding $6.3 billion. The company said the financing would support further development of its humanoid robots and the AI systems controlling them.
China’s advantage is partly rooted in its enormous manufacturing base.
The International Federation of Robotics (IFR) reported on Sept. 24 that Chinese factories installed approximately 354,000 industrial robots in 2025, representing 59 percent of global installations. American factories installed approximately 38,500 units during the same period.
“Industrial automation is progressing at high speed,” IFR President Jane Heffner said in the report.
The organization also reported that Chinese manufacturers supplied 55 percent of their domestic industrial robotics market in 2025, demonstrating the country’s growing manufacturing capabilities.
However, these figures primarily concern conventional industrial robots, which perform repetitive tasks. Developing autonomous humanoid machines that can operate independently in unfamiliar environments presents a substantially more difficult technological challenge.
US and China compete to develop intelligent robotic systems
While China possesses extensive robotics manufacturing capacity, American technology companies remain influential in developing the sophisticated AI models needed to control increasingly autonomous machines.
The emerging technology, often called embodied AI or physical AI, differs from conventional generative artificial intelligence.
Rather than simply producing written responses, images, or computer code, embodied AI systems must interpret visual information, understand physical relationships, and translate instructions into precise movements.
American companies including Physical Intelligence, Skild AI, and Generalist are developing foundation models designed to teach robots new capabilities without requiring extensive programming for each individual task.
In August, California-based Generalist introduced its GEN-1.5 model, reporting that it could acquire basic physical skills after observing demonstrations lasting between three and 12 seconds.
The company said its model could “learn a new task in seconds, from a single example,” without additional training.
In testing involving 10 relatively simple manipulation tasks, Generalist reported an average success rate of approximately 59 percent after a single demonstration. The company acknowledged that its technology remains limited to comparatively short and straightforward activities.
Chinese companies are pursuing similar capabilities.
Beijing-based X Square Robot, backed by investors including Alibaba, ByteDance, Xiaomi, and Meituan, has developed AI systems intended to help robots adapt to unfamiliar household environments.
“The real challenge is not repetition, but whether a robot can execute new, untrained actions in an unstructured environment,” company founder and CEO Qian Wang said in an April announcement.
The company has emphasized the importance of training robots using real-world experiences, allowing machines to encounter unpredictable objects, changing conditions, and obstacles that cannot always be reproduced in laboratory demonstrations.
American semiconductor giant Nvidia is also positioning itself at the center of the emerging industry.
During a March announcement, Nvidia introduced new simulation technologies and AI models designed to accelerate the development of intelligent robots.
“Every industrial company will become a robotics company,” Nvidia CEO Jensen Huang predicted.
Nvidia’s technology is already being incorporated into robotics development platforms used by manufacturers around the world, including companies operating in China.
The broader competition extends beyond robotics hardware to advanced semiconductors, AI training data, and the software infrastructure needed to develop autonomous machines.
In a September analysis, Beniamino Irdi of the German Marshall Fund of the United States identified different strengths in the two countries’ technological strategies.
“The United States leads on the first path, but China is ahead in building the substrate for the second,” Irdi wrote, contrasting American strengths in large language models with China’s manufacturing infrastructure and access to physical-world data.
The analysis suggested that future industrial competitiveness could depend on combining advanced AI capabilities with the manufacturing infrastructure necessary to deploy intelligent machines at scale.
Commercial challenges and technological limitations remain
Despite substantial investment, the practical capabilities of humanoid robots remain considerably more limited than their increasingly sophisticated public demonstrations might suggest.
A comprehensive April assessment by the Mercator Institute for China Studies (MERICS) concluded that “China’s humanoids still lack precision and dexterity.”
The researchers found that many Chinese humanoid robots remain confined to experimental deployments or narrowly defined industrial activities. They also identified significant weaknesses in software development, advanced components, and the ability to operate reliably without human intervention.
The report highlighted China’s continuing dependence on Nvidia technology and specialized mechanical components produced by foreign manufacturers, despite rapid progress in domestic robotics production.
Nevertheless, Chinese companies are beginning to demonstrate practical applications.
According to Reuters, battery manufacturer CATL said robots developed by Spirit AI were successfully connecting flexible wires during battery-pack testing at its Zhongzhou facility, reporting insertion success rates exceeding 99 percent.
Meanwhile, Ant Group’s robotics division, Robbyant, said its machines were processing orders at two pharmacies in Shanghai, although employees continued to verify medications before dispatch.
Such applications suggest that robots may initially become commercially useful in specialized environments rather than immediately replacing human workers across a broad range of occupations.
Financial analysts have also warned that investment enthusiasm could be moving faster than the industry’s commercial development.
“If embodied intelligence remains limited to demonstrations, marketing, and research fields, it will only have short-term commercial value,” Liu Yinghang, a partner at investment advisory firm Scale Partners, told Yicai in July.
Liu cautioned that rapidly rising valuations could create speculative risks if manufacturers fail to establish sustainable business models.
Similar concerns were raised in an August Reuters investigation, which documented the industry’s dependence on government support, limited commercial demand, and continued difficulty performing basic physical tasks efficiently.
The technological obstacles are particularly significant because robots must operate reliably outside controlled environments. Changes in lighting, object placement, physical resistance, or workplace conditions can cause failures that would rarely trouble a human employee.
Industry consultant Georg Stieler cautioned that even major advances in robotic AI would not necessarily result in immediate commercial readiness.
Asked about the time separating impressive demonstrations from reliable workplace deployment, Stieler told Reuters that he would “add years, not months.”
For both Washington and Beijing, the central challenge is therefore shifting from building increasingly sophisticated machines to developing artificial intelligence capable of making those machines consistently productive.
China’s manufacturing scale and rapidly expanding investment provide substantial resources for that effort, while American companies retain important advantages in advanced AI research and computing technology.
Original article
Comments