Beyond humanoid spectacle, China’s two-million-robot factory base reveals a deeper automation advantage
By Shahzaib Saqib I SCN
BEIJING — Humanoid robots that sprint, box and dance have become the most visible symbols of China's race to dominate the next generation of machines, but a quieter transformation already underway inside Chinese factories may prove considerably more important to global manufacturing. China had a record 2.027 million industrial robots operating in factories in 2024, according to the International Federation of Robotics, while 295,000 new units were installed that year — 54% of all industrial robots deployed worldwide. The numbers reveal that Beijing is not beginning its robotics experiment with humanoids; it is attempting to place artificial intelligence and increasingly flexible machines on top of an automation ecosystem already operating at a scale no other country matches.
The contrast has been visible across Beijing this month. At the World Robot Conference and World Humanoid Robot Games, Chinese companies displayed machines capable of running races, playing table tennis and performing increasingly sophisticated human-like movements, while investors poured money into companies including Unitree and Xpeng's robotics division. Yet practical tests showed how far humanoids still have to travel before they can reliably replace workers: robots capable of remarkable athletic performances continued to struggle with seemingly mundane jobs such as recognizing objects, aligning connectors and plugging in cables. Unitree founder Wang Xingxing has similarly cautioned that a breakthrough allowing robots to autonomously perform most unfamiliar tasks could still take years.
China's conventional factory robots tell a very different story because they are already doing commercially useful work. Industrial arms weld car bodies, move components, handle materials and perform repetitive production processes across automotive, electronics, metalworking and other sectors. IFR figures show China accounted for more than half of global robot installations in 2024, and its domestic manufacturers supplied 57% of robots installed in the Chinese market, up from 47% a year earlier. That shift matters strategically: China is not only becoming the world's largest user of industrial automation but increasingly developing the domestic manufacturing capacity to supply the machines automating its own factories.
The scale also challenges an older assumption that China's manufacturing advantage rests principally on abundant inexpensive labour. Robot density — the number of industrial robots per 10,000 manufacturing workers — has risen rapidly as Chinese factories automate production even while the country remains the world's dominant manufacturing base. That transformation gives Beijing a possible answer to one of its biggest long-term economic problems: an ageing population and shrinking workforce need not translate automatically into declining industrial output if factories can produce more with fewer workers. Automation also helps Chinese manufacturers compete in high-volume sectors where consistency, speed and cost increasingly matter as much as labour availability.
That is where the quieter robot revolution connects directly with the much louder humanoid race. China's huge electric-vehicle, battery and electronics industries have already created supply chains for motors, batteries, sensors, actuators, gears and other components that can also be used in advanced robots. Humanoid developers can therefore draw on an existing manufacturing ecosystem rather than constructing one from scratch. The International Federation of Robotics says a central element of China's humanoid strategy is precisely the establishment of a scalable supply chain for key components, while research into China's embodied-AI sector identifies its enormous installed industrial-robot base as an important foundation for the next stage.
That manufacturing advantage is beginning to shape the humanoid competition itself. Xpeng said this week that its robotics unit raised more than $900 million, valuing the business above $6.3 billion, and plans to reach production capacity of 1,000 IRON humanoids per month by the end of 2026 before broader commercial sales in 2027. Chery-affiliated AiMOGA says it has already delivered more than 3,000 robots globally, including about 2,000 overseas across more than 60 countries, while Unitree has become one of the best-known Chinese humanoid manufacturers and exporters. Those companies are trying to convert China's existing advantages in cars, batteries, electronics and precision manufacturing into the ability to mass-produce increasingly intelligent physical machines.
There are substantial reasons for caution. China's humanoid sector is attracting enormous capital before broad commercial demand has been demonstrated, and Chinese officials themselves have warned about excessive duplication among manufacturers. Reuters Breakingviews reported that officials had identified around 150 humanoid manufacturers, while more than 450,000 companies were registered in the wider intelligent-robotics field at the end of 2024. Unitree provides a vivid example of the speculative risk: its shares initially soared 460% on their Shanghai debut but have since fallen sharply, even as questions persist over how quickly spectacular demonstrations can be converted into profitable, reliable machines performing useful work.
The commercial problem is especially important because humanoids must solve something conventional industrial robots largely avoid. A traditional factory robot is engineered for a predictable environment and a narrow task; a general-purpose humanoid is expected to perceive an unpredictable world, understand instructions, manipulate unfamiliar objects and recover when something goes wrong. Recent competitions in Beijing illustrated that gap vividly: running quickly is impressive, but reliably connecting a cable or completing an industrial task without human intervention can be far harder. That is why the next decisive robotics breakthrough may come not from making a machine look more human, but from giving machines enough perception, reasoning and physical-world training data to operate reliably without constant supervision.
China nevertheless has an unusual environment in which to attack that problem. Millions of existing industrial robots, vast manufacturing plants, EV and electronics supply chains, aggressive robotics startups and government support create enormous opportunities to test machines, generate physical-world data and rapidly redesign hardware. Shanghai, for example, has set a goal of deploying 100,000 humanoid robots in factories by the end of the 2026-2030 Five-Year Plan period, while Beijing has elevated embodied intelligence into its broader industrial strategy. Whether those ambitions translate into commercially sustainable deployment remains uncertain, but they demonstrate that China's humanoid programme is being built as an extension of industrial policy rather than simply as a technology showcase.
The global competition therefore may be less about which country produces the first robot capable of an impressive demonstration and more about who can manufacture intelligent machines cheaply enough, deploy them widely enough and improve them quickly enough to change the economics of production. China followed a similar path in electric vehicles: large-scale manufacturing, dense supplier networks, domestic competition and falling component costs eventually produced companies capable of competing globally. Robotics is not guaranteed to follow that trajectory — humanoids remain far less mature commercially than EVs — but China's existing factory automation provides an industrial foundation its humanoid ambitions do not have to build from zero.
For competitors, that is the more consequential development hidden behind China's robot spectacle. The dancing humanoids attract cameras because they resemble the future; the two million industrial robots already working behind factory doors show that part of that future has arrived. If China succeeds in connecting its manufacturing scale with increasingly capable embodied AI, the competition will no longer be simply over who builds the smartest robot. It will be over who controls the factories, components, data and production economics needed to manufacture intelligent machines at global scale.