China’s Robot Army Is Growing — And the Race With America Is Entering a New Phase

China Wants Robots Everywhere — Beijing’s Humanoid Gamble Is Starting to Take Shape

China Puts Thousands of Robots on Display as the Humanoid Race Accelerates

China has transformed Beijing into a vast demonstration ground for the next stage of the robotics revolution. More than 300 exhibitors and thousands of robotic products are filling the 2026 World Robot Conference, with humanoids walking, boxing, dancing, serving customers and attempting the mundane physical tasks that engineers hope will eventually make them useful workers rather than expensive technological spectacles.

But the most important part of the exhibition is not what the robots can do for an audience. It is what China is trying to build around them. Beijing increasingly sees humanoid robotics and "embodied intelligence" — artificial intelligence able to perceive and act in the physical world — as a strategic industry capable of changing manufacturing, logistics, healthcare and eventually everyday life.

Beijing Has Turned Robotics Into an Industrial Mission

The scale of the conference shows how quickly China's robotics ecosystem is expanding. The official programme covers almost 50,000 square metres, with more than 2,000 announced exhibits and over 150 product launches, while demonstrations across the event have pushed the total number of robotic products on display into the thousands. The accompanying World Robot Contest is expected to involve more than 6,000 teams from over 20 countries and regions.

Humanoid robots are drawing much of the attention because they represent something fundamentally different from the industrial machines that transformed factories over previous decades. Traditional robots are normally designed around one controlled task. A genuinely capable humanoid could theoretically use the same doors, tools, workstations, shelves and environments already designed around the human body.

That is the attraction. Instead of rebuilding factories, warehouses and ultimately homes around specialist machines, businesses could deploy machines designed to operate inside a human world.

China has been preparing for exactly that possibility for years. The Ministry of Industry and Information Technology's humanoid robot development strategy called for breakthroughs in core technologies, batch production and internationally competitive companies, with the longer-term ambition of making humanoids an important new economic growth engine.

That policy is now becoming considerably more concrete.

In June, China's industry ministry and state-assets regulator launched a 2026 programme specifically designed to put humanoids and embodied-AI machines into genuine working environments. Manufacturing, maintenance, logistics, retail, healthcare, elderly care, emergency response and disaster relief are among the targeted areas. The programme aims to establish more than 100 high-value application scenarios and create the capacity for deployment at the scale of tens of thousands of units.

That is a crucial distinction. China is no longer merely funding better robots. It is attempting to construct the ecosystem in which those robots learn how to work.

The Robots Are Becoming More Impressive

Some of what visitors are seeing in Beijing would have looked extraordinary only a few years ago.

Humanoid machines can run, maintain their balance after disturbances, manipulate objects, perform coordinated movements and interact with people. Unitree has become particularly effective at producing highly shareable demonstrations involving dancing, fighting and athletic movement, while other Chinese manufacturers are concentrating on industrial manipulation and repetitive production tasks.

China's progress in locomotion has already been demonstrated outside exhibition halls. Humanoid robots have completed half-marathons, with increasingly sophisticated machines showing substantial improvements in endurance, balance and autonomous movement.

Yet walking dramatically across a stage is not the same as reliably performing an economically valuable job for eight hours.

That distinction may become the defining question of the industry.

A human worker can encounter a slightly misplaced object, recognise the problem almost instantly and improvise. A robot may fail because an item is several centimetres outside the location represented heavily in its training data.

Human hands are another extraordinary engineering challenge. Picking up fabric, untangling objects, opening different packages or manipulating oddly shaped items requires an enormous combination of touch, force control, vision and unconscious physical reasoning.

China's own industrial trials demonstrate both the progress and the limitations. Humanoid systems are already being tested on tasks including material loading, logistics sorting, electronics assembly and production-line operations, but developers acknowledge that many machines can perform individual processes without yet being capable of independently managing complete production workflows.

The Factory Floor Is the Real Battlefield

One of the most consequential developments is therefore happening far away from the robot boxing rings.

Chinese manufacturer AgiBot has been testing robots continuously on electronics production lines. Eight robots used in one tablet-manufacturing trial performed material loading, unloading and transfer operations while working with millimetre-level precision. The company has said it expects tens of thousands of units to be shipped during 2026 as deployments expand into electronics, semiconductors, automotive manufacturing, logistics and warehousing.

Other companies are following the same route.

UBTECH has deployed humanoid machines into automotive manufacturing environments, but its own figures illustrate how far the technology still has to travel. The company previously estimated humanoid productivity at only around 30 to 40 per cent of a human worker's output, while targeting substantial improvements by 2027.

Battery life remains another major weakness. Current humanoid robots commonly operate for only a few hours before requiring charging or battery replacement. Dexterous hands, thermal management, reliability and the cost of maintenance remain unresolved problems.

This is why some spectacular demonstrations need to be interpreted carefully.

A robot doing a backflip establishes excellent dynamic control. It does not establish that the same machine can spend thousands of hours assembling products more cheaply than a human without regularly breaking down.

Even at this year's conference, impressive machines have struggled with ordinary activities. One helper robot trying to fold clothing provided an almost perfect demonstration of the industry's paradox: movements that look trivial to humans can still be brutally difficult for machines.

China Has One Enormous Advantage

Whatever the weaknesses of current humanoids, China possesses something few competitors can replicate easily: industrial scale.

The country already operates more factory robots than any other nation. International Federation of Robotics data recorded more than two million industrial robots working in Chinese factories in 2024, while China installed roughly 295,000 new industrial robots that year — approximately 54 per cent of global installations. Chinese manufacturers also overtook foreign suppliers within their domestic market.

That existing ecosystem matters enormously.

Humanoid robots require motors, actuators, sensors, batteries, cameras, precision machining, gearboxes, electronics, AI processors and huge numbers of specialist components. China has spent decades building many of the manufacturing networks capable of producing them.

It creates the possibility of an industrial feedback loop: manufacture more robots, lower component prices, deploy more machines, gather more physical-world training data, improve their AI, generate more demand and manufacture even more robots.

China is already attempting to institutionalise that process.

Robot training centres recreate supermarkets, homes, factories, healthcare environments and workplaces. Humans repeatedly demonstrate tasks to machines while varying objects and environments, creating enormous datasets that can teach physical AI how the real world behaves.

Beijing is even supporting manufacturing infrastructure designed specifically for humanoids. One new "super factory" has initial annual capacity for roughly 10,000 units, with plans described by its operator that could eventually increase production dramatically if demand develops.

China Is Already Producing Robots at Extraordinary Scale

The numbers emerging from China's humanoid sector are increasingly striking.

Chinese authorities said more than 140 domestic humanoid manufacturers released over 330 models during 2025 alone. China has subsequently introduced its first national standard framework covering the humanoid and embodied-intelligence industry, another sign that policymakers are preparing for production at a scale where interoperability, safety and consistent manufacturing become essential.

Industry estimates differ over exactly how market share should be measured, but every major dataset points in the same direction: China has become the dominant source of humanoid hardware.

Some estimates place Chinese manufacturers at more than four-fifths of current global humanoid shipments, while narrower measurements of certain categories place the proportion even higher. The figures should be treated cautiously because the industry remains young and definitions vary substantially, but China's manufacturing advantage is no longer theoretical.

This creates an unusual technological contest with the United States.

America remains extraordinarily powerful in artificial intelligence. Its companies dominate much of the frontier-model ecosystem, advanced computing infrastructure and AI investment landscape. Tesla, Figure, Apptronik and others are pursuing their own humanoid programmes.

China's advantage is different: manufacturing density, component suppliers, aggressive pricing, local government support and access to enormous numbers of physical workplaces in which robots can train.

The humanoid race could therefore become a contest between American strength in machine intelligence and Chinese strength in physically producing and deploying machines.

Why China Wants Humanoid Robots So Badly

There is another reason Beijing is willing to spend heavily.

China is ageing.

Its working-age population is under long-term demographic pressure as decades of declining birth rates reshape the country's labour supply. Automation therefore offers more than a productivity opportunity. It could eventually become part of China's response to having fewer workers available to support an ageing population.

Factories are the obvious starting point because their environments are comparatively structured.

Warehouses could follow. Then commercial cleaning, inspection, hazardous maintenance, hospital logistics, elderly assistance and potentially retail.

Household robots are much harder.

A factory can define exactly where a robot should stand and what object it should handle. A home contains stairs, pets, children, cables, liquid, clutter, unfamiliar objects and countless unpredictable interactions.

The humanoid that reliably loads a factory component may arrive years before the humanoid that can safely clean an entire house and prepare dinner without supervision.

There Is Also a Huge Bubble Risk

China's apparent success carries another danger: government support can create demand before a genuinely sustainable commercial market exists.

Many humanoids are currently being purchased by universities, research institutes, government-supported training facilities and other organisations helping create data and develop the industry. That activity is technologically useful, but it does not prove that ordinary companies are prepared to buy millions of machines because those machines save them money.

This distinction matters because valuations are exploding.

Unitree's dramatic stock-market debut has illustrated the enthusiasm surrounding China's robot industry. Investors are effectively betting that humanoids could become another transformative hardware category, potentially following the trajectory of electric vehicles, smartphones or industrial automation.

The risk is that investors price the destination before engineers have solved the journey.

Thousands of robot companies and suppliers could emerge before consolidation leaves only a handful of winners. China has seen versions of this cycle before in solar power and electric vehicles: massive state-backed expansion, brutal competition, collapsing prices and eventually powerful global manufacturers.

Humanoid robotics could follow the same pattern.

The Coming 'ChatGPT Moment'

The industry's ultimate breakthrough may not come from stronger motors or better hands at all.

It could come from software.

Unitree founder Wang Xingxing has described robotics as approaching its own "ChatGPT moment": the point at which a robot can be placed into an unfamiliar environment, receive an ordinary language instruction and work out how to complete the task without engineers explicitly programming every movement. He has nevertheless cautioned that such capabilities could still be anywhere from roughly two to ten years away.

That is the real goal of embodied intelligence.

Imagine telling a machine: "Unload those boxes, place the damaged ones to the side and put the others onto shelf three."

Today, achieving that reliably can require extensive training, environmental mapping and task-specific programming.

A genuinely general-purpose robot would identify the boxes, understand what constitutes damage, navigate the room, recognise the shelf, manipulate objects of different weights and shapes, recover from mistakes and complete the instruction safely.

Once machines can do that reliably, humanoid robotics stops being primarily a hardware story.

It becomes an AI revolution with arms and legs.

The Race Could Reshape Global Manufacturing

This is why the spectacle in Beijing matters.

Dancing robots and boxing machines generate attention, but China is simultaneously constructing factories, data centres, training facilities, standards, supply chains and real-world deployment programmes around them. Its government wants humanoids to move from exhibition stages into productive economic activity.

Success is not guaranteed. Present machines remain expensive, energy-hungry, comparatively slow and frequently incapable of dealing reliably with the physical unpredictability humans navigate effortlessly every day.

But technological races are not always won by whoever initially produces the smartest prototype. They can be won by whoever learns how to manufacture a technology cheaply, improve it rapidly and deploy it at enormous scale.

China did not invent the electric car, the solar panel or the industrial robot. It nevertheless built formidable manufacturing positions in each.

Humanoid robotics may now be entering the same industrial experiment.

The thousands of machines appearing in Beijing are therefore less important individually than collectively. They show that China is prepared to make an enormous wager: that artificial intelligence will eventually escape the screen, enter the physical economy and become a new class of worker.

If that happens, the country capable of building those workers by the million could gain one of the defining industrial advantages of the 21st century.