China’s PLA Is Building the Industrial Base for Military Humanoid Robots

That distinction matters because China already dominates the commercial humanoid market. Bank of America Global Research estimates that Chinese manufacturers accounted for about 95 per cent of global humanoid robot shipments in 2025. The PLA is now building channels through which that industrial advantage can feed military development - through perception systems, training data, teleoperation and battlefield simulations.

The result could be a very different model of military robotics from the one currently taking shape in Europe.

The battlefield is being designed in data first

The clearest signal from the PLA's activity is that it is not simply buying robots. It is trying to teach them how to function in environments that are difficult for machines to understand.

In June 2026, the PLA allocated about $300,000 for a specialised system to collect and label camera, radar and spatial-motion data intended to improve humanoid navigation across complex terrain. An earlier procurement linked to the National University of Defense Technology (NUDT) sought an “embodied humanoid robot intelligent perception and dexterous operation system”.

Those projects point towards a familiar lesson from artificial intelligence: hardware is only part of the problem. A machine that can walk through a room is not necessarily capable of understanding what it sees, deciding how to move around obstacles, manipulating objects, or responding to an unpredictable human environment.

The PLA appears to be treating the collection of that experience as a military capability in its own right.

NUDT's August 2025 simulation offers a glimpse of the intended application. Researchers modelled an urban assault involving six robotic platforms - humanoids, quadrupedal robot dogs and unmanned vehicles - operating in two teams alongside human soldiers to clear a multi-storey building room by room.

That is a more significant development than the spectacle of a humanoid robot performing a demonstration. It places the technology inside a tactical formation.

The question is therefore less whether a humanoid can walk reliably today than whether China can accumulate enough data, testing and industrial experience to make such systems increasingly useful over the next decade.

Why bother with two legs?

Wheeled and quadrupedal machines are generally easier to stabilise and can already perform useful military tasks. Europe is consequently concentrating on unmanned ground vehicles (UGVs) for logistics, casualty evacuation and mine clearance, rather than betting heavily on bipedal platforms.

The PLA's interest in humanoids reflects a different calculation.

Military operations in cities take place inside infrastructure built for humans. Doors, staircases, ladders, corridors and vehicles all assume a human body. A machine with a broadly human physical configuration could operate in that environment without requiring the battlefield to be redesigned around it.

That makes the humanoid form potentially useful for precisely the tasks where sending people remains most dangerous: entering buildings, moving through confined spaces and carrying out high-risk physical work.

But the potential is still some distance from reality.

Humanoid machines consume substantial energy, remain vulnerable outside controlled environments and face difficult problems of balance, durability and embodied artificial intelligence. Military analysts and Chinese defence commentary cited by Reuters put practical battlefield deployment, including humanoid assault teams, roughly five to 10 years away.

That gap is precisely why the current investment matters.

If China waited until the machines were combat-ready before developing the doctrine and training infrastructure around them, it would lose valuable time. Building those systems now allows the PLA to work on the slower problem: teaching machines how to operate as part of a military force.

Teleoperation may be the bridge

China's defence industry is also signalling that full autonomy is not the immediate objective.

In August 2026, state-owned defence company Norinco unveiled Fuxi, a full-size humanoid designed for sentry duties, all-weather reconnaissance and high-risk operations. The system was paired with teleoperation technology, allowing a human operator to control the machine remotely.

That is an important compromise. Current artificial intelligence may struggle to make reliable decisions in an unpredictable battlefield, but a remotely controlled machine can still place hardware where sending a soldier would be costly.

Teleoperation also offers a way to accumulate operational experience before autonomy becomes sufficiently reliable. Humans remain in the decision loop while machines handle the physical exposure.

China's Ministry of Defense said in March 2026 that AI-powered weapons must remain under human control. The United States has likewise maintained requirements for legal review, operational testing, and human judgment over lethal force.

Yet the boundary becomes harder to define as machines acquire greater ability to navigate and act independently.

The International Committee of the Red Cross has warned that fully autonomous systems lack the contextual flexibility required to assess intent, recognise surrender and reliably distinguish combatants from civilians in dense urban environments. Humanoids operating alongside human troops could make that problem particularly acute, because they would be entering precisely the environments where those distinctions are most difficult.

China's industrial advantage is the real weapon

The strongest argument for taking China's military humanoid programme seriously is not that Chinese robots are already superior battlefield machines. It is that China has an industrial ecosystem capable of producing them at scale.

The 95 per cent share of global humanoid shipments gives Chinese manufacturers something military programmes traditionally struggle to obtain: volume.

A defence laboratory can develop a sophisticated prototype. An industrial ecosystem can produce thousands of machines, gather performance data from them, refine components and repeat the process. If commercial and military development are allowed to feed into each other, the boundary between the two becomes strategically important.

China's robotics industry is already being extended into Europe's immediate neighbourhood. In August 2026, a joint Chinese-Serbian factory operated by Minth and AGIBOT opened in Šabac, with plans to produce more than 5,000 humanoid robots annually and scale production to 20,000 units when robot dogs are included.

Serbia is not an EU member, but its location gives the development a significance beyond the factory itself. Chinese humanoid manufacturing is gaining an industrial foothold on Europe's periphery at precisely the moment European governments are trying to work out what physical AI means for defence.

The United States is also pursuing militarised humanoid technology. The US Army launched a competition in 2025 covering reconnaissance, obstacle clearance and urban combat, with up to $1.25mn in follow-on contracts. But China's manufacturing base gives Beijing a different starting position.

The contest is therefore not simply between individual robots. It is between industrial systems.

Europe is learning a different lesson from Ukraine

Europe's response has so far been more pragmatic.

Combat experience in Ukraine has accelerated the development of non-humanoid ground systems. Ukrainian forces conducted more than 112,000 UGV missions in 2026, providing a body of battlefield experience that European manufacturers are using to develop and scale practical robotic platforms.

European-Ukrainian projects include Quantum Tencore Industries in Germany, ARX Robotics' work with Roboneers, serial production initiatives involving Poland and Finland, and a Denmark-Ukraine joint venture producing the Ravlyk UGV.

Ukraine also formally integrated drones, ground robots and infantry into unified drone-assault units in April 2026.

This is a fundamentally different route into robotic warfare. Europe is taking machines that already solve particular military problems and pushing them towards mass deployment. China is investing more heavily in the longer-term proposition of machines capable of interacting with the physical world in something closer to a human way.

Neither approach has yet settled the future of warfare.

But the difference could become important if humanoid robotics crosses its current technological threshold. Europe's advantage lies in combat feedback from systems already being used. China's lies in combining that kind of military ambition with an enormous commercial manufacturing base.

For now, the PLA does not appear to have armed humanoids operating in its frontline units. The batteries are not good enough, the machines are not robust enough, and autonomous reasoning remains too unreliable.

The strategic competition is taking place before those problems are solved.