China Is Testing Humanoid Robots for Future Military Use—But They Aren’t Battlefield Weapons Yet

China is building the foundations for possible military use of humanoid robots by combining its large commercial robotics industry with People’s Liberation Army research, procurement, training systems and operational experiments. The available evidence shows an accelerated development effort—not a fielded force of armed humanoid robots. Reuters found no evidence that an armed humanoid has entered service with an operational PLA unit. 1

What Reuters found in Chinese military records

Reuters reviewed more than 100 Chinese military procurement notices, academic studies, patents, official publications, government records and defense-company materials. Its reporting found that activity related to military humanoids accelerated in 2025 and 2026. PLA-linked institutions were testing robots against military requirements, acquiring robots and training technologies, and examining how robots might operate alongside troops. 1

The work is focused on the less glamorous but essential building blocks of useful field robotics:

• Perception: enabling a robot to interpret surroundings and operate around personnel.
• Dexterous manipulation: improving the ability to handle objects and carry out physical tasks.
• Training data: collecting and labeling camera, radar, movement and terrain data, including information on whether terrain is traversable.
• Human-robot teaming: assessing how machines could support troops rather than simply perform isolated demonstrations. 1

Reuters identified a May 2025 PLA tender for a humanoid robot, installation and technical training, as well as a September 2025 notice linked to the National University of Defense Technology (NUDT) for an embodied humanoid intelligent-perception and dexterous-operation system. A June 2026 PLA budget of about $300,000 covered a system for collecting and labeling sensor and movement data. Reuters could not determine whether earlier tenders had been completed or which suppliers received them. 1

From robot exhibitions to military training concepts

The World Humanoid Robot Games took place in August 2026, not 2025. The event displayed gains in mobility and other capabilities, while also revealing conspicuous failures—an important reminder that impressive demonstrations do not equal battlefield reliability. 1

Two days later, the PLA Daily called for cutting-edge technology to move from laboratories into military training grounds for robotic “combatants.” That intervention gave public momentum to the idea of translating civilian humanoid progress into military experimentation. 1

Earlier signs of institutional interest included a 2024 forum session involving NUDT and the Academy of Military Sciences on military uses of humanoid systems. In 2025, an NUDT competition simulated missions including behind-enemy-lines infiltration, mapping and target location, while also considering base security, patrol and inspection roles. 1

Why China’s commercial robot industry matters

China’s manufacturers accounted for roughly 95% of global humanoid-robot shipments in 2025, according to BofA Global Research, as cited by Reuters. That commercial lead could give the PLA a broad domestic pool of platforms, components, software and development expertise. 1

The practical advantage is not that commercial humanoids are automatically military-ready. Rather, a large civilian market can support faster iteration, access to parts and potentially lower-cost scaling. It may also expand the data and engineering base available to military researchers.

That advantage has limits. A factory robot or exhibition model must still withstand uneven ground, poor weather, communications disruption, physical damage, limited battery power and unpredictable human behavior before it can be considered credible for combat operations. 1

The battlefield roles China is exploring

The most detailed reported concept is urban combat. An August 2025 NUDT paper modeled six “combat robots”—a mix of humanoids, robot dogs and unmanned vehicles—divided into two assault teams supporting ground troops as they cleared an enemy-held building room by room and floor by floor. The paper projected that relevant equipment could be available within five to 10 years. It did not spell out armament, rules of engagement or how civilians would be handled. 1

Other prospective roles described in PLA-linked material include:

• reconnaissance and sentry duty;
• security patrols and inspections;
• hazardous or dangerous missions;
• logistics and obstacle clearing;
• tasks in buildings, tunnels and ship interiors; and
• possible urban offensive and defensive operations. 1

For many of these missions, a humanoid form is not necessarily the most practical choice. Wheeled, tracked and quadruped robots can be simpler and less expensive to deploy for routine reconnaissance, transport or surveillance. Humanoids become more attractive where a human-like body could help navigate stairs, doors, rooms and environments designed around people—but that potential remains technically difficult to realize. 1

Norinco’s Fuxi claims and the role of teleoperation

China’s state-owned defense company Norinco says its full-size Fuxi humanoid can conduct sentry duty, all-weather reconnaissance, intelligent patrols and dangerous missions. Norinco’s materials also describe a teleoperation system that would allow a human operator to control the robot remotely. 1

Teleoperation is significant because it could keep humans involved where autonomous judgment is unreliable or inappropriate. But these are company claims rather than independently verified evidence of operational performance in combat conditions. 1

The biggest obstacles: power, reliability and judgment

Humanoid robots remain energy-intensive and unreliable outside controlled settings. Battlefield conditions compound the challenge: robots must retain balance, perceive changing surroundings and make sound decisions amid damaged infrastructure, difficult terrain, weather, interference and human unpredictability. 1

Chinese military commentary itself has identified unresolved problems in movement, perception and intelligence, while noting that further breakthroughs, lower costs and mass production would be needed for widespread use. 1

The evidence therefore supports a more restrained conclusion than visions of robot infantry: China is developing and testing the ecosystem for military humanoids, but it has not demonstrated that it can field capable armed humanoids at scale.

How the U.S. effort compares

The United States is also exploring military humanoids. In 2025, the U.S. Army launched a competition for “militarized humanoid capabilities,” considering uses including reconnaissance, security, obstacle clearing, hazardous-material tasks and urban offensive or defensive missions. The program could select up to 10 finalists for testing with military experts and offered up to $1.25 million for follow-on contracts. 1

Reuters assessed that the U.S. robotics industry trails China in two- and four-legged platforms, although U.S. military-funded innovation has influenced some leading Chinese robot-dog designs. 1

China’s edge is best understood as an industrial and supply-chain advantage, not proof of superior battlefield autonomy. The race depends on far more than manufacturing volume: robust mobility, perception, communications, power systems, human-machine command structures and operational testing all matter.

Human control is the decisive ethical question

The hardest issue is not simply whether a robot can move through a building. It is whether a machine can be trusted with decisions involving lethal force.

International humanitarian law requires parties to distinguish combatants from civilians and to protect wounded or surrendering fighters. The International Committee of the Red Cross has warned that autonomous systems may struggle to recognize surrender because it depends on context, behavior and intent. 1

China’s defense ministry said in March that AI weapons should remain under human control. A July 2025 PLA Daily article also discussed human verification before a robot could fire on a living target. 1

Those statements point toward a likely initial model: robots conducting surveillance, transport, patrol, reconnaissance or high-risk entry under close human supervision, rather than independently selecting and engaging targets.

What to expect over the next decade

Malcolm Davis of the Australian Strategic Policy Institute described humanoids as currently impractical battlefield systems, while judging that they could become viable within five to 10 years. 1

That timeline aligns with the NUDT urban-assault scenario, but it should be read as an aspiration rather than a deployment forecast. The strongest evidence today is that China is investing in the research, procurement and data infrastructure needed to experiment with military humanoids. The most plausible near-term applications are remotely supervised reconnaissance, patrol, logistics and dangerous-entry tasks—not autonomous armed formations. 1