Last month, robots jumped, boxed and danced their way through the World Humanoid Robot Games in Beijing, with some fleet-footed machines outrunning Usain Bolt’s 100 metre pace. It made for good television. Two days after the games ended, the People’s Liberation Army’s official newspaper, PLA Daily, drew a rather less playful conclusion, calling for researchers to accelerate the transfer of cutting edge robotics technology from laboratories to military training grounds for what it termed robotic combatants. That single editorial line is a useful hinge point for understanding where China’s military robotics programme actually stands.
What the research shows
A Reuters review of more than 100 previously unreported procurement and research records found that China’s defence establishment is accelerating research into humanoid robots’ military applications and actively planning for their eventual wartime deployment. Military institutions are testing humanoids against battlefield requirements, acquiring the robots and training technologies needed, and studying how machines might operate alongside human troops. That work gathered real momentum through 2025 and 2026, with the technical focus centred on perception, manipulation and the training data needed to make robots useful in unpredictable environments rather than choreographed demonstrations.
China’s civilian robotics industry gives the PLA an unusually large base to draw from. Chinese manufacturers accounted for roughly 95 percent of global humanoid robot shipments in 2025, according to BofA Global Research, and the country’s civil-military fusion policy explicitly encourages redirecting commercial manufacturing expertise, supply chains and engineering talent toward military use. That is not a hypothetical concern raised by outside analysts. It is stated government strategy.
From research institutions to defence contractors
By 2026 the effort had moved beyond university labs and into China’s defence industrial base. Norinco, the state-owned defence group, has developed a full-size humanoid called Fuxi, which company material published in August described as suited to sentry duties, all-weather reconnaissance, intelligent patrols and hazardous assignments. Fuxi can also operate through teleoperation, allowing a human controller to direct the robot remotely, which sidesteps the much harder problem of requiring a machine to make every battlefield decision on its own.
Unitree, the commercial firm responsible for most of that 95 percent shipment share, published footage in early September claiming to show fully autonomous humanoid robot combat, driven by what it described as a real-time world model capable of instant planning and dynamic interactive execution. Commercial demonstration footage of this kind should be read with some caution, since a company marketing video is not the same evidence as a fielded military capability, but it illustrates how closely China’s civilian robotics frontier and its military ambitions now sit next to each other.
Quadrupeds, not humanoids, are the part already happening
Set the humanoid headlines aside for a moment, because the more immediately relevant development is not walking, talking robot soldiers. It is armed robot dogs. China has demonstrated drone-deployed quadrupeds, four legged robots dropped from heavy-lift drones that activate on landing and begin patrolling or engaging targets, combining the reach of an aerial drone with the persistence of a ground-based platform. The PLA has run urban warfare exercises featuring robot dog squads advancing alongside infantry, used for reconnaissance, for drawing enemy fire to expose positions, and for carrying explosive charges into breaching operations. Their appeal is partly economic. Individual units reportedly cost under $30,000, cheap enough to field in numbers that can overwhelm defences built to handle far smaller quantities of far more expensive systems.
Money and modest budget lines pointing to something larger
One detail from procurement records is worth flagging on its own. The PLA budgeted approximately $300,000 in June 2026 specifically for camera, radar and motion data collection systems, precisely the components needed to train robots on battlefield perception. That figure is small by military spending standards, which is itself the point. A line item that modest looks far more like one visible piece of a much larger, less visible programme than like the whole of it.
Why the timeline matters more than the demonstrations
It would be a mistake to read any of this as robots replacing soldiers imminently. Malcolm Davis, a senior analyst at the Australian Strategic Policy Institute, has been direct that humanoid robots are not yet practical battlefield systems, while adding that in five to ten years they very well could be. That estimate lines up with the broader assessment across the robotics and defence commentary, which is that fully autonomous humanoid soldiers remain some way off because current systems still lack the general judgement war fighting requires, and that human-machine teaming, with a person supervising or directly controlling the robot, is the realistic near-term model rather than machines operating entirely on their own initiative.
The Ukraine war offers the clearest evidence of what is actually transformative right now, and it is not humanoids. Cheap first-person-view drone swarms have done more to reshape that battlefield than any legged robot, alongside semiautonomous drones using artificial intelligence for navigation and targeting, and simpler robotic platforms used for carrying supplies and retrieving casualties. China’s humanoid push sits downstream of that lesson rather than ahead of it, aimed at filling a gap, ground-level tasks in places drones cannot easily reach, rather than superseding what drone warfare has already proven.
The competitive picture is not one-sided
China’s manufacturing advantage is real, but it is not the whole story. The United States Army has launched its own competing humanoid military programme, though it trails China significantly in manufacturing volume and industrial base. That asymmetry, a Chinese civilian sector producing the overwhelming majority of the world’s humanoid robots against a smaller and more fragmented Western effort, is likely to be one of the more consequential and least discussed aspects of the broader technology competition between the two countries, arguably mattering more in the long run than any single demonstration video.
What this actually amounts to
China is not on the verge of fielding robot armies. It is, however, unmistakably building the institutional, industrial and doctrinal groundwork to do so within roughly a decade, moving with a deliberateness that spans university research programmes, state defence manufacturers and a dominant civilian supply chain that can be redirected on policy instruction. The PLA Daily editorial calling for laboratory technology to reach training grounds was not a throwaway line. It read like an instruction, and the procurement records Reuters reviewed suggest the instruction is already being followed.
Key takeaways
Reuters has found evidence that China’s military is accelerating research into humanoid robots for combat roles, backed by a civilian robotics industry that produced roughly 95 percent of global humanoid shipments in 2025. Norinco’s Fuxi humanoid and PLA-tested armed robot dog quadrupeds represent the more immediate steps, with fully autonomous humanoid soldiers assessed as five to ten years away rather than imminent. Drone swarms, not humanoids, remain the most transformative robotic weapon actually in use, in Ukraine, while China’s programme is aimed at extending robotic warfare into ground environments that drones cannot easily reach.
