The Military Humanoid Race: How the US and China Are Automating the Battlefield

Out of 103 initial pitches thrown into the ring, the US Army picked five winning entries for its first-ever xTechHumanoid run:

• Florida Institute for Human and Machine Cognition (IHMC): Alex, a heavy-set frame with 29 joints, custom impact-resilient actuators, protective shielding, and a split control setup—half neural network, half human operator holding the reins via direct teleoperation.

• Hydroplane Ltd.: A compact mobile power generator made for off-grid field spots where running a power cable is entirely out of the question.

• ACES Group, LLC: Modular power architectures and energy storage built to feed everything from forward logistics hubs to onboard battery units without breaking down.

• RAS Labs, Inc.: High-sensitivity tactile skin sensors for fingers, capable of reading force, surface shape, and microscopic slip as it happens.

• Velocity Explorations: Interface code that translates raw military comms into actual physical movement commands on the fly.

The Quantico trials mark a proper structural shift in how defense planners buy hardware. Rather than dropping a massive contract on a single proprietary machine from one big prime contractor, the Pentagon chose one heavy-duty chassis alongside four specialized subsystem tech builds.

‘Humanoid robots are part of a broader autonomy push to increase the depth and breadth of robotic platforms and components industry can provide for a range of missions,’ said US Army Chief Technology Officer Dr Alex Miller. ‘We are identifying the problems, mission needs, and opportunities to enable our formations with the right mix of tools. xTech is our method of surveying the industry landscape and fast-tracking promising technology rather than falling victim to years-long studies.’

The primary operational target is as straightforward as it gets. You swap flesh and blood for synthetic frames in high-threat tasks—explosive ordnance disposal, forward ammo runs, and heavy combat engineering support.

Built for Polish, Dropped in Dirt

Let's be completely realistic about Silicon Valley's output for a second. Modern humanoids from the likes of Tesla, Figure, or Agility Robotics were drawn up for climate-controlled fulfilment hubs. They like flat polished concrete, predictable overhead lighting, and a charging dock right down the hall.

Warzones don't work like that.

Field logistics demand a machine that can scramble across broken rubble, haul itself up steep naval ladders, and handle dangerous kit in driving rain. Off-the-shelf plastic casings shatter on first impact. Standard electric actuators freeze up under shock loads.

That explains why IHMC’s Alex was the only full platform to make the final cut at Quantico. Pairing 29 joints with impact-resistant housings, automated balancing, and direct human control isn't some luxury add-on—it's the bare minimum for surviving five minutes near a field position.

Deconstructing the Platform

By handing four out of the five award slots to component developers, the Department of Defense is deliberately dodging single-source traps. They’re ripping the humanoid concept down into workable, modular layers:

• Power Generation and Storage: Hydroplane brings mobile power, while ACES handles storage. Without decent power, bipedal platforms are little more than expensive lawn ornaments after forty minutes.

• Tactile Sensing: RAS Labs' tactile fingertips give mechanical hands the subtle touch needed to handle live artillery rounds, where camera feeds miss tiny shifts in friction.

-

• Control Systems: Velocity Explorations builds the translation bridge between high-level field orders and motor articulation.

• Integration: IHMC supplies the heavy skeleton holding the whole apparatus together.

Put them together, and you see an acquisition model built on interchangeable parts rather than wrapped-up proprietary monoliths.

The China Angle

This whole US push isn't happening in a vacuum.

Over in China, the People’s Liberation Army (PLA) is driving its own hardware program hard. Following the World Humanoid Robot Games, PLA Daily published piece after piece urging research teams to drag prototypes out of university labs and push them straight onto military proving grounds.

Looking through a hundred-plus Chinese military tenders, academic papers, patents, and official notices reveals a push that is anything but theoretical:

• Production muscle: BofA Global Research reckons Chinese outfits churned out roughly 95 per cent of all global humanoid shipments in 2025. Firms like Unitree scaled up fast, occasionally building quadrupeds on initial concepts that trace back to older US military research grants.

• Field runs: PLA institutions like the National University of Defense Technology (NUDT) have been running trials focused on perimeter security, terrain navigation, and target identification. Tenders from 2025 explicitly cover humanoid frames and vision hardware, followed by a $300,000 push in June 2026 purely to map and tag motion data.

• Industrial weight: State defense giant Norinco brought out Fuxi, a full-sized humanoid designed for base security and guard duties in rough weather. It uses a remote teleoperation harness so a human operator can step in the moment the onboard autonomous system gets confused.

• Building assaults: An NUDT study from August 2025 laid out tactical models where six ground bots—mixing humanoids, quadrupeds, and small tracked platforms—move alongside regular infantry to clear multi-storey buildings room by room. Two teams of three bots handle the front doors, taking the point position while soldiers follow behind.

Much of this draws direct lessons from Ukraine, where strike drones, automatic supply buggies, and robotic casualty stretchers completely upended classic infantry tactics.

As UCLA Professor Dennis Hong put it: ‘Going where we don’t want humans to go is one of the most compelling reasons to develop robots in the first place. If losing a robot means avoiding the loss of a human life, then the robot can become expendable.’

Mind you, the technical roadblocks are still massive.

Carnegie Mellon Professor Aaron Johnson and ASPI analyst Malcolm Davis both point out that current humanoids drink power, stumble on uneven ground, and need a constant human safety net.

Wang Yonghua over at China’s Academy of Military Sciences sounded a similar note, warning that unit costs, mobility, and sensor noise have to be sorted out before anyone deploys these en masse—though he admitted that once those hurdles fall, ‘humanoid robots will stream onto the battlefield.’

Then there's the legal side of things.

Autonomous targeting and recognizing a soldier trying to surrender are absolute landmines. While the Red Cross keeps pointing out that machine vision lacks the contextual nuance to spot someone throwing up their hands, current Pentagon guidelines still mandate strict legal reviews and human authority over any lethal call. Beijing says much the same on paper, keeping official control tied to human operators.

Why Two Legs Matter

Why bother with two-legged platforms at all when wheels and tracks are cheaper and far more reliable?

Truth be told, it comes down to basic architecture.

The physical infrastructure of war was designed by humans, for humans.

Warships have narrow gangways and near-vertical ladders. City streets feature broken stairs, narrow doorways, and fallen masonry. Supply crates are sized for human shoulders and arms.

Tracked vehicles hit a hard wall when faced with vertical drops or tight stairwells, and wheeled buggies need something resembling a paved track.

A heavy-duty bipedal machine can move through human spaces without forcing an army to rebuild its bases, ships, or transport vehicles just to fit the hardware.