The Army Is Testing Armored Humanoids

More Human Than Human

In 1968, science fiction writer Philip K. Dick asked a question: Do Androids Dream of Electric Sheep?

That was the title of his novel about a future in which artificial humans had become nearly indistinguishable from the real thing.

They looked human. Moved like humans. Thought like humans.

Loved like humans. Killed like humans.

They were also stronger and smarter than the people who created them.

PKD explored the fragile boundary between human and machine…

And whether empathy was what separated the two.

Nearly six decades later, that question is leaving the pages of science fiction.

In September, the U.S. Army announced the winners of its first xTechHumanoid competition.

It received 103 proposals. Five companies won.

One demonstrated a humanoid called Alex, with impact-resistant joints, ballistic protection, and a combination of artificial intelligence and human control.

Now, this wasn't a competition to buy an army of armed androids.

It was an early assessment of what the technology can actually do.

But consider how we got here… and where we're headed.

Over the past three issues, we followed autonomous machines into the air, across the ground, and beneath the ocean.

Now we reach the most ambitious possibility.

A war machine designed to operate in a world built for humans.

Why Give a Robot Two Legs?

Think about a naval ship.

Its ladders, hatches, valves, tools, and controls are designed for a human body.

So are a truck's pedals and steering wheel.

And most logistics equipment.

We've spent centuries constructing a civilization around two hands, two arms, and the ability to walk upright.

Now imagine a robot that can enter that world and use its equipment without requiring everything to be redesigned and replaced.

A wheeled robot may carry more weight. A robot dog may be more stable on rough terrain. A dedicated machine will often outperform a humanoid at a single task.

But the humanoid's potential advantage is versatility.

The same machine might open a door, operate a tool, unload supplies, and inspect equipment designed for a human technician.

For humanoids, pulling a trigger is a long way off.

The nearer prize is a pair of hands that could work inside vehicles and facilities across air, land, and sea.

That's a different kind of automation.

Seeing Isn't Enough

In Part I of this series, I explained how computer vision detects objects.

A neural network might recognize a coffee mug sitting on a counter.

But recognizing the mug isn't the same as picking it up.

A robot has to estimate its shape, reach for it, control the pressure of its fingers, and respond if the mug starts slipping.

And before it acts, it helps if the robot can predict what happens next.

Researchers call systems that model how the environment changes "world models."

Computer vision gives a machine eyes.

World models, planning, touch sensors, and motion control help it act.

And getting all of those systems working reliably together remains a tremendous engineering challenge.

It's the final AI frontier.

The Pentagon Is Paying Attention

The U.S. Army isn't hiding its interest.

Its September humanoid competition examined complete machines along with tactile sensors, power systems, and control software.

The awards were modest, with limited potential follow-on funding.

Then, on October 13, the Army announced plans to redesignate a major command as the Army Futures and Autonomous Systems Command.

Its job is to bring autonomy into every part of how the Army fights.

Now, we won't see humanoid fighters anytime soon…

But the earliest useful military roles for humanoids may be loading supplies, handling hazardous materials, inspecting dangerous spaces, and maintaining equipment.

Governments are starting to take the technology seriously.

And some want humanoids on the front line sooner than you'd think.

In September, Ukraine's former defense minister announced an "Army of Robots" initiative to develop humanoids capable of reloading machine guns, changing batteries, and eventually taking over the most dangerous infantry jobs.

Russia has experimented with humanoids, too. It even flew one, called FEDOR, to the International Space Station in 2019.

At a Russian defense exhibition in 2022, a robot dog that looked like a Chinese-made Unitree was displayed carrying a rocket launcher.

China is now developing military humanoids of its own, including machines for reconnaissance, patrols, and dangerous assignments.

The civilian opportunity is easier to measure.

The Business Research Company estimates the global humanoid-robot market at $8.32 billion in 2026, reaching $39 billion by 2030.

That's 47.1% annual growth… and where you want to be looking for the next moonshot.

To be clear, this isn't a forecast for robot soldiers…

And that distinction matters for these three humanoid robotics stocks.

Three Humanoid Robotics Stocks I'm Watching

Tesla (TSLA): The Apple of Robotics

Tesla is pursuing the integrated approach.

Think about the early personal computer industry.

Apple designed the hardware, software, and overall experience together.

The PC world evolved around interchangeable parts and specialized suppliers.

Tesla wants to control most of its humanoid's design: intelligence, electronics, mechanics, and manufacturing.

Its Optimus project builds on capabilities developed for electric vehicles: batteries, motors, cameras, AI chips, and large-scale manufacturing.

But building a useful general-purpose humanoid is vastly harder than producing an impressive demonstration.

Tesla reported $28.05 billion in third-quarter revenue, roughly flat from a year ago and almost all of it from established businesses.

Optimus isn't yet a demonstrated, material commercial revenue stream.

Last week, Tesla said early Optimus production had started at Fremont. Musk warned the ramp would be "somewhat bumpy" because it's such a radically new product.

Tesla plans more than $25 billion in capital expenditures during 2026, across multiple projects.

And here's the defense angle readers should understand…

Musk's SpaceX and xAI have been reported as competing in a Pentagon autonomous-drone initiative.

But that's not a Tesla contract, and I haven't found a confirmed military purchase of Optimus robots.

Could the military eventually adapt useful humanoids for logistics, maintenance, or dangerous operations?

Certainly possible.

I'll be keeping a close eye on this story over the coming months for you…

Including actual Optimus production, working hours per robot, manufacturing costs, and paying customers.

Plus, any Pentagon initiative rumors.

NVIDIA (NVDA): The PC Ecosystem of Robots

NVIDIA is pursuing the opposite model.

Instead of building a finished humanoid, it supplies the computing and software used by robot developers and manufacturers around the world.

Its Jetson Thor computers run AI models directly onboard machines.

Its Isaac tools let engineers train and test robots in simulation.

Its GR00T models provide a starting point for learning physical tasks.

In June, NVIDIA introduced an open humanoid reference design combining robot hardware, dexterous hands, Jetson Thor, and GR00T.

The strategic advantage is obvious: a robotics company doesn't necessarily need to create every AI model, simulation tool, and processor from scratch.

And NVIDIA can sell into warehouses, factories, medical equipment, and potentially defense robotics without betting on one winner.

In its latest quarter, NVIDIA reported $96.2 billion in revenue, up 106%. Data-center revenue accounted for $89 billion.

That's extraordinary financial strength.

But it also shows the problem for investors looking for a pure robotics bet…

NVIDIA doesn't separately report robotics as a major revenue business.

So, it'll take substantial adoption to make a material difference to a company this large.

Allient (ALNT): The Robot Still Needs Muscles

You can give a robot brilliant AI…

But it won't accomplish much if it can't move its fingers.

A humanoid needs motors, actuators, precision gears, encoders, and electronic controls.

These determine how accurately it moves, how much energy it consumes, and whether it can work all day without breaking down.

And unlike a software model, you can't improve a poorly engineered joint merely by training on more data.

That's why I'm watching Allient, a smaller manufacturer of precision motion and control technologies serving industrial, aerospace, defense, and medical customers.

It already sells products where reliability and engineering expertise matter.

Second-quarter revenue was $153.8 million, up 10%.

New orders reached $201.3 million, up 49%, and backlog grew to $298 million, up 26%.

Management expects most of that backlog to become revenue within three to nine months.

Those are real customer commitments across its existing businesses.

The investment thesis is that increasingly sophisticated robots could become another attractive market for the precision components Allient already knows how to build.

So, I'll keep an eye on new robotics design wins, orders converting to sales, margins, and whether customers find its components difficult to replace.

The Last Human in the Loop

Four issues, four environments: air, land, water, and now the human form.

We began above the battlefield, where cheap drones can threaten expensive aircraft.

Then came ground robots that carry supplies, evacuate wounded soldiers, and sometimes carry weapons.

Underwater, autonomous vehicles navigate where radio and GPS barely work.

And now, humanoids.

Picture an army of mechanical soldiers marching toward an enemy.

The more startling possibility is quieter: intelligence that can see, predict, and handle the physical world could eventually operate much of the equipment people already use.

That possibility extends far beyond warfare.

It reaches into factories, farms, hospitals, warehouses, and homes.

Technology wants to be everywhere. And AI is just taking its first steps.

The investment opportunity could be substantial.

So could the consequences of getting it wrong.

Today we're teaching machines to use their hands.

What they eventually do with those hands is a human decision…

Until these robots start making decisions on their own.

So it goes.

Always be prospering,

Harry Seldon
Moonshot Minute

🔓 Premium Content Begins Here 🔒