The U.S. Military Funded the Breakthroughs. China Built the Robot-Dog Empire.

A new investigation traces some of Unitree’s most successful robot dogs back to breakthroughs developed through U.S. military-funded research. America helped advance the technology. China figured out how to manufacture it at scale.

Unitree Go1 robot at Romanian Land Forces Day in Bucharest in 2023. Photo: Sgt. Amber Edwards, U.S. Army. Public domain.

There is something almost painfully ironic about this story.

For years, the U.S. military poured money into research trying to teach four-legged robots how to run, jump, recover from falls and move across difficult terrain. Researchers at places including MIT and the University of Pennsylvania made major breakthroughs and published much of what they learned openly so other engineers could build on it.

China did exactly that.

Today, Unitree Robotics is one of the biggest robot manufacturers in the world, selling quadruped robot dogs at prices dramatically below many American alternatives. Its robots can run, climb stairs, roll over, map environments and perform movements that looked like science fiction not very long ago.

And according to a new Reuters investigation, some of the engineering principles behind Unitree's most successful robot dogs can be traced directly to breakthroughs developed through research financed by the U.S. Army and other American military programs.

The strange part is that Unitree did not steal secret military technology.

The research was published.

That was the point.

America helped invent better ways for robot dogs to move. China became much better at turning those ideas into something people could actually buy.

The similarities were apparently hard to miss

One of the most important robots in this story is MIT's Mini Cheetah.

MIT unveiled it publicly in 2019. It weighed around 20 pounds, could run across uneven ground, recover after being knocked over and even perform a standing backflip. MIT described it as a relatively inexpensive, modular research platform designed so other laboratories could experiment with advanced robotic movement.

Benjamin Katz worked on the Mini Cheetah in MIT's Biomimetic Robotics Lab. When Reuters asked him about Unitree's later Go-series robots, his description was striking.

Their dimensions were almost identical “to the millimeter,” he said.

That does not mean Unitree simply photocopied an American robot.

Another MIT researcher, Joao Ramos, told Reuters that Unitree borrowed heavily from the Mini Cheetah's overall shape, structure and design but also credited the Chinese company with doing the difficult engineering needed to turn those concepts into reliable mass-produced machines.

And that second part is really important.

Having a brilliant prototype in a university laboratory is one thing.

Building thousands of them cheaply enough that people actually buy them is something completely different.

Unitree's founder openly studied the American research

There is no secret spy-thriller reveal here.

Unitree founder Wang Xingxing referenced MIT's work in his own 2016 master's thesis.

Reuters reviewed the thesis and found Wang cited earlier MIT Cheetah research and acknowledged the lab's pioneering work. He also referenced work from University of Pennsylvania researchers involved in U.S. Army-funded robotics programs.

Unitree itself says it independently developed important components including its motors, reducers, controllers and some sensors.

So the more accurate story is not:

“China stole an American robot.”

It is considerably more interesting:

American researchers developed and openly published major breakthroughs. Chinese engineers studied them, improved the practical implementation, built a supply chain around them and then manufactured robots at a scale the United States never achieved.

That is the part America is now struggling with.

Then came the $1,600 robot dog

Unitree launched the Go2 in 2023 with an entry model priced around $1,600, according to Reuters. The company's current official store lists newer configurations at higher prices, but that original price point was a huge part of the product's appeal.

Unitree robots, including a Go2 quadruped, displayed at the 2025 Japan Mobility Show. Photo: Yuki Uchida / Wikimedia Commons, CC BY 4.0.

The Go2 weighs around 15 kilograms and can be equipped with 4D LiDAR, cameras and onboard computing. Depending on the configuration, Unitree lists maximum speeds reaching around 5 meters per second, or roughly 18 km/h.

Watch the official Unitree Go2 demonstration

That video is useful because you can actually see why these machines caught people's attention. They are no longer awkward laboratory robots taking careful little steps.

They run.

Jump.

Climb.

Recover from falls.

And some versions cost closer to a high-end computer than a military vehicle.

China then did what China has become extremely good at doing

Scale.

Unitree sold more than 18,000 quadruped robots and 5,500 humanoid robots last year, according to company filings reviewed by Reuters. The company is valued at roughly $9 billion ahead of its Shanghai stock-market debut.

Compare that with the United States.

Reuters reports that no American company is mass-producing comparable robots at Unitree's scale. Ghost Robotics produces military-focused quadrupeds, while Boston Dynamics makes its famous Spot robot, but neither has achieved Unitree's combination of volume and low pricing.

Ghost Robotics has shipped roughly 1,000 Vision 60 robots, mostly to military and government customers. Those machines can cost upwards of $165,000.

That comparison is not perfectly apples-to-apples. A hardened military robot and a consumer Go2 are designed for very different jobs.

But the price difference tells you why Unitree matters.

China did not merely make quadruped robots work.

It made them cheap enough to spread.

Even Elon Musk has acknowledged China's advantage

Tesla has spent years developing its Optimus humanoid.

China already has companies shipping thousands of humanoid and quadruped robots.

On a Tesla earnings call earlier this year, Elon Musk described China as “incredibly good” at scaling production.

That might be the central lesson of this whole story.

The U.S. is still exceptionally good at research.

China has become exceptionally good at taking promising technologies and surrounding them with factories, component suppliers, manufacturing expertise and enough capital to make them inexpensive.

Motors.

Gears.

Magnets.

Actuators.

Sensors.

Battery components.

If all of those suppliers sit relatively close to your robot factory, building the thousandth robot becomes much easier than building the first one.

Personally, this is what I find more interesting than the usual “China copied America” argument. Copying an idea does not automatically give you an industry. Somebody still has to manufacture the thing reliably, make it affordable and convince thousands of customers to buy it.

Unitree did that.

Then the robot dogs started appearing with soldiers

This is where the story stops being just an argument about manufacturing.

Unitree has repeatedly said its robots are intended for civilian use. In 2022, it joined Boston Dynamics and other robotics companies in an open pledge opposing the weaponization of advanced mobile robots.

But Unitree-built machines have nevertheless appeared in military settings.

Chinese state television has shown Go2 robots operating alongside People's Liberation Army troops, while footage has also shown a modified Unitree B1-series quadruped firing a mounted assault rifle. Reuters does not report that Unitree itself performed the weapon modification.

That distinction matters.

A company can sell a general-purpose robot.

Someone else can attach something to it.

Watch Reuters' investigation and footage

Reuters' August 18 video report includes the robotics research, Unitree machines and military context behind the investigation.

The United States has experimented with the exact same idea.

In September 2023, U.S. Marines conducted an exercise in which a rocket launcher was attached to a Unitree Go1 for remote firing.

Yes, an American military unit experimented with attaching a weapon to a Chinese robot dog based partly on technology that emerged from American military-funded research.

This story keeps getting stranger.

The U.S. Army even tried to buy Unitree robots

Possibly the most awkward detail came in 2023.

According to procurement records examined by Reuters, the U.S. Army obtained a federal waiver allowing it to seek two Unitree Go2 robots despite rules prioritizing American-made products.

Why?

The procurement documents said comparable U.S.-produced alternatives were not available in sufficient commercial quantities and satisfactory quality.

Reuters could not establish whether the purchase was ultimately completed.

Still, think about what that means.

The U.S. military funded research that helped advance quadruped robotics.

A Chinese company turned similar ideas into inexpensive commercial products.

Then a branch of the U.S. military considered buying those Chinese products because the domestic market did not have enough suitable alternatives.

You could not design a better example of the difference between inventing technology and industrializing it.

Washington is now trying to close the door

The relationship has changed dramatically.

In June, the Pentagon added Unitree to a list of companies it considers contributors to China's defense industrial base. The designation is not the same as a full sanction, but it restricts future U.S. military relationships with the company. Unitree maintains that its products are civilian.

Then in July, the Federal Communications Commission moved to ban future imports of certain foreign-made humanoid and quadruped robots, including Unitree products, on national-security grounds. China has threatened retaliation.

The obvious question is whether bans solve the underlying problem.

Several researchers Reuters interviewed do not seem to think so.

Their argument is basically that America cannot protect its way back into manufacturing leadership.

You still need factories.

Suppliers.

Engineers.

Capital.

Skilled workers.

And companies willing to turn research prototypes into millions of physical products.

The researchers do not regret publishing the work

This surprised me.

The scientists Reuters interviewed were not arguing that MIT and other universities should have hidden their discoveries.

The U.S. Army Research Laboratory also defended publishing the research, saying open scientific work strengthened America's broader robotics ecosystem.

That makes sense.

If every university immediately classified every robotics breakthrough because somebody overseas might use it, scientific progress would slow dramatically.

MIT's own goal with Mini Cheetah was almost the opposite. Researchers wanted other laboratories to have robots they could experiment with, modify and use to accelerate robotics research.

The problem was not that somebody else learned from the research.

The problem was what happened afterward.

America had the breakthrough. China built the factory.

That is really the story. The U.S. military spent years funding ambitious research into robotic movement. American universities produced some extraordinary machines. American engineers published the results. Chinese researchers studied them. Unitree engineered commercially viable versions. China's manufacturing ecosystem supplied the components.

Prices fell.

Production increased.

And now Unitree sells more quadruped robots than companies in the country where many of the underlying breakthroughs were originally developed. Robotics analyst Reyk Knuhtsen described America's failure to commercialize its own research as a “dropped ball.”

I think that is probably the most important part of this investigation. The technology race is not won when somebody publishes the best paper. It is not even necessarily won when somebody builds the best prototype.

It is won when somebody figures out how to build the technology 10,000 times, cheaply, reliably and fast enough that the rest of the world starts buying it.

Right now, Unitree seems to understand that extremely well.

And tomorrow, when the company begins trading publicly in Shanghai, investors will finally get to put a price on just how valuable that advantage has become.