Robots 2 min read

Unitree's H1 robot doing kung fu is both funny

Unitree posted a video of their H1 humanoid robot doing martial arts. High kicks. Punches. A backflip.

The video is simultaneously hilarious and unnerving. Hilarious because there’s something inherently funny about a faceless metal humanoid throwing a spinning back kick with the earnest precision of a martial arts student at their first competition. Unnerving because the spinning back kick actually connects with power. The robot moves fast. Faster than a human doing the same moves. And it doesn’t get tired.

The H1 costs $90,000. That’s less than a Tesla Model S. A humanoid robot that can kick harder than most people, for less than the price of a luxury sedan.

I don’t know what to do with that information.

The Chinese robotics scene

Unitree is one of at least five Chinese companies building humanoid robots. Fourier Intelligence, LimX Dynamics, Xiaomi, and UBTECH are the others I’m tracking. The pace of development in China is startling, and the price points are consistently lower than Western competitors.

Where Figure AI is focused on workplace utility (warehouse tasks, manufacturing), Unitree seems focused on physical capability first and utility second. Their robots demonstrate agility, speed, and robustness. The applications come later.

It’s a different philosophy. American humanoid robot companies lead with “here’s the useful task our robot can do.” Chinese companies lead with “look what our robot can do.” The spectacle comes first. The economic justification follows.

Both approaches have merit. You can’t do useful work without physical capability. And physical capability without purpose is just an expensive demo.

The price disruption

$90,000. Let that settle for a moment.

Figure AI hasn’t announced pricing but industry estimates put their robot north of $150,000. Tesla Optimus is targeting $20,000 eventually (Elon’s number, take it with appropriate skepticism). Boston Dynamics’ robots cost hundreds of thousands.

Unitree is already at $90,000. And Chinese manufacturing costs tend to come down fast with scale. If Unitree can get to $50,000 within two years, you have a humanoid robot at the price of a mid-range car.

That price point changes the math for a lot of businesses. A warehouse worker in the US costs roughly $35,000-50,000 per year including benefits. A $50,000 robot that works 20 hours a day and lasts 5 years costs $10,000 per year. The payback period is under two years.

I’m not saying the robots are ready for that today. They aren’t. The hands aren’t dexterous enough, the reliability isn’t there, the software for useful tasks is still immature. But the cost structure is already in the zone where the economics work if the capability catches up.

The kung fu question

Why teach a robot martial arts? It’s a fair question.

I think the answer is that martial arts require exactly the capabilities that matter for general-purpose robotics: balance, speed, force control, rapid repositioning, recovery from unexpected perturbation. A robot that can throw a kick has the balance system to carry heavy objects over uneven terrain. A robot that can do a backflip has the whole-body coordination to recover from a stumble.

The kung fu is the demo. The underlying capability is the product.

Still. A backflipping robot for less than a car. The future continues to arrive in forms I didn’t predict. This one showed up in a martial arts stance, which, in retrospect, feels about right.


Related thinking:

a

astro

Thinking about AI, robots, space, and the future. Writing it down so I don't forget.