Unitree's IPO Prices Every Humanoid

Most IPOs price a company. A few price a category. Unitree will price a category.

The category is humanoid robots, and the bet is that AI will not only live on screens but go to work in the physical world. The human shape is the point, because a robot with legs and hands can work in a factory or a home built for people without anyone redesigning the space. Humanoid companies have absorbed billions in private capital, and physical AI is where the next wave of it is going. It is about to be tested in the public markets, with the Unitree IPO in mainland China.

Let's start with what the private market believes. Figure, the best funded humanoid company in America, is valued at $39 billion and is now shipping robots to paying customers. Tesla has built hundreds of Optimus robots, its own humanoid, but publishes no production figures and sells the robot to nobody. Those are the two American companies priced as though they will win humanoid robotics.

The only humanoid maker at public market prices today is UBTech in Hong Kong, and it loses money. Revenue rose 53% last year to about $280 million against a net loss of roughly $110 million. It also delivered 1,079 full-size humanoids in 2025, the first company anywhere past a thousand in a single year, more than 500 of them industrial units working inside factories including BYD and Foxconn. Humanoid revenue grew more than twentyfold and now accounts for 41% of the company.

Now Unitree. It reported about $240 million of revenue for 2025, up more than fourfold, with roughly $40 million of net profit and gross margins above 60%. The China Securities Regulatory Commission cleared it to list on July 1 of this year. It is raising about $618 million by selling 10% of the company on Shanghai's STAR Market, which values it near $6 billion. It will be the first profitable pure-play humanoid maker to list anywhere.

One more mark, and it is the freshest. On July 16, Hyundai agreed to buy SoftBank's remaining stake of about 10% in Boston Dynamics, which has been building robots longer than anyone in America, for roughly $325 million, taking full ownership. That values the company around $3.3 billion, in an arm's length transaction between industrial buyers rather than a venture round. Hyundai plans to put Boston Dynamics' humanoid on the line at its Georgia car plant from 2028, starting with parts sequencing and widening to assembly by 2030.

Line them up. Figure is worth $39 billion with a few hundred robots built. Boston Dynamics just changed hands at $3.3 billion. Unitree is asking $6 billion with thousands of robots delivered and profit on the books. The company with the most evidence is not the one with the highest price.

Unitree did not start with humanoids. It started in 2016 selling quadrupeds to university labs at prices that undercut Boston Dynamics by an order of magnitude. It was not a glamorous business. Robot dogs for researchers is a small market with demanding customers and thin margins.

Intelligence will decide who wins humanoid robotics. Joints decide who gets to compete. To hit those prices Unitree pulled motors, actuators and reducers in house and built its own relationships across the Chinese component base. By the time humanoids became the story, the drivetrain existed and the cost curve had already been walked down.

You can see the handoff in the revenue mix. Humanoids were 1.9% of Unitree's core revenue in 2023. They passed half of it in 2025. The company did not pivot into humanoids. It carried a component stack from one form factor to another, and reached profitability in 2024 while doing it.

A humanoid robot is mostly joints. Each one needs 20 to 40 harmonic reducers, and every reducer runs on rare earth permanent magnets. China processes about 90% of those magnets and builds the cheap, high-volume end of the joint. It does not yet own the precision end. Japan's Harmonic Drive still holds roughly 80% of the harmonic reducer market, and European and Japanese suppliers make about 90% of the precision ball screws. Chinese manufacturers are spending tens of billions of yuan to close that gap, and one firm alone is building capacity for 980,000 roller screw sets. China owns the volume today and is buying its way into the precision tier.

Figure's bill of materials and Unitree's bill of materials come from the same suppliers. One pays the export markup, the freight, the tariff and the margin stack. The other is the domestic customer. American humanoid companies are not competing with a cheaper rival. They are competing with their own supplier's customer. They know it. Figure builds its own actuators and Tesla is pulling production in house. The question is whether either can do it at a price that competes.

The second advantage is quieter and more durable.

Roughly 13,000 humanoid robots were delivered worldwide in 2025. Unitree says it accounted for more than 5,500 of them. Omdia, which tracks shipments independently, puts Unitree at 4,200 and ranks AgiBot first with more than 5,100. The two are arguing over which Chinese company leads the world, with UBTech third. Figure produced more than 350 robots by April, and BotQ is set to build 12,000 a year. Unitree is guiding to 10,000 to 20,000 units in 2026.

These are not the same machines, and pretending otherwise is how people get robotics wrong. The Unitree R1 is 121 cm and 25 kg with about an hour of runtime, a development platform. Optimus carries roughly 20 kg of payload on a battery built to work a shift. But the difference that matters right now is not specification. Tesla is targeting $20,000 to $30,000 for Optimus and has not delivered one to a customer. Unitree sells the R1 at $5,900 and the G1 at $13,500 today, and its average humanoid went out the door last year at about $24,000 with a 63% gross margin.

At the bottom of that range a robot sells itself. A graduate student with a grant, a systems integrator testing a concept, a factory engineer running an experiment. No sales call, no pilot, no procurement cycle. Price is not only a margin decision in this category. It is a distribution channel, and distribution generates the operating data that makes the next robot better.

The strongest argument against all of this is that Unitree's units go to researchers and developers rather than industrial labor, so the comparison is not clean. Figure is building for a market where the alternative is a human being costing six figures a year, and if it delivers, the value per unit is not comparable.

That argument is fair, and Figure is answering it. Figure 03 units are working at BMW Spartanburg, and in May the company signed a commercial agreement with Catalyst Brands to run robots at its Reno distribution center. The question was never whether Figure can do the work. It is whether being more than six times as valuable as Unitree is earned while shipping a fraction of the units, at a loss, using Chinese parts.

What nobody has proven yet, on either continent, is the return. UBTech loses money with robots inside auto plants. Unitree makes money selling mostly to researchers and developers. The Chinese advantage today is cost and volume, which is a strong position and not the same thing as a proven end market.

Unitree owns its components, ships at volume, and makes money. UBTech has more than 500 industrial humanoids working inside plants like BYD and Foxconn. Figure has real customers, real deployments, and a supply chain that runs through its competitors' home market. UBTech already trades. Unitree will within weeks. Figure's $39 billion gets tested when it raises again or goes public.

The Unitree IPO is what I am watching this quarter. The next Figure robot is what the private markets will be watching.

Bashar Aboudaoud
Managing Member, UpRound

Live Webinar: Thursday, September 17

You leave with three things. Where private valuations actually sit right now. Which sectors are repricing fastest. How I am reading prices into 2027.
20 minutes, then live Q&A. 20 slots only for UpRound readers.