When Alvin Graylin toured Unitree’s headquarters and factories, he asked about the customers. Nearly all were research labs, he recalled being told. Others bought robots for demonstrations, including kickboxing. Very few of the machines, he said, were doing commercial work.
But he also heard about a change in shape. Unitree staff told him they were moving toward an upper-torso model, without legs, and that those machines were selling better to commercial customers.
Graylin brought the contrast to Moonshots with Peter Diamandis: a robot built for useful factory work may have less reason to resemble a whole person than one built to demonstrate what robotics can do.
The case against feet
Humanoid robots borrow the human body plan: a torso, arms and, in walking versions, two legs. Graylin’s objection is that a repetitive task may need only part of that package.
He recalled a Figure demonstration in which a robot sorted packages for roughly ten hours. His criticism was about cost rather than whether the robot could sort. A simpler machine with one or two arms could have done the same job at a fraction of the price, he argued. “You didn’t need a full body to do that,” he said.
His account of the Unitree visit followed the same reasoning. A legged machine has to keep balancing. Falls can damage it, bringing repairs and maintenance. Removing the legs means fewer components; a large base can accommodate a larger battery and allow longer operation.
For a repetitive job at one station, that makes a fixed arm a plausible alternative. Where movement is needed, an upper body on a wheeled base offers another way to keep the arms without paying for a walking mechanism. Graylin’s argument is to choose the body around the work, rather than assume the work requires a human-shaped machine.
The panel reached for an aircraft analogy. Human bodies reflect billions of years of biological evolution, constraints included. A machine need not reproduce that history: a plane flies without moving through the air the way a bird or insect does.
Factory automation is a different market
Asked whether China’s robotics push reflected government incentives or workforce pressures following the one-child policy, Graylin first separated established automation from the humanoid boom.
Factories have already been automating in response to demographic pressures, he said. He described highly automated plants, often called “dark factories,” where a small number of people oversee the machinery. That is a different picture from a research lab buying a walking robot or a company putting one on stage.
The International Federation of Robotics supplies a measure of the industrial scale. In its September 25, 2025 release, it reported that China installed 295,000 industrial robots in 2024, up 7% from the previous year. That represented 54% of worldwide installations that year. China’s operational stock—the accumulated fleet still in use—reached 2.027 million.
Those figures count industrial robots broadly, not humanoid shipments. And the 54% figure describes China’s share of new installations in 2024, not its share of every factory robot operating worldwide. The expansion also reached beyond traditional sectors: food-and-beverage installations rose 86% to 8,900.
A crowded field and a shared parts supply
Graylin sees a much smaller market supporting the proliferation of humanoid companies. He recalled seeing more than 200 businesses demonstrating humanoid-related products at a conference, while estimating global humanoid volumes in the tens of thousands in the previous year. By his estimate, 80% to 90% came from China, concentrated in two or three companies.
He expects consolidation over the next year or two, potentially leaving a single-digit number of surviving humanoid companies. That is his forecast of a shakeout, not a count of businesses already disappearing.
The commercial calculation also depends on access to parts. Discussing possible restrictions on Chinese robots, Graylin estimated that 90% of robot components came from China. He repeated rumors of US robotics companies buying components and actuators—the mechanisms that produce movement—in China and carrying them home in suitcases because they could not obtain them domestically.
He used those accounts to argue that domestic manufacturing ambitions cannot quickly undo dependence on overseas suppliers. “I don’t think we can stuff that genie back in the bottle,” he said of globalization.
For buyers choosing a robot now, his Unitree account points to a more immediate decision: whether legs help enough to justify maintaining them. The commercial interest he described was in an upper torso that retained the useful arms and left the walking mechanism behind.