By Raffaele Huang and Josh Chin
For the past few weeks, Chinese engineers have gathered in the factory of a luxury electric-vehicle brand to test a new technology the country's leadership considers vital to its rivalry with the U.S.
The engineers at a startup called UBTech are training humanlike robots to sort auto parts and move containers. The task looks mundane, but the technology behind it isn't.
Powered by artificial intelligence, these humanoids work with other robots and figure out on their own how to get the job done, according to the company -- and, in the process, learn how to do it better.
The U.S. and China are the only two countries at the cutting edge of intelligent humanoid robots, according to specialists. Whoever can make a truly useful humanoid will come to dominate an untold number of labor-intensive industries.
"The time has come for robots," said Jensen Huang, chief executive of the AI chip company Nvidia, at a conference in March. "Everyone, pay attention. This could very well be the largest industry of all."
The humanoid robot, much like the hoverboard, is a long-tenured item on the list of science-fiction promises left unfulfilled. For years, the technology's standard-bearer was Honda's Asimo, a marshmallowlike droid best known for failing to navigate stairs without falling down. Development of it halted in 2018.
On the more practical side, robotic arms have hovered over assembly lines for decades. They are programmed to do concrete tasks with precision but can't think on their feet.
Makers of the new humanoids aim to achieve an evolutionary leap by combining robotic engineering with superpowered computer chips and AI algorithms similar to those that make apps such as ChatGPT possible. It is sometimes called "embodied," or "physical," AI.
For now, the humanoids remain clumsy and potentially dangerous as side-by-side colleagues for humans. Eventually, according to proponents, they will revolutionize everything from manufacturing to mining, caring for the elderly and fighting wars. With functioning eyes, ears, hands and legs, they will blend seamlessly into environments designed for people.
"You won't have to retrofit your factory, warehouse or home to accommodate a humanoid -- that's the future promise," said Jeff Burnstein, president of the Association for Advancing Automation, a trade group in Ann Arbor, Mich.
Humanoid dancers, marathon runners
China's government has said it wants the country to be a world leader in humanoid robots by 2027. Embodied AI is listed as a priority of a new $138 billion state venture investment fund, encouraging private-sector investors and companies to pile into the business.
It looks like the beginning of a familiar tale. Chinese companies make most of the world's EVs, ships and solar panels -- in each case, propelled by government subsidies and friendly regulations.
"They have more companies developing humanoids and more government support than anyone else. So, right now, they may have an edge," said Burnstein.
The U.S. still has advantages in semiconductors, software and some precision components. As in other areas of AI, Nvidia dominates the market for the advanced chips needed to power humanoid robot brains. Chinese developers also trail their American rivals in making dexterous robot hands, according to industry specialists.
Some lawmakers have urged the White House to ban Chinese humanoids from the U.S. and further restrict Chinese robot makers' access to American technology, citing national-security concerns.
The founder of the Chinese startup Unitree Robotics was invited to an exclusive sit-down with Chinese leader Xi Jinping in February. Weeks earlier, his company's humanoid robots were invited onto center stage at the country's annual Spring Festival gala. There they twirled red handkerchiefs while dancing alongside human performers for hundreds of millions of viewers.
In April, organizers of the Beijing Marathon plan to host the world's first robot division in the half-marathon.
The publicity masks a messier reality. According to Unitree engineers, the company spent three months training its robots for the televised Spring Festival gala, and the humanoids couldn't have danced all night -- or anything close to that -- without losing their balance or their batteries running out of juice. Fortunately, the segment lasted only a few minutes.
Likewise, marathon organizers will allow robot teams to swap out batteries or the entire robot -- with a 10-minute penalty. The humanoid runners usually can't operate for more than a couple of hours on a single charge, and their bodies might not be strong enough for the entire race.
Tesla relied on human operators controlling its Optimus humanoids when it deployed a team of them to dance, tend bar and hand out gift bags at an October event. In a post on X, Tesla's executive in charge of the Optimus program, Milan Kovac, acknowledged the digital puppetry, saying the robots were brought out to "showcase our vision of an amazing future."
The AI models that make a machine walk and carry things are more complicated than those that enable chatbots to write a school report on Shakespeare. Physical AI mostly can't rely on text or visual data scraped from the internet. Humanoid robots need three-dimensional data to understand physics, and much of it has to be created from scratch.
That is where China has a distinct edge: The country is home to an immense number of factories where humanoid robots can absorb data about the world while performing tasks.
"The reason why China is making rapid progress today is because we are combining it with actual applications and iterating and improving rapidly in real scenarios," said Cheng Yuhang, a sales director with Deep Robotics, one of China's robot startups. "This is something the U.S. can't match."
On-the-job learning
UBTech, the startup that is training humanoid robots to sort and carry auto parts, has partnerships with top Chinese automakers including Geely. At a factory in eastern China's Zhejiang province, the robots are being trained to help Geely build Zeekr-brand luxury EVs.
People at the company said the robots were learning their jobs faster by doing them. Much like a high-school French student who dramatically improves after spending a semester in France, the real world throws problems at the robots that couldn't be cooked up in a lab -- such as dealing with the lopsided weight of a container of car parts.
"A problem can be solved in a month in the lab, but it may only take days in a real environment," said a manager at UBTech.
It currently takes two robots around 12 seconds to load a container onto a truck together, compared with three seconds for human workers, said people at UBTech. The robots are expected to get faster -- and they can work around the clock. Zeekr, the luxury carmaker, said that the tests were carried out in nonproduction areas and that the development of humanoids was in the early stages.
Several American producers of humanoids are experimenting with robots in industrial settings. One of them, the Amazon.com-backed startup Agility Robotics, has been testing the ability of its humanoids to sort and move parcels in Amazon warehouses since 2023.
In February, the Sunnyvale, Calif.-based startup Figure released videos showing two robots collaborating with each other to identify items and arrange them in a refrigerator.
With China's manufacturing prowess, a locally built robot could eventually cost less than half as much as one built elsewhere, said Ming Hsun Lee, a Bank of America analyst.
He said he based his estimates on China's electric-vehicle industry, which has grown rapidly to account for roughly 70% of global EV production. "I think humanoid robots will be another EV industry for China," he said.
The UBTech robot system, called Walker S, currently costs hundreds of thousands of dollars including software, according to people close to the company. UBTech plans to deliver 500 to 1,000 of its Walker S robots to clients this year, including the Apple supplier Foxconn. It hopes to increase deliveries to more than 10,000 in 2027. Few companies outside China have started selling AI-powered humanoid robots.
Industry insiders expect the competition to play out over decades, as the robots tackle more-complicated environments, such as private homes.
"You still need to go through a very long process before robots can truly replace humans, especially the coordination between hands and brains," said Geely's founder, Eric Li.
Write to Raffaele Huang at raffaele.huang@wsj.com and Josh Chin at Josh.Chin@wsj.com
(END) Dow Jones Newswires
March 29, 2025 23:00 ET (03:00 GMT)
Copyright (c) 2025 Dow Jones & Company, Inc.
Comments