China’s running, dancing, kung-fu fighting humanoid robots are better at being influencers than anything else. But soon they will have to get real jobs, and fast.
In this episode, editor Tom Mitchell and reporter Savannah Billman discuss the explosive growth of Chinese humanoid robots this year. China expects to produce 100,000 of them, much more than anticipated, but is the technology developed enough for customers to buy them all?
Transcript
Tom: Hi, and welcome to The Wire China podcast. I’m Tom Mitchell, the Wire’s Features Editor.
In the podcast, we aim to take you behind the scenes of the stories we cover in our magazine. This week, I’m joined by our staff writer and podcast producer, Savannah Billman. Last week, Savannah wrote about the potential exponential increase in the number of humanoid robots that China will produce this year.
We will be discussing that in a minute, but first a quick note about our other articles in our coming edition, which goes online Sunday evening, New York time.
In other pieces, Victor Shih writes about the legacy of Zhu Rongji, China’s former and formidable premier, who died on August 12th, age 97. Our cover story by Rachel Cheung and Peiyue Wu is about the latest big AI moment in China, the release of Moonshot AI’s new model, Kimi K3, which has impressed experts, but also worried U.S. politicians. Rachel and Peiyue also profile the company’s 34-year-old founder, Yang Zhilin.
Savannah also profiles Geely, the Chinese auto champion, in this week’s Big Picture. To get this story and the rest of our issue in your inbox this weekend, you can sign up for our free newsletter via the subscribe link in the episode description, or go to our website, thewirechina.com.
Hi, Savannah.
Savannah: Hello, Tom.
Tom: Today, we are discussing your article on China’s humanoid robots industry. Tell us first what the Chinese Communist Party envisions for humanoids.
Savannah: The CCP has been fairly clear about its AI ambitions from the jump in recent speeches, and policy documents, including the 15th Five-Year Plan. China calls for the implementation of strategic deployments in AI, opens the door for, quote, unconventional arrangements to cultivate more AI talent, and to use AI everywhere from education to the government. But as Xi Jinping said in his July speech at the World AI Conference in Shanghai, China, AI is moving from the digital to the physical world.
And that’s where a field called embodied intelligence comes in. Embodied intelligence is basically giving the brain of an AI model a machine body. That might be something as mundane as a smart robot vacuum, or something as complex as a humanoid.
Robots are identified as the strategic emerging industry in the 15th Five-Year Plan, along with aerospace and smart cars. But within that group of robotics, which is very broad, humanoids are a very small but very interesting and eye-catching sector.
Tom: And what are the projections for humanoid robot production this year in China, and how does that compare to 2025?
Savannah: As I mentioned, humanoids are a very small part of the overall robots produced and used in China.
When I speak to people about the robotics industry, there’s an interesting, more philosophical question. Why do the robots have to look human? Why can’t these robots have three or four arms or more legs to do more tasks? And in China, it is kind of true that robots not shaped like humans are doing much better. In 2024, China deployed nearly 300,000 industrial robots, more than half of all global deployments.
And for the first time, that number was mostly drawn from Chinese domestic manufacturers, rather than foreign robot makers. In comparison to industrial robots, which are not shaped like humans, humanoid sales are lagging. China’s two leading robot firms, Unitree and Agibot, combined sold fewer than 11,000 humanoids last year.
This year could very well mark a turning point, however. Also at Shanghai’s World AI Conference, Chinese officials announced that the country is expected to produce 100,000 humanoids this year, which is far higher than any other estimate that analysts have put forward. So this is definitely looking like it will be a banner year for humanoids in particular, although the robotics field in China has been growing much more than that for some time already.
Tom: Why such a big year-on-year increase for humanoid robots this year?
Savannah: The practical answer is that many robot makers now have the factory capacity to produce humanoids at scale. Leju Robot has launched a new factory in Guangdong this year that can produce 10,000 robots.
Agibot is another interesting case. It took the company two years, from 2023 to December 2025, to produce 5,000 robots. But in just the first six months of 2026, it produced an additional 10,000 robots. That firm’s capacity has scaled rapidly as well.
The bigger picture answer from the perspective of the industry is that the companies need more real-world data to build better bots.
Last year, I visited a robotics incubator in Shanghai, and I saw some of these humanoids. They’re about as big as a fourth grader, and they could dance, shake your hand, wave at you, and recover from a stumble if you lightly pushed them.
It was a bit creepy. They mimicked these human movements very well, but we had to remember that all of these movements were coded and built by engineers.
In a real-world working scenario in a factory, you have a human worker on the line. If you need to switch an employee from one task to another, you can do so very easily, and the human worker can obviously adjust to different pacing and different actions. A robot can’t do that yet. So the use case of these humanoids is still quite small and niche.
For the humanoid industry to actually be useful, to scale, and to make a profit, these companies simply just need thousands more robots in real-world situations collecting data for them to build more movements. So this is the bigger picture reason. The industry has to get much bigger first before it can do really anything else.
Tom: Two of the humanoid robot makers you mentioned, Unitree and Agibot, are expected to have a combined 80 percent of the market this year. Why are they and market leader Unitree in particular so far ahead of the competition?
Savannah: So Unitree built its reputation on quadruped robots, aka robot dogs, for years before expanding into humanoids. So they already had a lot of experience in manufacturing these sorts of devices, working with embodied AI, and they already had a customer base that was interested in advanced robotics.
In 2023, Unitree’s humanoids were less than two percent of its revenue, but over 50 percent of its revenue in 2025, according to its IPO prospectus. To put it in a different perspective, in 2023, Unitree only sold five humanoids, and it sold well over 5,000 last year. So that just illustrates the explosive scale and speed of Unitree in particular.
And I already mentioned how quickly Agibot has scaled its production.
Some of this also relates to where humanoids are currently being sold and deployed. The majority of these robots work in manufacturing, logistics centers, or automotive plants.
A small subset of them work as what we can call entertainers, doing backflips, dancing at theme parks, things like that going viral on TikTok. Unitree and Agibot make those sorts of robots, but there are other companies in China building specialized humanoids for other industries. Some firms are making healthcare-specific robots.
That field hasn’t totally scaled yet. Chinese robotics firm UBtech has also just launched a line of what they call ultra-bionic humanoids, which are supposed to meet human needs for companionship and look very eerily human-like. That use case hasn’t really scaled up yet either. It’s not like we’re seeing people walking around Shenzhen with robot boyfriends at this time.
So Unitree and Agibot not only have the production capacity to scale much larger, but they’re making the types of humanoids that society can absorb at this moment. And there are much clearer use cases for their products.
Tom: So you mentioned the various real-world applications for humanoids, logistics facilities, that sort of thing. But we know that they won’t be working in the U.S. because the Federal Communications Commission has banned the import of foreign-made humanoid robots, which is basically just a way of banning Chinese-made humanoid robots. What is the FCC so afraid of?
Savannah: To quote from the FCC’s notice, which not only includes humanoids, but also quadruped robots and all other mobile robots. That also includes robot vacuums.
According to the FCC, these devices contain unacceptable risks, quote, including that these devices could create supply chain vulnerabilities that could disrupt U.S. economic and national security and could create a cybersecurity risk that threatened American critical infrastructure. But while these existing models can still be sold in the United States, firms will need to apply for exemptions for any new models produced.
And that’s a major blow to robot makers with existing U.S. sales, like Unitree. Our colleague Noah Berman wrote last year about how Unitree robots are indeed being purchased by police stations in the U.S. and even some branches in the U.S. military, simply because they’re at a much lower price point and perform well. You can just go on Amazon and buy a Unitree quadruped very easily.
Unitree’s prospectus did disclose that the company earned more than 40% of its revenue from outside of China, and many of its largest buyers are overseas. So this new guidance is likely to have an effect not just on Unitree’s now revenue-boosting humanoid business, but also on many other embodied intelligence firms in China, even something as mundane as a smart robot vacuum.
Tom: Is it possible that humanoid robots are going to be over-deployed in China? For example, in many Chinese cities, you can go into a store and a humanoid robot will get an item for you, bring up the sale.
But when I was flying through Boston’s airport this summer, there was a much simpler and perfectly effective alternative. You walk into a store, put the items you want in a bag, then tap your credit card to exit the store. The exit gate sensor automatically adds up the cost of your items.
There’s no robot, but basically the same service. Doesn’t that kind of solution make a lot of sense?
Savannah: I do think that is the big outstanding question in this industry. Aside from some very niche industrial options and, of course, the robots that are being sent to universities or research labs where the science of it all is very exciting and interesting, the industry is definitely at a point where it needs to figure out where all these robots are going to go and what they’re going to do, especially important because they are about to start producing so much more.
We seem to be at the very early stages of a model that we’ve seen many times before in other parts of China’s manufacturing industry, for example, the EV industry. You have this boom of new technology, dozens or even hundreds of players rush in and start building the technology. It scales very, very quickly.
Suddenly, we have this overcapacity. Competition drives the price down, weeds out the weaker players, and then you’re left with just a few top players and a really clear niche and clear use case. We seem to be at the very beginning of that cycle with humanoids.
We’ll have all these robots going out into factories and logistic centers. Are the engineers going to be able to update them quickly enough so they’re actually useful workers, more useful than a human? There’s a lot that’s still unknown, and that’s to say nothing about the more specialized humanoids for health care or human companionship and friendship. Whether those will work out as anything more than a novelty is also very much up in the air.
This is definitely something that we will continue to be watching as we go forward.
Tom: I remember living and working in Beijing when the shared bike craze really took off. For a while, it worked very well, and then it just stopped working so well.
The bikes would break down. It just wasn’t as reliable, and I gave up on it. Then there were lots of pictures of just heaps of dead bicycles stacked up around various places in Beijing.
I hope it doesn’t come to that for the poor humanoids. That would look a bit grim. Thanks so much, Savannah.
For more on this story and others, please do go to thewirechina.com. Thank you for listening.
