The argument about whether humanoids have a real job usually plays out on a stage, with a script and a crowd. The more useful test is a shop where nobody is watching. I spent two hours in a drugstore in an ordinary part of Shanghai, standing next to a robot that was filling online orders. Customers walked past it. Staff walked past it. Nobody looked up.
The machine comes from a robotics company inside Ant Group, the group behind Alipay. Founded in March 2025, barely eighteen months old, and it has just raised 210 million dollars with more rounds expected this year. It also bumped into me while I stood in its way, which tells you roughly where the technology sits today.
The Money Goes Into the Brain
The hardware still looks like a prototype, panels open on the side. That is the plan rather than an accident: build it now, make it pretty later, keep the data from every awkward run.
What the funding pays for is the software. Six versions of the control system in the past few months alone. The job on the floor is deliberately narrow: an order arrives online, the robot finds the item on the shelf, carries it to the front, and in the target version packs it. None of that is impressive to watch. All of it is measurable, which is the point. And the eventual case is not a Shanghai drugstore, it is the rural county where filling a prescription already takes half a day.
Two Humanoids Doing Paid Work at CATL
CATL is the biggest battery maker in the world. Two humanoids from two different suppliers have been working in its plants for months, each on one job.
At the Dongguan plant, a robot from a young Hangzhou company sitting on the Alibaba campus stands at the end of the battery pack line. Before a pack leaves the building it gets a final function test, and someone has to plug in the connectors at several hundred volts. That is the task the robot took over. It has been there since December, close to a thousand packs, zero faults. The connector goes in correctly more than 99 times out of 100, and throughput at that station is three times what it was.
At the Ningde plant, a second machine called the S1 moves and picks material on the module and pack lines. Fifty kilos with two arms, eight hours on one charge, running on a CATL battery. Signed off in March, around the clock since, and in June CATL and that supplier signed a global partnership, so this is meant to travel to other plants.
Both robots got one station and one number. So the useful question is not whether to buy a humanoid, it is which station in your building repeats all day and already gets measured. We are running exactly that with a German automotive supplier now: four tasks picked on their own floor, the robot training in Shenzhen, then further training in the plant here in Shanghai.
Read the Unitree Investor List Twice
On August 6, Unitree published the strategic investors for its listing. Nine names, eight million shares, about twenty percent of the offering.
DeepSeek is one of them, with 933,000 shares for 141 million yuan, locked for 36 months. An AI lab whose entire business lives on a server is paying to sit on the shareholder list of a company that builds legs and arms. That is a bet on where its models run next.
Then look at the rest. Tencent came in through its Shanghai investment arm with 136 million yuan. The investment arm of the state oil group put in the same amount, as did the finance arm of the power grid operator for southern China, and China Telecom's capital arm is on the list too. Oil, power and telecom own the work nobody wants: climbing a pylon, walking a pipeline, checking a substation at two in the morning.
That list is a customer list with a share certificate attached.
One correction to the previous episode. The final issue price came in at 150.8 yuan per share, not the 104 that was being passed around, which puts the company near 8.5 billion dollars rather than six. Retail demand was the other headline: 9.8 million private investors applied on August 10, the offer was covered more than 8,000 times, and 0.018 percent of applicants got any shares at all.
The Shenzhen Pattern, Running Again
Nine out of ten consumer 3D printers sold anywhere in the world now come from four companies, and all four sit in Shenzhen. China exported 3.6 million of them in the first half of this year, ninety percent more than last year. Ten years ago those names were American and European. The reason for the switch is boring and decisive: parts, factories and engineers within a two hour drive of each other. It is why most robots bought this decade will come out of Shenzhen, Suzhou or Hangzhou.
What None of This Proves
A thousand packs at one station is not a fleet, and the pharmacy robot is a prototype that walked into me. Nothing here shows that a humanoid pays for itself across a whole plant, and anyone selling that story is ahead of the evidence.
What has changed is the entry point. A Unitree G1 sells for 14,000 dollars, which is a purchase order rather than a research grant. The open part is whether these single station numbers survive the second and third station. Nobody has that answer yet, including the companies running the first one.
[00:00] China is putting robots into real shops now. I stood in a pharmacy in Shanghai and watched one work. It comes from a company called Ling Bo and Ling Bo belongs to Ant Group, the company behind Alipay. And what they are after is the brain of the robot. I will take you in there in a minute. Story number two takes us into a factory. The biggest battery maker in the world already has humanoid robots working on its lines. Two different suppliers, two different jobs, and one of them has been doing [music] it since December. After that, we look at who is putting money on the table. Deep Seek just bought into Unitree and locked itself in for 3 years. And on the same list, you find China's oil group and the power grid operator for the south of the country.
[00:47] Then I close with a quick round of short news. Nine out of 10 3D printers sold to consumers worldwide now come from companies in one Chinese city. A Chinese robot taxi is going to Denmark, and there are two robot events this month that everyone in this [music] industry is flying to, and so am I. And stay until the end. I have another cute robot clip for you. Every week, I go through these Chinese sources, the one nobody translates for you, and I bring you the things that matter for your business, often weeks before they show up in the English press. And why me? I live in Shanghai, I speak the language, I read the prospectuses and the industry reports in Chinese, and I stand on these factory floors myself. So, over the past year, I have taken more than 100 CVs,
[01:37] board members, and founders into robotics companies here in China. From the names you know, like Unitree, to the suppliers nobody has ever heard of in the west. Someone asked me, which robot works? The answer [music] comes from what I saw on the floor, not from a marketing video or a press release. If you want to walk one of these floors yourself, the link is in the description. Just write me a message. AsiaBase. This story is about a company some of
[02:35] you might be familiar with. This is the Robby Ant, which is built by Antling Bohr, which is a mother company of Ant Financial, and the most famous product of them is Alipay, which is one of the biggest payment providers here in China. And this company just raised 210 million US dollars, and they will have a couple of more rounds this year. So, the fundings are crazy in China, and let me tell you something, this company was founded last year in March 2025, so it's not even 18 months old. So, and today I want to show you directly
[03:22] where this money goes. I am here in a drugstore in the middle of nowhere in Shanghai, and we see a robot here doing work. So, the money goes directly into this hardware, but mainly in the software. So, what Antling Bohr is building is the brain of the robot. They want to use this money to develop a really good brain, and they released six versions of the brain in the last months and we see the job he is doing is somebody order some something online and he gets the task to find it. So, what do we have here? We have here some drug that it needs, but I'm maybe
[04:11] disrupting his work and he is now looking where this drug is and then he will get it and bring it to the cashier. And the final goal of this robot is that he does all the work himself. You see, okay, he got it. Then he brings it to the front and in the end he does the packaging. If this works 100% we all know this makes so much sense. If you live in a rural area in Europe or in the US and um you need to get some medication, you need to get some pills, we know nowadays it's very hard to get it. So, this is something that in a couple of years we will see these robots do. And now we are
[05:00] in the prototype phase. We see it. There's a lot is not finished yet. You see we have some open parts here. What they want to do, we innovate first and then we do the perfect version later. But all the I just got hit by a robot. So, all the stuff we learn now will be very important. Oh, oh, he just hit me. >> [laughter] >> So, all the data we gather now, all the learnings we gather now, will be very valuable in the future. So, it's very important to understand how far they are already with these robots here. They part of everyday life and I spend two hours in this drugstore and the customers here, the employees, nobody
[05:49] cares about it. So, everyone is very open to technology. Seeing a robot is nothing special anymore in China. You also should understand what is happening already and that this is a serious business because if you don't do, your competitors will do. This is a story of Ant Ling Bul. We will continue to report from the ground, so be sure to leave a comment down below, give us a like, and subscribe to this account. Now, let's move from the pharmacy into a factory. CATL is the biggest battery maker in the world. If you drive an electric car, there's a good chance the battery inside it came from there. And in their plants, humanoid robots have been doing paid work for months now. Two different
[06:37] companies on two different jobs. At the Dongguan plant, there is a robot called Moss, built by a company called Spirit AI. I visited them last month with a German automotive supplier. It's a young company sitting in Hangzhou on the Alibaba campus, and their focus is the brain. Their robot stands at the end of the battery pack line. Before a pack leaves the building, it gets [music] one final function test, and someone has to plug in the connectors for that test. It is several hundred volts, and that is the job the robot took over. At the Ningde plant, there is a second one, the S1 from Gelbot. It moves and picks material on the module and pack lines. 50 kilos with two arms, 8 hours on one charge. And the battery inside that
[07:26] robot is also a CATL battery. Now, let's have a look at the numbers because that is where this got interesting for me. Moss has [music] been on that line since December, close to a thousand battery packs, and zero faults. The connector goes in correctly more than 99 times out of 100 and the output at that station is three times what it was. The S1 from Gelbort was signed off in March and has been running around the clock ever since, three months straight. Then in June, CATL and Gelbort signed a global partnership. So, this is meant to travel to other plants and maybe also worldwide. Here is what I take from this. Both robots got one station each, one task with a number on it. That is
[08:14] how these machines walk into a plant, one station then the next one. So, if you run a factory, the useful question is which station or which [music] task in your building repeats all day and gets measured. That is where this starts and it's is much smaller decision than buying a humanoid or a humanoid fleet. So, we are doing exactly that right now. We put a German automotive supplier together with a Chinese robotics company. We went to the plant, picked four tasks and the robot is being trained on those four tasks [music] in Shenzhen at the moment. Once it has enough data, it comes to the plant here in Shanghai and keeps training on the floor. And if it hits the same numbers as the robots at CATL, they roll it out globally. In my opinion, that is the way
[09:03] to do this. One clear job, one number and from there you spread it. Now, I'm curious about your factory. Which station in your building would you hand over first? Write it in the comments. I read them and [music] I will answer them in the next video. This is my favorite story of the week. [music] Remember the deep seek shock last year? A Chinese AI lab that nobody outside China had on the radar put out a model that knocked the wind out of the US tech market for a week. That company has now taken its money and put it into a robot maker, and it [music] cannot sell a single share for 3 years. Here is what happened. On August 6th, [music] Unit republished the list of strategic investors for its listing. We have nine names, 8 million
[09:52] shares, about 20% of the entire offering. Deep Seek is one of them. 933,000 shares, 141 million yuan, which is around $20 [music] locked for 36 months. So, sit with that for a second. Deep Seek writes models. Their whole business lives on a server, and they are paying to sit on the shareholder list of a company that builds legs and arms. That is a bet on where their models run next, and I think it is a very smart one. Now, let's have a look at who else is on that list. We have Tencent, WeChat, the app a billion people here use for messaging, payments, and pretty much everything else. They came in through their Shanghai investment [music] arm with 136
[10:41] million yuan. Kunlun Capital, the investment arm of the Chinese state oil group, same amount. Then, we [music] have the finance arm of China Southern Power Grid, the company that runs the electricity network for the south of the country, [music] same amount. And finally, we have China Telecom Capital. So, read that list again. We have oil, power, telecom. These are the companies that own the work nobody wants to do. Climbing a pylon walking a pipeline, checking a substation at 2:00 in the morning. And it has already started. I saw videos of robot dogs and drones inspecting power lines [music] and substations in China. That list is a customer list with a share certificate attached. [music] Now, a correction from the last episode. So,
[11:29] the final issue price came in at 150.8 yuan per share. In the last episode, I used the estimate that was going around about 104, and I said the company would come out at around $6 billion. At that price, the company comes out at about 8 and 1/2 billion dollars, and it raised [music] roughly 850 million for 10% of itself. And these numbers are still fluctuating. And the interesting thing is ordinary people wanted [music] in 9.8 million private investors applied for shares on the 10th of August. 0.018% of them got any. The offer was covered more than 8,000 times over. Think about that number for a second. 9.8 [music]
[12:18] million people trying to buy a piece of a robot company. So, why does it matter for you as a founder or company owner? Take an hour and walk your own operation. Mark every task where a person goes somewhere unpleasant to look at something. Tank inspections, pipe runs, roof [music] checks, anything at night, anything where you need a second person there for safety. That is the list worth making this week. We have a project running with a large German specialty chemical company right now. They use robot dogs to inspect their sites. Same logic as the battery plant. One task that repeats, one number to hit, and the machine that does not mind where it has to stand. And if you want this before it reaches the English press, I write it up every week [music]
[13:06] in our newsletter. Link is in the description. Before we wrap up, four quick ones. Let's start with 3D printers. Nine out of 10 sold to consumers anywhere in the world come from companies, and all four of them sit in Shenzhen, where I will go tonight, actually. [music] China exported 3.6 million printers in the first half of this year, 90% more than last year, and the value of those exports more than double. One of the four is Bambu Lab, started by a former DJI engineer. 10 years ago, the names in the consumer 3D printing [music] were American and European. Today, four companies in one Chinese city have the whole market. I think [music]
[13:53] that is a preview for robots. Most of the machines you will buy this decade [music] will come out of Shenzhen or Suzhou, Hangzhou, and a handful other cities here. [music] And the reason is the same or one that moved 3D printing. The parts, the factories, and all the engineers are within a 2-hour drive of each other. From printers to cars. [music] A Chinese robot taxi is coming to Denmark. WeRide signed with Green Mobility, a Danish car-sharing operator, running more than 1,500 electric cars in the Nordics. Public service is planned for the first half of [music] 2027 with the Danish road authority sitting at the table. Denmark is their [music] sixth market in Europe this year. Spain and Slovakia came before. If you're in Europe, watch this one. It arrives through your local operator with no
[14:41] Chinese brand on the door. Staying with cars for the next story. Chery exported [music] 200,000 vehicles in July alone, 70% more than a year ago, and the fifth monthly record in a row. No Chinese car maker [music] has crossed that line before. that in a robotics show for one reason. The same factories, the same suppliers, and the same engineers are the [music] ones building robots now. And one for your calendar, August is the loudest robot month of the year here in China. The Embodied AI conference in Shanghai is happening right now, and from the 19th [music] to the 23rd, the World Robot Conference in Beijing is happening. Over 300 companies [music] and more than 150 products shown in public for the first time. I will be there for all of 5 days
[15:30] walking the floor and talking to the people who are building these machines. So, whatever comes out of that week, you will get it here first. And hit me up if you're there, maybe we can meet on site. So, let me zoom out for a second. Four things this week and they all point the same way. A robot working in a pharmacy in Shanghai. Two robots doing paid work in the biggest battery plant on Earth. An [music] AI lab tying its money to a robot maker for 3 years. And four companies in one Chinese city holding the consumer [music] 3D printer market. Every one of those is either a machine already doing the job or money betting that it will. The part that matters for you is timing. A Unitree G1 sells for $14,000. That is a machine you can put
[16:21] on a purchase order, not a research grant. When a price sits at that level, the question in your company moves from whether to when. And whoever starts this year will have 2 years of data on their own task before their competitors have picked a supplier. So, here is my one ask. If you run a factory, a warehouse, a hospital, or a network of sites and there's one task you would hand to a machine tomorrow, write me. Tell me your industry and tell me that task. I will tell you straight whether a trip to China makes sense for you this year or whether you're better off waiting. I take groups into these companies almost every week and I would rather tell you no than sell you a trip you don't need. The link to contact me is in the description. On the 19th, I'm in Beijing
[17:10] at the World Robot Conference all 5 day. Whatever comes off that floor, I told you, you get it here first. So, be sure to subscribe, give me a like, and a comment. And here is your clip for the week. Somebody parked a unitary humanoid on the sidewalk, speaker next to it, payment code on the ground, the robot sits there and begs. So, every robot in this episode has a job. This one >> [music] >> went freelance. See you next week with Asia Bits Robotics News.
Get These Insights Every Monday
Join 18,000+ professionals reading Asiabits. Free, every Monday, straight from Shanghai.
Subscribe Free →
