Why the Best-Selling Robot Can't Do Anything ｜ Unitree G1

These robots, Unitrix G1, are everywhere. From viral kung fu shows in China, to greeting their co-workers in Japan. From chasing wild boars in Poland, to becoming a monk in South Korea. Half the internet thinks they are ready to replace all human labor, the other half think they are useless and only built for viral clips. So what can they actually do? Let's start with the obvious dream, house chores. We've all imagined having a robot at home handling everything we don't want to deal with. Making dinner, watering the plants, watching the kids. And from the demos floating around for Unitrix G1, it looks like we're just months away from that reality. But we're really not. It might be hard to notice when these robots are performing on stage, but when they stand with the real person, you will soon realize they are kind of short, like really short. At 130 centimeters tall, the Unitrix G1 is much smaller than other humanoid robots, like Neo at 168 or Figure at 173. We are basically talking about the height of a 9-year-old kid. Even if they wanted to, they'll struggle to reach the kitchen counter or grab something off the top shelf. The G1 is this size for a reason, which we'll get to, but that reason has nothing to do with helping around the house. Size aside, the G1 is also missing some critical hardware. Take Figure 03 for comparison, it had 16 degrees of freedom hands and tactile fingers that can detect the weight of a paper clip. So like in this video, it knows whether it's holding a storage basket or a TV remote, and how much pressure is needed to press a power button. Quite impressive, actually. The Unitrix G1, by contrast, uses non-functional placeholder hands on the base model, and even if you buy the upgraded hands separately, they still don't come close to what Figure offers. But the hardware gap is actually the easier problem to fix. The more fundamental issue is the brain. It sounds simple, but doing housework requires a really intelligent robot. AI can beat a human really easily in Go, but all those daily tasks that feel effortless for humans, like folding a shirt, closing the door of the dishwasher, are extremely difficult for robots. Even something basic like picking something off the ground, for the remote control NEO, this would take a minute. Or minutes. You get the point. Companies are making progress here. Figure's Helix model, for example, can efficiently make a bed and coordinate in real time with another robot, even nodding at each other before they start, which is really impressive. Unitrix do invest in this area too, but nowhere near the same level. In the first three quarters of 2025, their total research spending is only $12 million. That's less than one-tenth of Figure 8's spending. There's also a computing problem. Running something like the Helix model demands serious processing power that G1 simply cannot support it. The base model heavily relies on remote control. The higher-priced EDU version can run local models, but the computing power compared with Figure 03 is like a Nintendo Switch versus a decent gaming desktop. So home use is pretty much off the table for now. What about factories? There's another problem there. Payload. The G1 has a frame of a nine-year-old, but it's actually weaker than one, carrying only three kilograms per arm. Compare that to Optimus at 10 kilograms, or Atlas at an impressive 30 kilograms. That rules out a huge range of industrial work, like carrying parts around the floor, or just holding heavy tools, because an industrial drill alone could be heavier than three kilograms. On top of that, the battery only lasts around two hours, and thermal management isn't great either, meaning it has to pulse very frequently just to cool down. So if G1 cannot really do housework, and it's not quite ready for factory floor, then what's the point? It is just a robot designed to dance and nothing more. It is just a robot slop, like Sora and the AI slop videos. Well, first of all, dance performance and entertainment could be counted as use cases for robots, and the G1 can do it extremely well. Remember they are too short to do housework? That's actually on purpose. A smaller frame and a lighter body reduces the moderate density you need and make it a lot easier to do quick movements, flips, and acrobatics. The G1 can even fold itself down to fit inside a large suitcase, which makes logistics surprisingly simple. That's why beyond kung fu shows, these robots are showing up in concerts and theme parks too. But bigger reason for their popularity is price. Most high-end humanoid robots don't really sell to public. Even if they do, it would be a crazy price tag. Atlas is estimated to cost around 150k. So Unitrix G1 has a starting price of 13.5k. Even more impressive, they still have a more than healthy margin of over 60%. For a hardware company, that's almost unheard of. Apple, one of the most profitable companies on earth, averaged around 40% overall. Unitrix achieved this by designing and manufacturing roughly 80% of their key components in-house, which cut costs down significantly. So Virality has also created an entirely new ecosystem around the robots. A company called Botshare now rents robots out for performance and personal use, making short-term use even cheaper. All of this has made Unitrix so dominant in the entertainment space. But entertainment is not all they can do. They sold 5,500 units in 2025. That doesn't sound impressive, but all American companies combined in 2025 delivered only a few hundred. If you look at where all the units were sold, actually only 17% went to entertainment. And over 70% of the sales, surprisingly, were for research and education. So what is research and education usage, and why do they need so many humanoid robots? Well, AI has unlocked huge progress in robotics research, and the appetite for actual humanoid hardware to test on has exploded. Researchers from different institutes, like Amazon working on autonomous movement algorithms, or UC San Diego working on trials of remote surgery, they all need a real robot to test their work. The G1 is one of the very few available humanoid robots you can actually buy. You don't have many other options unless you really want to buy all the parts and build one yourself. Unitrix was also well-positioned here because their robot dogs, their main product before the humanoid line, were heavily adopted by universities and research institutes. So why haven't American companies target the same market? A few reasons. First one is, the research market is really niche. There are only 2,600 universities in the US. Compare with 135 million households. And even if American company tried, Unitrix has a massive cost advantage from China's manufacturing supply chain. For companies like Tesla and Figure, it makes much more sense to focus on the bigger price, that would be household and industrial applications. Interestingly, even Chinese companies and investors were skeptical of the research market. China's startup culture is fiercely competitive. When food delivery apps first launched, over 100 platforms were fighting for market share almost immediately. Yet the G1 was announced in May 2024. And two years later, you still don't see much competition for Unitrix G1. That tells you something about how unattractive investors consider this research market to be. And yet, Unitrix quietly build their business there anyway. To their credit, they've served the market well. The EDO version has an open architecture. They even have an app store where you can buy and sell certain movement routines, like a Michael Jackson dance sequence. At its core, Unitrix's success comes down to identifying an underserved niche market, building a product that fits it well enough, and at a price nobody else could match. In a sense, it was also the practical choice. The founding team didn't have a prestigious background or large founding rounds, so they cannot really compete with Silicon Valley giants like Tesla, even if they want to. Whether they can sustain that momentum and eventually break into broader use cases is a real question. And only time will tell. But for now, at least, it looks like they'll have a successful IPO in China, raising roughly 600 million dollars. If you want to see how Unitrix robots were beaten so badly by another company in the robot marathon, I have another video here for you. Otherwise, thanks for watching. I'll see you in the next one.