Figure and 1X Are Making a Lot of Humanoids. But For Who, Exactly?

You might be wondering why this matters. Here's the thing: humanoid robotics is at a weird inflection point. The hardware is getting genuinely impressive. IEEE Spectrum recently highlighted Boston Dynamics' Atlas moving a mini-fridge, using whole-body control and reinforcement learning to handle the weight and inertia. Agility's Digit is learning to deadlift 65 pounds. Unitree is showing off voice-commanded real-time control. The robots are doing things that would have seemed impossible five years ago.

But doing impressive things in controlled demos is not the same as being useful in someone's home or warehouse. Figure released a video of one of their robots tidying a bedroom, and while it's cool, it also kind of illustrates the gap. The robot is clearly comfortable with certain movements and clearly not with others. It's methodical in a way that a human wouldn't be. That's fine for a research demo, but is it ready for a customer who paid actual money?

There's also the question of what "commercial use-case development" actually means in practice. If you look at where robots are genuinely delivering value in warehouses right now, it's mostly not humanoids. Companies like Berkshire Grey are deploying systems that look nothing like people but move packages efficiently and reliably. The humanoid form factor makes sense for environments designed for humans, but those environments are also the hardest to operate in. Homes are cluttered, unpredictable, full of weird edge cases. Factories can be adapted to robots. Your living room can't.

I talked to a few people in the industry about this (off the record, because nobody wants to be quoted being skeptical about the humanoid hype), and the consensus seems to be that we're in a "build it and they will come" phase. The theory is that scale will fix everything. Make enough robots, collect enough data, train better models, and eventually the use cases will emerge. It's not crazy logic. It's how a lot of technology development works.

But it's also not guaranteed to work. And it requires a lot of capital and a lot of patience from investors who may or may not have either.

What I find genuinely interesting is the vertical integration play that 1X is making. Building everything in-house, from motors to final assembly, is expensive and hard. But it also means they can iterate faster and control quality in ways that companies relying on suppliers can't. If humanoid robotics is going to become a real industry, someone needs to figure out the manufacturing side. Maybe that's what 1X is actually doing, less "building robots for consumers" and more "building the capability to build robots at scale." That would make the 2026 consumer shipment timeline less of a promise and more of a target.

Figure's approach seems different, more focused on software and learning. Their bet appears to be that the hardware is good enough, and the real challenge is training robots to do useful things. Hence all the data collection. Hence the 55 robots per week.

Honestly, I'm not sure which approach is right. It's too early to say. Both companies are making big bets with limited evidence that the market they're building for actually exists. That's either visionary or reckless, depending on how things shake out.

What I do know is that the humanoid robotics industry is about to get a lot more crowded. Between Figure, 1X, Unitree, Agility, and a dozen other companies I haven't even mentioned, there are going to be a lot of humanoid robots in the world very soon. Whether anyone outside of research labs and corporate innovation departments actually wants them remains unclear. But we're going to find out.

James Chen · Yesterday · 4 min