China Is Iterating Faster Than We're Thinking

I just returned from a few weeks in China, across Beijing, Shanghai, Hangzhou, and Shenzhen, where I met with few tens of humanoid and robotics builders, operators, and supply-chain vendors. Since coming back, I've been thinking about the country in a quieter, more reflective way than I expected.

Not in headlines. Not in geopolitics. Not even in competition.

Just in intuition.

There's a particular texture to being inside a place that has decided it will build something at scale. You can sense it before anyone says it out loud. It's in the density of suppliers within a few blocks. It's in how quickly a conversation moves from concept to torque specifications. It's in the way iteration is assumed, not debated.

What stayed with me wasn't just the speed. It was the lack of hesitation.

In the U.S., when we talk about humanoids, there's still a layer of abstraction. Research papers. Demos. Vision-language-action models. World models. We are exceptional at cognition. We stretch the frontier of what is thinkable.

In China, I felt something else: manufacturability as instinct.

Factories that once made consumer electronics are pivoting toward robotics without existential angst. Supply chains are not theoretical — they are spatially real. You don't wait weeks for a custom component; you walk across the street and negotiate a modification. The distance between idea and embodiment is physically shorter.

That proximity changes psychology.

When you ask what's possible in three years, the answer isn't philosophical. It's framed in deltas: payload increases, actuator density, cost curves at 10,000 units. Humanoids are discussed as product lines. SKUs. Margins. Export markets.

It feels industrial, not aspirational.

And yet, for all the velocity, there are fragilities. Reliability is not solved. Dexterous manipulation remains temperamental. Tactile sensing is improving but not yet intuitive. Scaling autonomy is still bridged by teleoperation and structured human oversight.

But even that bridge is strategic.

There is a seriousness about compressing the learning loop, about generating embodied data at industrial scale. Not as an academic experiment, but as an economic lever.

It made me think about where value will actually accrue.

If hardware becomes abundant, if the marginal cost of a capable humanoid falls faster than expected, then the center of gravity shifts. The moat isn't in assembling metal and motors. It's in reliability. Deployment. Compliance. Integration into real institutions with real constraints.

The West may continue to lead in cognition, in the architecture of intelligence itself. China may continue to dominate manufacturing density and iteration speed. The enduring layer may sit between them: the infrastructure that makes robots trustworthy in messy, regulated, human environments.

I didn't come back feeling alarmed.

I came back feeling clarified.

We are exiting the era where humanoids are speculative artifacts. They are becoming industrial products. Once that transition is internalized, everything else reframes.

The question is no longer whether robots will exist at scale.

It's where the seams of power will settle when they do.

And that feels less like competition, and more like the beginning of a new equilibrium.

Updating to include a brilliant read "Breakneck: China's Quest to Engineer the Future" by Dan Wang.