Why the West Won’t Win the Humanoid Race

Pick almost any humanoid robot you’ve seen go viral in the last year and it probably came from a Western lab. Tesla’s Optimus tightening bolts. Figure’s robot sorting parts at a car plant. Agility’s Digit hauling totes. Boston Dynamics doing backflips. The demos are slick, the funding is enormous, and the story writes itself: this is the West’s moment in physical AI.

Then you look at who actually ships. Chinese firms put roughly 13,000 humanoids into the world in 2025, taking more than four fifths of the market, while US makers shipped a few hundred units between them. That gap isn’t about vision or talent. It’s about a supply chain and a set of customers that sit almost entirely on the other side of the planet.

That’s the uncomfortable argument Patric laid out on the show. Not that Western robots are bad, but that the game is being played on a field China built.

The cluster is the moat

Robots are a supply-chain game before they’re an AI game

Strip a humanoid down and most of its bill of materials is mechanical: motors, harmonic drives, actuators, precision bearings, dexterous hands, batteries, and sensors. Silicon is a small slice. And China makes the mechanical stuff at a scale nobody else can touch. It handles roughly 90% of the world’s permanent-magnet processing and holds big shares of precision bearings, motors, and power electronics. Actuators alone can be about 70% of a robot’s cost, and China controls a majority of that component base.

The reason is boring and hard to copy. All of those parts sit right next to China’s electric-vehicle and electronics factories, clustered so tightly that a designer can go from drawing to production in days, inside a couple hundred kilometers. One analysis found that building Tesla’s Optimus without Chinese suppliers would push its parts bill from around $46,000 to roughly $131,000. Same robot, triple the cost, just for sourcing it somewhere else.

This is Patric’s Detroit-and-Stuttgart point. Industries win where the parts, the factories, and the people already sit shoulder to shoulder. Cars clustered in Detroit and southern Germany for a reason, and once that gravity forms it’s brutal to shift. For robots, that cluster is in China, and it wasn’t an accident. It was decades of industrial policy.

Demand-pull beats supply-push

China has customers, the West has demos

Here’s the other half of the problem, and it might matter more. Most Western humanoid companies are building an impressive machine and then hunting for a job it can do. That’s supply-push. China is doing the reverse: the state is actively pulling robots onto real factory floors. Power giants, oil majors, and steelmakers already run robots for inspection and hazardous work, EV maker Zeekr has humanoids on its lines, and Beijing has named embodied AI a national priority, backed by a state fund reported at $138 billion.

Demand plus low prices creates a flywheel. Unitree lists one humanoid at $5,900 and another near $13,500, well under the $20,000-plus most Western rivals can manage. Cheap robots get bought, bought robots generate real-world data, and that data makes the next version better and cheaper. Over five years China filed more than 7,700 humanoid patents to the US’s roughly 1,500. The lead isn’t just in units shipped. It’s compounding.

Where the West can still win

The brain, and the jobs humanoids are actually good at

Now the pushback, because “the West won’t win” is too clean. China’s edge is hardware, and by its own analysts’ account the software, the actual intelligence, still lags the machine. The frontier-AI money sits in the US, where private AI investment ran about $109 billion in 2024 against roughly $9 billion in China, and export controls keep the best training chips out of Chinese hands. The likely outcome isn’t one winner. It’s a split: Chinese bodies, Western brains, with a running fight over which layer captures the value.

There’s a deeper question hiding under all of this, too. Maybe the humanoid shape is the wrong bet entirely. On a real jobsite, a two-legged generalist loses to a specialized machine that does one narrow task extremely well, over and over. As Patric puts it, a biped is a peak performer at nothing. Humanoids look most useful in controlled spaces like warehouses and factories, plus a few edge cases such as elder care and tactical work, not as the core workforce anywhere messy.

So if humanoids aren’t the endgame, “winning the humanoid race” matters less than the headline suggests. The team that owns the right specialized robots, and the software running them, may quietly win the war while everyone watches the walking robots.

What this means if you build or buy

Plan for a world where the hardware is Chinese

For founders, competing head-on with the cluster on hardware cost is close to suicidal. The defensible plays are the software and autonomy layers, verticalized applications tuned to one industry, and purpose-built machines for a specific scope where you can out-execute a generalist. For contractors and operators, expect your first truly useful robots to be specialized, and increasingly Chinese-built or Chinese-supplied, which makes procurement, security, and spare-parts support real planning questions rather than afterthoughts.

For investors, the move is to separate category excitement from supply-chain reality. A cluster advantage this deep is hard to underwrite against, and a slick Western demo doesn’t change the parts bill. Watch who ships, at what cost, into what actual demand. That’s the scoreboard that counts.

Related Articles
Why Most Robotics Pilots Die on the Jobsite
Construction Robotics in 2026: 4 Workflows Where Robots Are Actually Working on Site