Banning Chinese Humanoid Robots

Japanese matcha is available in cafés and shops from London to Los Angeles. What you cannot get, anywhere outside Japan, is the highest-grade ceremonial product, harvested in the first days of spring and stone-ground by the finest producers. The best matcha stays in Japan.

There is a version of this principle forming in humanoid robotics, and it may determine who gets what when the time comes.

The domestic imperative

In the last week of March 2026, a Chinese manufacturer called Agibot announced that it had shipped its ten-thousandth humanoid robot. In the same week, a bipartisan group of US senators introduced the American Security Robotics Act, a bill to ban the federal government from purchasing or using unmanned ground vehicles, including humanoid robots, manufactured by foreign adversaries.

Senator Schumer described it as a response to China "running their standard playbook, this time in robotics, trying to flood the U.S. market with their technology." The assumption embedded in the bill is that China is building humanoid robots to sell to the West.

The assumption may be incomplete.

China does intend to dominate the global humanoid supply chain. Its 15th Five-Year Plan places embodied intelligence in the same strategic tier as nuclear fusion and Chinese industrial robot exports grew roughly 65% per year between 2022 and 2024.

But China's population has now shrunk for four consecutive years. Its fertility rate has fallen to 1.09 births per woman, roughly half the replacement rate. By 2050, China is projected to lose 250 million people. The manufacturing base that made China the world's factory was built on young, abundant, affordable labour, and that workforce is ageing out faster than any policy intervention can reverse.

China is building humanoid robots because it is running out of workers.

The United States has its own version of this logic. In February, Elon Musk told the Dwarkesh Podcast that without AI and robotics, the US is "actually totally screwed" because of the national debt. He described his work at DOGE as addressing waste and fraud in the near term in an attempt to hold the fort until physical robots are ready, physical robots, develops. He called humanoid robots "the infinite money glitch."

As such, my take is that the two largest competitors in the humanoid race are each building for domestic survival.

China needs robots to replace a workforce that was never born. The United States needs robots to sustain an economy that cannot afford its current trajectory.

Both will export. But the matcha principle will apply. The domestic market will get the first tier. Everyone else gets what the producer is willing to sell.

Seller's terms

The conventional expectation is that humanoid robots will follow the pattern of most manufactured goods: competition drives down prices, supply expands, and the buyer chooses. That pattern requires surplus production capacity and competitive pressure among multiple sellers.

The conditions forming in this market look different.

The fastest manufacturer is building for a domestic demographic emergency. The second-largest programme is building for fiscal survival. Europe has one well-funded company with unproven production capacity. South Korea has a single concentrated bet through Hyundai and Boston Dynamics. Japan has the engineering base and has not yet committed to the market.

If demand arrives simultaneously from healthcare, logistics, manufacturing, construction, mining, elder care, agriculture, defence, and data centre maintenance, the supply question becomes acute. Who has units available? At what capability level? With what data governance? Running whose software? Countries that did not invest in domestic humanoid manufacturing may find the experience feels less like a purchase and more like an allocation. The terms set by the supplier. The regulatory constraints set by the government. The buyer operating in whatever space remains between the two.

The bill in Washington is being discussed as a China story. The more durable observation is simpler. The countries building these machines need them for themselves. What that leaves for the rest of us is not yet clear.