US Builds AI Brains While China Dominates Robot Bodies
A new bipartisan report warns the US leads in AI software but is losing the physical AI race to China due to a massive deployment gap.
Policy · Source: Unite.AI
What happened
The National Security Commission on Robotics for Advanced Manufacturing just dropped a 90-page preliminary report. The verdict is brutal. America leads the world in building the brains of modern robotics. We dominate artificial intelligence, computer vision, and autonomy software. But we completely lack the scale to actually build and deploy the physical machines. The commission calls this physical AI. They argue that global leadership will be determined by productive deployment and resilient factory capacity. Having the smartest software models is no longer enough.
The numbers reveal a massive deployment gap. According to the International Federation of Robotics, Chinese factories installed 354,200 industrial robots in 2025. That is 59 percent of all worldwide installations. The United States installed just 38,400 units in the same year. China now operates over two million industrial robots across its manufacturing sector. That stockpile is more than five times larger than the American total of 393,700 robots. The US currently ranks eighth worldwide in robot density. We have roughly 307 robots per 10,000 manufacturing employees.
The report identifies a massive market failure blocking American progress. They call it a supply-demand paradox. Manufacturers are waiting for robotic systems to get cheaper, more adaptable, and highly proven. Meanwhile, robotics companies cannot lower their costs or improve their hardware without the massive deployment volume that manufacturers provide. Small and medium manufacturers make up 95 percent of the US market. They face severe barriers including high integration costs, limited financing, and major knowledge gaps.
Key facts
- 354,200 — Industrial robots installed by China in 2025
- 38,400 — Industrial robots installed by the United States in 2025
- 2 million — Industrial robots operating in Chinese factories by the end of 2024
- 12.1% — Percentage of US manufacturing plants reporting robot usage in 2021
- 1.9 million — Potential unfilled US manufacturing jobs by 2033
Why it matters
If you are building physical AI, software moats are not enough to survive. You can train the best vision model in the world. But if you cannot manufacture the chassis or find factories willing to test it, you lose. Most industrial robots installed in the United States are currently imported. Meanwhile, Chinese suppliers accounted for 55 percent of installations in their domestic market in 2025. American builders are relying on overseas hardware to house their cutting-edge brains. This creates a fragile supply chain and limits domestic innovation cycles.
The second-order effect hits the broader American industrial base directly. The US manufacturing sector faces a projected shortage of 3.8 million jobs by 2033. Up to 1.9 million of those jobs could go completely unfilled. Without aggressive robotic automation, American factories will fail to meet basic production demands. Furthermore, federal support remains highly fragmented. Government robotics programs are scattered across at least four executive departments. There is zero centralized accountability for national-scale outcomes. The National Science Foundation even ended the National Robotics Initiative in 2022.
For builders
Bridge the integration cost gap
Manufacturers desperately want automation but cannot afford the massive upfront integration costs. Founders who build robotics-as-a-service models or zero-integration deployment systems will win lucrative contracts. You will capture the 95 percent of US manufacturers that are small or medium-sized and currently priced out.
Target domestic hardware production
The United States relies heavily on imported robot components to function. Startups that can manufacture critical robotic parts domestically or provide cheap alternatives to overseas hardware will attract massive interest. Defense contractors and policy-driven investors will pay a premium for secure local supply chains.
Solve the liability and safety puzzle
Factory owners are terrified of cybersecurity risks, safety standards, and legal liabilities. If your physical AI product includes built-in compliance and bulletproof security frameworks, you instantly remove the biggest friction point. Enterprise buyers will choose your system over cheaper alternatives that carry higher systemic risks.
My take
We are entirely too obsessed with training larger foundation models while ignoring the physical world. Software eats the world, but robots actually have to build it. If we do not figure out how to manufacture cheap and reliable hardware at scale, America will just end up writing the code for Chinese machines.
Original reporting: Unite.AI. This is my rewrite and opinion.