U.S. Humanoid Robotics Has a Scale Problem, Not an Ideas Problem

What ITIF Is Arguing

The report's central point is blunt: the United States helped invent modern robotics, but invention is not the same as industrial leadership. In humanoids, ITIF argues that the U.S. market is producing compelling companies while China is producing unit volume, supplier depth, lower-cost hardware, and a policy environment built around adoption.

The unit comparison is the clearest part. ITIF cites Unitree at 5,500 G1 robots sold in 2025, AgiBot at 5,168 A2 robots, and UBTECH at 1,000 Walker robots. It contrasts that with about 150 robots each from Figure AI, Agility Robotics, and Tesla. The result is an estimated 12,868 units for the named Chinese companies against 450 for the named U.S. companies.

Those numbers should be read carefully. They refer to sales, not necessarily production capacity, long-term deployment quality, revenue per robot, autonomy rates, or uptime. Still, they matter because robots improve through repetition. Higher sales can mean more supplier orders, more field data, more customer feedback, more manufacturing learning, and more pressure to bring down cost.

The Sales Gap

Humanoid robotics is still early enough that every sales number deserves skepticism. A low-cost education platform, a research robot, a teleoperated demo fleet, and an industrial robot under a paid automation contract are not the same thing. But early volume has strategic value even when the robots are imperfect.

China's advantage is that low-cost platforms can spread fast. Developers buy them. Universities buy them. Systems integrators experiment with them. Factories test them. Videos and benchmarks multiply. Suppliers get larger orders for actuators, batteries, cameras, sensors, reducers, hands, and controllers. That cycle is messy, but it creates more chances to learn than a tiny fleet of high-cost units locked inside a few private pilots.

This is where the debate becomes more practical than ideological. The United States can point to better-funded startups, stronger AI labs, elite universities, and large end customers. China can point to volume, price pressure, fast supplier iteration, state-backed adoption priorities, and growing export ambition. In robotics, both sides matter. A brilliant robot that cannot be manufactured or bought at scale loses ground to a rougher robot that improves every quarter through field use.

AI-generated image

Humanoid competitiveness depends on the parts stack as much as the walking demo: actuators, sensors, compute, batteries, hands, safety systems, and manufacturing process control. Source: Biped.News AI editorial illustration.

Why Adoption Policy Matters

ITIF places heavy weight on national strategy. The argument is that countries with strong robotics policy create the first market for their own manufacturers. Subsidies, public procurement, tax incentives, standards work, training programs, research funding, and adoption targets can all pull robots out of the lab and into factories, warehouses, hospitals, utilities, and public agencies.

That matters because buyers are conservative. A plant manager will not buy a humanoid because a demo went viral. They need a task that fits the robot, a safety case, a maintenance plan, insurance comfort, worker training, support coverage, uptime targets, and a financial model. Policy cannot solve all of that, but it can reduce the friction that slows adoption.

The United States has pieces of the puzzle. It has the National Robotics Initiative legacy, defense robotics spending, advanced AI companies, large logistics networks, world-class manufacturers, and startups that can raise huge rounds. What it lacks is a single, durable robotics strategy that connects research, manufacturing, adoption, exports, workforce, safety standards, and procurement into one plan.

The Cheap Robot Advantage

The uncomfortable part for U.S. companies is price. ITIF points to earlier work arguing that many Chinese robots are cheaper while good enough for customers to buy. That pattern has already appeared in drones, solar hardware, batteries, electric vehicles, and consumer electronics. Robotics may follow a different path because reliability, safety, and support are harder, but the economic pressure is familiar.

Cheap robots do not need to be best-in-class to shape a market. They need to be available, serviceable, hackable, and cheap enough for developers and buyers to take risks. Once thousands of units are in circulation, the ecosystem starts to form around them. Training data, spare parts, code examples, repairs, third-party tooling, reseller channels, and customer expectations all attach to the installed base.

That is why the U.S. response cannot be only "build better robots." It also has to be "build more robots, with more domestic demand, at lower cost, with clearer deployment rules." The industrial policy debate around humanoids will probably sound less like AI model policy and more like semiconductor, battery, and advanced manufacturing policy.

What the Report Gets Right, and What Still Needs Proof

ITIF is strongest when it separates invention from scale. U.S. companies can produce some of the most impressive humanoid demos in the world and still lose the adoption race if the robots remain expensive, scarce, privately tested, or dependent on a thin supply chain. Physical AI will be judged in factories and warehouses, not only in model benchmarks.

The report is also right to connect humanoid robotics with broader industrial competitiveness. A country that can build and deploy robots at scale gets more than robot revenue. It gets productivity gains, manufacturing process knowledge, component companies, maintenance jobs, data infrastructure, export leverage, defense options, and bargaining power in future automation standards.

The open question is how much the 2025 unit gap predicts long-term leadership. Some Chinese sales may be lower-end developer or education units. Some U.S. companies may be deliberately avoiding low-margin sales while they focus on enterprise reliability. Agility's warehouse strategy, Figure's manufacturing partnerships, and Tesla's internal production ambitions cannot be reduced to a single shipment table.

Even with that caveat, early volume has gravity. If the U.S. waits until the market is mature before treating humanoid robotics as a strategic industry, it may find that component supply, customer habits, developer mindshare, and export channels have already formed elsewhere.

AI-generated image

The next robotics race may be decided by deployment systems: procurement, safety validation, service networks, fleet monitoring, and operator training. Source: Biped.News AI editorial illustration.

Frequently Asked Questions

The Bottom Line

The July 14 ITIF analysis lands because it reframes the humanoid robot race around industrial capacity. The issue is not whether the United States can produce impressive robots. It can. The harder question is whether it can produce enough robots, deploy them widely enough, lower their cost fast enough, and build the supplier base underneath them.

China's early unit lead does not settle the race. The robots are still immature, the categories are uneven, and most commercial deployments remain constrained by task, supervision, cost, and safety. But the lead is a warning. Robotics is one of those sectors where field experience compounds. The country with the most robots in circulation often gets more data, more suppliers, more buyers, and more pressure to improve.

For U.S. humanoid companies, the next proof point is not another viral demo. It is repeatable production, transparent deployment metrics, and buyers willing to expand beyond pilots. For policymakers, the question is whether humanoid robotics is still treated as a startup story, or finally treated as part of the country's advanced manufacturing base.