The ceiling on humanoid robotics is not artificial intelligence. It is the magnet. Humanoid robots are often described as a breakthrough in software, but their physical bodies depend on a supply chain that is narrow, geographically concentrated and slow to expand. Actuators account for 40 to 60 percent of a humanoid robot's bill of materials, and the rare earth magnets inside those actuators are mostly refined in China. The result is that the hardest problem in physical AI may not be the model at all, but the magnet.

Key facts at a glance

• Actuators account for 40 to 60 percent of a humanoid robot's bill of materials.
• The gearbox accounts for 30 to 50 percent of the actuator's cost.
• China controls about 90 percent of rare earth magnet processing capacity and 69 percent of extraction.
• Europe's Critical Raw Materials Act aims for no single third country to supply more than 65 percent of any strategic raw material by 2030.
• From 2028, magnets above 0.2 kilograms require a label, digital record and recycled content figure under Articles 28 and 29.
• A German motion technology company is preferred supplier for more than half of one humanoid robot startup's joint actuators, with a seven-digit unit agreement through 2031.

Actuators are the real constraint

An actuator is the component that converts electrical energy into motion. In a humanoid robot, every joint needs one. The shoulder, elbow, hip, knee and ankle all require compact, high-torque actuators that can support the robot's weight while performing fast and precise movements. These units are not commodity parts. They involve precision engineering, specialized materials and extensive testing.

The cost share is substantial. Actuators make up 40 to 60 percent of a humanoid robot's total bill of materials, according to a consulting firm analysis cited in the original commentary. The gearbox, the part that increases torque while reducing speed, accounts for 30 to 50 percent of the actuator's cost. That means a robot's mechanical core is also its economic core.

One American manufacturing sourcing executive argued in a widely read column that new actuator capacity takes years to build, not quarters. He should know: he co-founded a manufacturing sourcing platform that was later acquired by a Japanese components supplier for $350 million, and now leads that supplier's AI division. The point is not abstract. If a robot maker wants to scale from thousands to millions of units, it cannot simply order more actuators.

The gearbox queue

The gearbox is where the queue forms. High-performance robots rely on strain wave or harmonic drives, which offer high reduction ratios in a small package with low backlash. That combination is crucial for smooth and accurate joint movement. But the number of qualified suppliers is small. The established names in precision gearboxes have decades of accumulated manufacturing expertise, and they are not adding capacity overnight.

Qualification cycles are long. A robot manufacturer cannot just buy a gearbox off the shelf and trust it in a humanoid joint. It must test the component for durability, repeatability, temperature tolerance and fail-safe behavior. The cycle can take years. If the supplier is also producing for automotive, aerospace and industrial automation, humanoid robots may not get priority.

The result is a bottleneck that cannot be solved by a software update. Even if a company's AI model is perfect, the physical joints need to exist. And the companies that control gearbox production are not scaling at the speed of the AI hype cycle.

The magnet dependency

Inside the motor sits an even harder dependency. Every high-torque joint motor needs a neodymium magnet. Neodymium-iron-boron magnets offer the strongest magnetic field per volume, which allows robot designers to shrink the motor and fit it inside the joint. Without these magnets, a humanoid robot's limbs would need to be larger, heavier or less powerful.

Neodymium is a light rare earth element, and magnet production requires more than mining. It requires separation, refining and alloying. China controls roughly 90 percent of the processing capacity for rare earth magnets. The mining figure is lower, around 69 percent, but that does not solve the problem. Refining is the binding constraint, and China's dominance there is overwhelming.

This concentration creates a strategic risk for every robot maker outside China. A geopolitical dispute, export restriction or shipping disruption can halt production even if the AI software is ready. The rare earth supply chain is not just an economic issue; it is a national security and industrial policy issue.

Europe's policy response

Europe has written a law about exactly this. The Critical Raw Materials Act identifies light and heavy rare earths as strategic raw materials. It sets a target that by 2030 no single third country supplies more than 65 percent of the European Union's annual consumption of any strategic raw material. The intent is to reduce dependency on China.

The regulation also reaches the robot itself. Articles 28 and 29 explicitly name industrial robots. Starting in 2028, magnets heavier than 0.2 kilograms will need a label, a digital record and a recycled content figure. This means every humanoid robot sold in Europe must track the provenance and recycled share of its magnets.

The policy is ambitious, but the supply side is not ready. Europe currently has almost no commercial refining capacity for the type of rare earth magnets used in high-torque actuators. A cap dated 2030 and a labelling regime dated 2028 are difficult to meet if the underlying refining industry does not exist. The law creates demand for transparency, but it does not create smelters.