During the launch of the second-generation Tesla Optimus, Elon Musk repeatedly emphasized cost control. He stated the goal of pushing the unit price below $20,000, utilizing integrated die-casting processes to reduce structural component costs, and optimizing wiring harness layouts to save on copper consumption.
However, he omitted one crucial figure: each Optimus requires 3.5 to 4 kilograms of NdFeB (Neodymium Iron Boron) rare earth permanent magnet materials.
This consumption exceeds that of the drive motor in a standard electric vehicle. Meanwhile, 94% of the global NdFeB production capacity is located in China.
On August 4, 2026, Bernstein released the report "Electric Revolution: Resilience — A Critical Chokepoint for Humanoid Robots and EVs," shifting the focus from algorithms to this fundamental material often overlooked by the industry.
01. Why Does a Single Robot Consume So Much Magnetic Material?
NdFeB is the core component in motors. Its magnetic energy product is several times that of ferrite and Alnico, determining how small the motor can be made and how much output power it can deliver. A humanoid robot has 40 to 50 servo joints, and each joint must be packed with a high-power-density motor. With space at an absolute premium, performance relies entirely on the capabilities of the magnetic materials.
High-temperature resistance presents an even greater technical challenge. Joint motors generate continuous heat during complex movements, and standard NdFeB will demagnetize at high temperatures. The solution is to add heavy rare earth elements such as dysprosium and terbium to enhance coercivity. However, these elements are scarcer, and their prices are several times higher than those of basic magnetic materials.
Assuming humanoid robots enter the mass production stage of one million units, the new annual demand for NdFeB would be 35,000 to 40,000 tons. This is equivalent to the sudden addition of magnetic material demand for 12 million to 15 million NEVs (New Energy Vehicles) in the market.
Magnetic materials account for 15% to 20% of motor costs, and the motor is the core of the joint actuator. Fluctuations in upstream rare earth prices directly lead to a narrowing of the gross profit margin for the complete machine. Previously, the industry believed that scaling up could rapidly reduce costs, but magnetic materials represent a rigid cost that cannot be compressed.
02. There Are Mines Overseas, So Why Are They Still Choked at the Bottleneck?
China's rare earth advantage does not rely solely on resource reserves.
According to 2024 data from the IEA (International Energy Agency): China accounts for 59% of global rare earth mining, 91% of refining and separation, and 94% of permanent magnet manufacturing. The further downstream, the higher the share. The barrier lies in the processing stage.
Both the United States and Australia have rare earth mines. What happens after extraction? The ore must undergo heap leaching, solvent extraction separation, metallothermic reduction, and then strip casting, hydrogen decrepitation, sintering, and post-processing. Every step requires long-term process accumulation and equipment debugging. Throwing money to build factories cannot solve the problem.
It is estimated that by 2030, the NdFeB production capacity in the United States will be less than 20,000 tons, the European Union less than 14,000 tons, and the rest of Asia combined about 19,000 tons. China's capacity is 290,000 tons. A tenfold gap. Non-Chinese capacity can only serve as a high-cost supplement and cannot change the supply landscape.
Can recycling alleviate the pressure? In 2023, rare earths provided by globally recycled magnetic materials accounted for only 3% of total demand. Even by 2050, this proportion can only rise to about 10%. Electronic products have a long end-of-life cycle, high dismantling costs, and there are losses in repurification.
03. How Will Export Controls Change the Rules of the Game?
Following the implementation of the Export Control Law in 2020, China has gradually tightened its control over rare earth technologies. In 2023, rare earth extraction and separation technologies, along with permanent magnet preparation technologies, were added to the catalog of technologies prohibited from export. In 2024, the Regulations on the Management of Rare Earths officially came into effect. In April 2025, export controls were imposed on seven categories of medium and heavy rare earths, including samarium, gadolinium, terbium, and dysprosium, and a "minimum content traceability rule" was introduced—overseas manufactured products containing controlled rare earths of Chinese origin valued at more than 0.1% will be subject to jurisdiction.
In November 2025, it was announced that some measures would be suspended until November 2026, but the signal has already been sent.
The market has reacted. In the month when the control measures took effect in April 2025, China's export volume of rare earth permanent magnets dropped significantly. Downstream manufacturers are re-evaluating their inventory strategies. Even if overseas robot manufacturers build magnetic material processing plants in Southeast Asia or Mexico, as long as the raw materials or core processes involve controlled items from China, their final shipments will face compliance risks.
Chinese domestic robot manufacturers with stable internal supply chains have a natural advantage in supply security. This advantage will be amplified during times of geopolitical tension.
The competitive logic for actuator and motor component suppliers has changed. Previously, it was about cost-effectiveness; now, it is about delivery certainty. Companies that can stably secure raw materials will win more orders, even if their prices are slightly higher.
Overseas magnetic material companies can capture a "backup premium," but their cost structure dictates that they cannot serve the mainstream market. Customers will pay more for supply chain security, but they will not bear a cost difference of three to five times.
04. What to Watch in the Next Three Years?
Over the past two years, the narrative in the humanoid robot field has revolved around large models, embodied intelligence, and end-to-end training. Algorithms can be rapidly iterated through code, whereas capacity building and process refinement require years or even decades.
Practitioners and investors are staring at demo videos of new robots, but they should pay more attention to rare earth export quotas, dysprosium and terbium price trends, and the production scheduling plans of leading magnetic material companies.
Industrial differentiation depends not only on software capabilities but also on who can lock in key upstream raw materials. From 2026 to 2028, this factor will predict industry trends better than any product launch event.
(For more in-depth articles on the humanoid robot track, please follow the WeChat official account "Humanoid Lecture Hall")