The Rare Earth Demand Hidden Inside Millions of Robot Joints

Global humanoid shipments were only about 20,000 units in 2025, but Chinese manufacturers accounted for roughly 95% of them. China's Ministry of Industry and Information Technology expects the country to build more than 100,000 humanoids in 2026, while BofA Global Research forecasts global annual shipments of 1.2 million units by 2030.[1] For rare-earth investors, the important variable is not simply the number of robots. It is the number of motorized joints inside every robot.

A humanoid needs compact, high torque actuation at its hips, knees, ankles, shoulders, elbows, wrists, hands, and other moving assemblies. High-performance neodymium-iron-boron magnets are particularly useful because they allow electric motors to generate substantial torque without becoming too large or heavy. That turns humanoid adoption into a repeated materials problem: one robot creates demand for dozens of motors, and every increase in dexterity can add more magnet-bearing actuators.

The demand multiplier is the joint count

A simple sensitivity model shows why this matters. These are modeled assumptions, not a disclosed bill of materials for any specific robot.

Assume a humanoid contains 30 joints using permanent-magnet motors. Assume each of those motors contains 50 to 100 grams of NdFeB magnet material. That produces an estimated 1.5 to 3.0 kilograms of NdFeB magnets per robot. If roughly 30% of the magnet mass consists of magnet rare-earth elements such as neodymium, praseodymium, dysprosium, and terbium, the modeled rare-earth content is about 0.45 to 0.90 kilograms per robot.

At 1.2 million robots annually, roughly the 2030 shipment level cited by BofA, those assumptions imply 1,800 to 3,600 tonnes of incremental NdFeB magnet demand and approximately 540 to 1,080 tonnes of contained magnet rare earths each year. At five million robots, the same model rises to 7,500 to 15,000 tonnes of magnets and 2,250 to 4,500 tonnes of magnet rare earths. At ten million robots, it becomes 15,000 to 30,000 tonnes of magnets and 4,500 to 9,000 tonnes of rare-earth content.

Those figures are scenarios, not forecasts. Motor architectures can change, some joints can use alternatives to rare-earth permanent magnets, and manufacturers have strong incentives to reduce material intensity. But the calculation exposes the variable investors should watch. A more dexterous robot can consume more magnetic material even if unit shipments remain unchanged. Robot demand therefore scales through both units and joints per unit.

That distinction becomes more important because the magnet supply chain is much more concentrated than the robot OEM landscape. The International Energy Agency estimates China produced 60% of mined magnet rare earths in 2024, 91% of refined output, and 94% of sintered permanent magnets. It also identifies metallisation and finished magnet manufacturing, rather than simply mining, as major constraints on supply diversification.[2]

The investable chain starts before the robot factory

The relevant public-company map therefore extends well beyond Tesla or the emerging humanoid manufacturers.

At the raw-material and separation layer, Lynas Rare Earths and MP Materials belong in the research perimeter because they represent major attempts to establish magnet rare-earth supply outside China. MP Materials is also moving downstream into metals, alloys, and finished NdFeB magnets. Inside China, companies such as China Northern Rare Earth sit closer to the enormous domestic refining and materials ecosystem that already supports electric motors, industrial automation, and advanced manufacturing.

The next layer may ultimately matter more. JL MAG Rare-Earth produces high-performance NdFeB magnets, placing it between separated rare-earth material and the motors that turn electrical energy into joint motion. Farther downstream, Leader Harmonious Drive Systems supplies harmonic reducers and integrated rotary actuator technology, while Shenzhen Inovance Technology operates in servo motors and motion control. These companies occupy the mechanical and electromechanical layers where a rare-earth input becomes a functioning robotic joint.

Finally come robot manufacturers such as UBTech, Unitree, Tesla, and other embodied-AI developers. This is a functional supply-chain map, not a claim that every named company directly supplies the next one. The important investment research question is which of these relationships actually exist, how concentrated they are, and where the economic exposure sits.

That is where robot adoption could produce a very different earnings sensitivity from the headline narrative. A large robot manufacturer may treat motors or magnets as only one part of a complex bill of materials. A smaller magnet or actuator supplier could experience a much larger change in revenue if it wins a high-volume platform. The relevant questions become whether several robot manufacturers share the same magnet, motor, or actuator supplier, whether one customer represents an unusually large share of a supplier's business, and whether that dependency has been increasing as production moves from prototypes toward standardized platforms.

The more important bottleneck may also migrate over time. Rare-earth ore can exist outside China without creating a diversified robot supply chain. The material still has to be separated, converted into metal and alloy, manufactured into high-performance magnets, incorporated into compact motors, and qualified inside actuators that must survive millions of motion cycles. The IEA's finding that magnet manufacturing remains one of the least diversified stages suggests that simply opening more mines does not remove the dependency.[2]

Conclusion

Humanoid robots do not need to reach smartphone scale to create a new rare-earth demand pathway. A market measured in several million robots would represent tens or hundreds of millions of motorized joints, turning improvements in robot dexterity into incremental demand for high-performance magnets.

That makes joint count as important as robot count. If humanoid production scales, the most consequential exposure may sit upstream with the companies converting rare-earth elements into magnets and magnets into compact motion systems. Quantifying which robot makers actually depend on which magnet, motor, and actuator suppliers, and how economically important those relationships are to each side, will matter more than simply identifying companies with "robotics exposure."

Sources

How to Cite This

For research inquiries or data access: press@altsets.com

Go Deeper