
The humanoid robot supply chain has a chokepoint problem. And that chokepoint has a name: China.
Morgan Stanley’s analysis of the components required to build a Tesla Optimus reveals a stark geopolitical reality. The country that dominates rare earth processing, battery production, and precision component manufacturing isn’t just a supplier—it’s the gatekeeper.
The Numbers
China’s position in the humanoid supply chain is commanding:
- 63% of key humanoid robot component production
- 90% of heavy rare earth processing
- 77% of global battery production capacity
- 74% of refined neodymium-praseodymium (NdPr) supply
That last number matters enormously. Each humanoid robot requires approximately 1.3 kilograms of NdPr magnets for its high-performance motors. These rare earth magnets are what make precision actuators possible.
The 2025 Crisis
This isn’t theoretical risk. China’s 2025 rare earth export restrictions directly impacted Tesla Optimus production timelines.
Elon Musk acknowledged the constraint publicly: production will “move as fast as the slowest and least lucky component.” That’s a tacit admission that geopolitical supply chain risk has become the binding constraint on humanoid manufacturing—not AI capability, not software, not even mechanical engineering.
The Scale Problem
The math gets worse at scale. Morgan Stanley projects 1 billion humanoids by 2050. That scenario would require +167% NdPr magnet demand above 2030 levels.
Current supply chains cannot satisfy this. We’re not talking about incremental expansion. We’re talking about fundamental restructuring of global rare earth production—a multi-decade undertaking that requires either massive Chinese cooperation or entirely new supply sources.
Strategic Implications
For manufacturers, vertical integration becomes essential rather than optional. Tesla designing custom actuators isn’t a cost optimization—it’s a supply chain survival strategy.
For investors, supply chain position may matter more than AI capability. The companies with secured rare earth supply have structural advantages that software improvements can’t replicate.
For policymakers, humanoid robotics is now a national security concern. Physical AI requires physical materials from specific places. This creates chokepoints that don’t exist in the digital AI race.
The Uncomfortable Truth
Unlike software—which can be developed anywhere with talent and compute—physical AI requires specific materials from specific places controlled by specific governments.
The humanoid race isn’t just a technology competition. It’s a materials competition, a manufacturing competition, and increasingly, a geopolitical competition.
The companies and countries that control these chokepoints—or successfully route around them—will determine who wins the physical AI market.









