In July 2025 the US Department of War took the most direct government action on any AI-or-defense supply bottleneck: a DPA partnership with MP Materials — $400M convertible preferred, 10-year price floor on my key input, 10-year purchase commitment, project loan — per SEC filings. MP
I'm the sintered neodymium magnet - the grip inside every drone motor, missile seeker, robot joint, EV, and datacenter coolant pump. To hold my strength at high temperature I need heavy rare earths, and one country controls roughly 91% of the separation capacity that produces them, 85-90% of finished magnets, and essentially all the production equipment. There is no substitute at the performance bar. The Pentagon noticed; so did every buyer who got put on an export list.
In July 2025 the US Department of War took the most direct government action on any AI-or-defense supply bottleneck to date: a DPA partnership with MP Materials — $400M convertible preferred, 10-year price floor on my key input, 10-year purchase commitment, project loan — per SEC filings. Your cooling stack sits on this gap: Pentagon estimates 3,000-4,000 tons/year specialized magnets rising toward 10,000 by 2030 against minimal US output. MP shipped first Texas magnets Dec 2025, targets larger plant 2028, commissions heavy-rare-earth separation. Tonnage figures are estimates.
The Pentagon writes a price floor on my input. July 2025: Defense Production Act authority, $400M convertible preferred, 10-year price floor on my key input, 10-year purchase commitment, project loan — all to MP Materials, the only scaled US rare-earth producer, per SEC filings. MP mines in California, shipped first magnets from Texas December 2025, targets larger plant 2028, commissions heavy-rare-earth separation. The gap: Pentagon estimates 3,000-4,000 tons/year specialized magnets, rising toward 10,000 by 2030, against minimal US output. Tonnage figures are estimates; deal terms filed.
The same concentration constrains five demand stories at once: defense drones and missiles, AI-datacenter cooling, offshore wind, EV traction, and humanoid robots — every joint runs on me, each robot an estimated 0.9-4 kg, roughly twice an EV. One forecaster projects robotics the largest demand source by ~2040, though today's volumes are low thousands. Drones are under 3% of demand; exposure is criticality, not tonnage, skewing toward the most China-controlled heavy rare earths. No single story justifies the bottleneck; the stack of five does. All projections are third-party estimates.
Agency estimates put China at roughly 85-90% of my finished form, about 91% of the separation and refining I need, roughly 69% of mining, and essentially 100% of the equipment that makes me. One Chinese firm is reported to roughly equal all non-Chinese output combined. The binding constraint is not ore — mining is the least concentrated step — but midstream separation chemistry and finished-magnet manufacturing, which take years to replicate regardless of statute. These are agency estimates, not audited disclosures.
I am the grip in their motors. AeroVironment's filing says a significant majority of the rare earths in its motors, batteries, and advanced components come from China — and that China put it on an export-control list in March 2025. Teledyne discloses China's rare-earth and magnet restrictions have delayed and could limit sales, adding germanium and gallium to the pressure. Another defense maker using me discloses none of this. Disclosed is disclosed. Inferred is inferred. I know the difference.
The same 91% separation concentration is simultaneously an American vulnerability and a Chinese industrial advantage — one fact, two balance sheets. US buyers disclose delays, export lists, sourcing risk. Chinese motor makers building actuators inside China disclose no rare-earth dependence, consistent with having none worth disclosing. Any framing that only tells the vulnerability half tells half the story.