Rare-earth magnets & motor materials

Who owns it. Every humanoid actuator is a frameless torque motor wound around sintered NdFeB magnets, and China makes roughly 90% of them. The constraint isn't the ore — Mountain Pass and Mount Weld prove the West has rock. It's three decades of separation chemistry, metal-making, strip-casting and sintering at scale (Adamas puts global NdFeB output near ~200,000+ t/yr), plus the scarce heavy rare earths (dysprosium, terbium) that keep a magnet from demagnetizing at the temperatures inside a joint running flat-out, and a licensing regime Beijing has now twice weaponized (CSIS). What makes robotics different from every prior magnet demand wave is multiplication. A humanoid carries 40-plus servo motors, each needing 50-100 grams of NdFeB — roughly 2-4 kg per robot, with Tesla's Optimus pegged near 3.5 kg and Morgan Stanley modeling ~1.3 kg of NdPr per unit (atechmagnet, Adamas Intelligence). Tesla converted its Fremont Model S/X line to build up to a million Optimus a year and targets 50,000 units in 2026 (InvestorPlace). Multiply 50,000 robots by ~3 kg and the math is small today and ballistic if the ramp is real.

What breaks it. Not mining — magnet-making, plus heavy-RE feed, plus a price floor to survive Chinese flooding. Washington has underwritten all three. The DoD's July 2025 deal with MP Materials set a $110/kg NdPr floor for ten years and backstops the 10X magnet campus (~7,000 tpa, the anchor of ~10,000 tpa of total planned US capacity) MP broke ground on in Northlake, Texas (MP Materials). MP produced its first commercial NdFeB magnets at its Fort Worth Independence plant in December 2025 and begins commissioning a Mountain Pass Dy/Tb circuit (~200 t/yr) in mid-2026 (Bloomberg Law). USA Rare Earth commissioned its first sintered-NdFeB line at Stillwater, Oklahoma in March 2026, ramping to 600 t/yr by Q4 2026 and 1,200 by Q1 2027, and signed an offtake/distribution pact with Arnold Magnetic (USA Rare Earth). Noveon — privately held — is the only operational US sintered-NdFeB maker, with GM and ABB supply deals and a $215M January 2026 round (PR Newswire). The binding constraint upstream is heavy RE: Lynas's Malaysia circuit is the only commercial ex-China Dy/Tb source, Energy Fuels passed initial qualification benchmarks for its first US heavy-RE oxide in December 2025, and USAR's pending $2.8B Serra Verde deal would add the only scaled four-magnetic-element mine outside Asia (Mining.com, Energy Fuels, USA Rare Earth).

What forces the reprice. Two dated triggers. First, China's one-year suspension of its October 2025 extraterritorial controls lapses on November 10, 2026 — and the April 2025 licensing regime on seven heavy rare earths never lifted (China Briefing). Second, DFARS 252.225-7052 bans NdFeB and SmCo magnets touched by Chinese melting or fabrication from defense systems on January 1, 2027 (Crowell & Moring). Both squeeze the same thin pool of compliant supply that robot programs need, and markets should start handicapping it in H2 2026. Vertical integration by the robot makers does not close this gap: Tesla designs and assembles its own actuators (planetary roller screws plus brushless motors), but that integration stops at motor design — nobody is integrating back to magnet sintering, so the merchant magnet story survives even an in-house actuator (optimusk.blog).

Be honest about size and earliness. This is a forecast, not a backlog. No Western magnet maker has disclosed a humanoid design win; the named contracts (MP-GM, Noveon-GM/ABB) are EV and industrial, and robotics shows up as a future application in earnings calls, not a line item (magneticsmag). Today's humanoid fleet runs almost entirely on Chinese magnets — sintered NdFeB is ~91-94% Chinese-made (Adamas Intelligence). MP's entire Q1 2026 was $90.6M revenue on 917 t of NdPr; total announced US magnet capacity reaches maybe ~30,000 t/yr by 2028-29, a tenth of China's and still short of heavy-RE feed. The demand wave is multiplicative and real on paper; the revenue is small, the design wins are absent, and the timeline runs through years of unglamorous metallurgy with slippage already visible. Own the few balance sheets that can legally sell a compliant magnet or its heavy-RE feed into a 2027 supply gap — and size it as the option it is.