AI supply-chain thesis — mapping bottlenecks, focus companies, and supply-chain exposure for investors.
**Bottleneck theme:** Heavy rare-earth alloying (Dy, Tb, Sm) for NdFeB magnets **Focus:** $ALOY — REALLOYS INC. **Horizon:** 18–36 months for binary execution; staged milestones every 6 months. ReAlloys ($ALOY) is the underappreciated bet on ex-China heavy rare-earth metallization — the missing middle of the magnet supply chain that suddenly matters because humanoid robotics is finally on a real production curve. A single humanoid carries 40–50 NdFeB-wound actuators (every joint, every finger, every wrist), which is roughly **10–20× the rare-earth content of an EV per unit**. Unitree's $608M Shanghai STAR IPO at 67% gross margin on a $27,300 G1 humanoid is the public market's first humanoid pure-play; Tesla Optimus, Figure, and Apptronik are racing it; every one of them needs a non-China magnet path for export-control and defence-customer eligibility. China controls 90%+ of global heavy rare-earth (Dy, Tb, Sm) separation and alloying capacity. Light rare-earths can be mined and refined in the West today — $MP, $LYC.AX, $ILU.AX, $UUUU all sit on that side. The heavy stack is the chokepoint. Beijing's escalating rare-earth export controls have made standing up domestic alloying capacity a national-security imperative, and the same chokepoint applies to the magnets that go into Unitree's BLDC motors, Optimus's hip actuators, and Apptronik's hands. Without the Dy/Tb alloy stage, separated rare-earth oxides do not become magnets, and Western humanoid OEMs cannot ship to defence or strategic customers. ReAlloys (formerly USA Rare Earth's metallization arm) is one of a small handful of Western players with the metallurgical know-how to take separated Dy/Tb/Sm oxides and turn them into the alloys and ingots used in NdFeB sintered magnets. The Pentagon's interest is not theoretical — DPA Title III dollars, IBAS funding, and direct prime-contractor offtake agreements are flowing to qualified domestic alloyers — but the larger demand pool now is robotics. Every Unitree, Optimus, Figure, and Apptronik unit that ships through 2028 needs to know whether its magnets came from a Chinese sintering line that may not survive the next export-control escalation. **The pair trade is structurally cleaner than the single name.** $MP is the upstream miner. $USAR runs the sintered-magnet plant downstream in Oklahoma. ALOY is the metallization layer between them. Long any of them is long the same thesis; long the chain is long the bottleneck. Without ALOY-type metallization between mining and magnets, $MP's REO does not reach an Optimus motor and does not reach an F-35 actuator. **Three reasons this re-rates, not just incrementally appreciates:** 1. **The humanoid TAM dominates the defense TAM.** Even at conservative humanoid unit forecasts (1M units by 2030 at 5–8 kg NdFeB per unit), the magnet demand from humanoids exceeds the entire current F-35 + missile + radar magnet stack by 5–10×. Defense was the catalyst funding the build; robotics is the demand that pays back the capex. 2. **Beijing's export controls re-price every kg of qualified ex-China heavy RE alloy.** The Dy/Tb price spread between Chinese benchmark and ex-China-qualified material is the real revenue line for ALOY — and that spread widens, not narrows, every time export rules tighten. 3. **Verticalized OEMs reset the buying pattern.** Unitree, Tesla, BYD, and the Chinese humanoid cluster will buy domestically when they can. Western OEMs cannot — they need a parallel non-China stack, and ALOY is one of the few names that can sell into it. This is binary execution. Scale a process that has never run at industrial volume outside China, in a politically protected market with subsidised capex and pre-committed offtake. The miss case is that the demand pull is real but ALOY cannot scale fast enough to capture it before a competitor (Quadrant, LCM, Solvay, or a fast-following Korean alloyer) takes the offtake. ## Validation metrics The thesis is **confirmed** if most of these green-line in 12-month windows, **invalidated** if 3+ red-line in 6-month windows. 1. **DPA Title III / IBAS / DOE awards to ALOY** — cumulative ≥$150M by Dec 2027. 2. **Annual heavy-RE alloy throughput** — ≥250 tons/yr by 2027; ≥1,000 tons/yr by 2029. 3. **Named offtake agreements** — ≥3 signed offtakes with magnet OEMs or humanoid platforms by mid-2027. 4. **Customer disclosure naming ALOY** — $MP, $USAR, JL Mag, Quadrant, Vacuumschmelze, or a humanoid OEM citing ALOY by name in earnings or 10-K (rumoured offtakes do not count). 5. **Dy/Tb spot-price spread vs. Chinese benchmark** — widening spread (ex-China > China) confirms pricing power; compression invalidates it. 6. **Humanoid OEM ex-China magnet path announcements** — Tesla, Figure, Apptronik, or 1X explicitly designating an ex-China magnet supplier. 7. **Quarterly cash burn vs. subsidy + pre-payment offset** — net burn ≤$40M/q absent capex catalysts; below this and a bridge financing/dilution event is likely. 8. **Qualification timelines with magnet OEMs** — first commercial-grade qualification complete by Q2 2027 (magnet qualification is typically 12–18 months and gates offtake conversion). 9. **Inventory of separated Dy/Tb feedstock on-site** — ≥6 months of forward production demand. 10. **Pentagon strategic-stockpile commitments** — multi-year DLA-Strategic Materials contract covering ≥30% of nameplate output. **Falsifiability:** if by Dec 2027 ALOY has shipped <100 tons/yr cumulative, has no named-OEM offtake, and the Dy/Tb spread has compressed back to Chinese benchmark, the thesis is wrong regardless of price action. ## Horizon and exit triggers | Window | What to watch | Position posture | | --- | --- | --- | | **0–6 months** | DPA Title III announcement; first humanoid OEM letter of intent; Beijing rare-earth export rule updates | Build into IPO/grant catalysts; size for binary outcome (≤3% of book) | | **6–12 months** | First commercial qualification; offtake count 0→2-3; Unitree post-IPO trading reset prices in the magnet chain | Add on confirmed offtake; cut on missed milestones | | **12–24 months** | First production-volume runs (250+ tons/yr); humanoid OEMs disclose supply chain in 10-K | Trim into news events; binary outcome resolves here | | **24–36 months** | Steady-state ramp at 1,000+ tons/yr; multi-year offtake stack visible in financials | Hold winners; exit losers cleanly | **Exit (positive):** ≥1,000 tons/yr with ≥4 multi-year offtakes by end-2028 → trim 50% into the re-rating, hold the rest. **Exit (negative):** any of (a) DPA Title III award delayed past Q2 2027; (b) first commercial qualification slips past Q4 2027; (c) Dy/Tb spread compresses for two consecutive quarters; (d) a competing Western alloyer (LCM, Quadrant, Solvay) wins the same offtake ALOY is bidding on. This is not investment advice. Do your own research and confirm the supply-chain wiring on Macroplane before sizing a position.
Browse all AI supply-chain theses · Macro trends · Industries · Product categories · Baskets