Live Casefile · Opened April 13, 2026

The Market Counted Mines. It Missed Magnets.

The Pentagon does not need rare-earth optionality. It needs qualified, traceable, non-Chinese magnet supply — and the clock is almost gone.

Updated June 8, 2026 · MP v USAR lawsuit · $1.6B CHIPS · 3 bypass paths · Exhibit 4 capacity table · strategic-equity sidebar

Updated: Jun 10, 2026 · ~28 min read Author: Ahmed Mir Confidence: Moderate–High Framework: Cascade casefile Horizon: Jan 1, 2027 Press: [email protected]

Built on ForcedAlpha’s proprietary supply chain intelligence graph (4,851+ nodes, 21,311+ edges). NDAA rare-earth subgraph snapshot 3ae2dc751710 — 55 nodes, 195 edges. Data sourced from NDAA Public Law 118-31 § 854, 10 USC § 4872 (Cornell LII), MOFCOM Announcements 18/55-58/61-62 of 2025, DFARS Clause 252.225-7052, USGS Mineral Commodity Summaries 2025, Adamas Intelligence, SEC filings (MP Materials, USA Rare Earth, REalloys, Energy Fuels, Neo Performance Materials), and company press releases (Lynas, ZF, Honda/Daido, Tesla Q1 2025 earnings call).

This is not investment advice. This casefile maps structural supply chain dependencies using public data. All claims are sourced and falsifiable. Do your own research.

Claim Confidence

Verified Primary source confirmed

Derived Analytical conclusion from verified inputs

Modeled Estimate or projection

Tags appear throughout this casefile. Full methodology →

Enforcement Scenarios

Three regimes. The middle one is the base case.

Who Wins Under Waivers

The base case — a managed waiver bridge — is not bearish. It is an official admission that compliant supply does not exist at scale. The trade shifts from “the deadline forces immediate demand” to “the waiver proves the shortage and unlocks multi-year federal funding, offtakes, and qualification programs.”

The Law Forcing the Shortage

Congress did not write one law. It wrote six, each closing a loophole the last one left open. The result is a mine-to-magnet ban that takes effect January 1, 2027.

The ban is not absolute. The Pentagon can grant nonavailability waivers — each individual determination is capped at 36 months, and class determinations covering multiple contracts require USD(A&S) approval with at least 30 days advance Congressional notice. The ladder explains why the deadline can slip while the capital flows cannot.

The law is Section 854, NDAA FY2024 (Pub. L. 118-31), implemented through DFARS 252.225-7052. It covers NdFeB magnets, samarium-cobalt magnets, tungsten, and tantalum.^[2]

Today, the ban covers materials “melted or produced” in China. On January 1, 2027, it expands to “mined, refined, separated, melted, or produced” — the full mine-to-magnet chain.^[3]

The critical word is any. Before 2027, Chinese ore could be shipped to a Japanese factory and the finished magnet was compliant. After January 1, that loophole closes. If the neodymium was mined in Baotou, the magnet fails — no matter where it was sintered.^[4]

Full statutory genealogy, the § 857 mis-citation issue, COTS exemption, waiver mechanics, and pending legislation: Appendix · Legal Detail →

The market prices rare earths as a commodity cycle. The Pentagon is turning them into a compliance asset with a deadline.

The Capacity Illusion

Capacity without qualification is not supply.

Analysts count 8,000–10,000 tonnes of non-Chinese NdFeB capacity. On paper, that covers defense demand. In practice, most of it runs on Chinese feedstock — and the law bans the feedstock, not the factory. Strip out everything that has not requalified on non-Chinese oxide, and the real number is roughly half.

The Nameplate Illusion

Total non-Chinese NdFeB magnet capacity worldwide is roughly 8,000–10,000 metric tonnes per year. Japan alone produces approximately 4,500 MT/yr through TDK Corporation, Shin-Etsu Chemical, and Proterial (formerly Hitachi Metals). Europe adds roughly 2,000 MT from Neo Performance Materials’ Estonia plant, and a further US line is ramping at eVAC Magnetics’ new Sumter, South Carolina facility toward 2,000 MT. US-based producers — MP Materials, USA Rare Earth, and Noveon Magnetics — contribute 1,600–2,000 MT in various stages of commissioning.^[40][46][34]

Against US defense demand of 3,000–4,000 MT/yr (Commerce Department estimate), this looks comfortable. The gap appears manageable.

It is not.

The Feedstock Constraint

The NDAA does not ban magnets made in Japan. Japanese magnets are compliant — as long as the rare earth oxide inside them did not come from China.

Today, Japanese magnet producers import the vast majority of their rare earth oxides from China. Switching to non-Chinese feedstock — from Lynas (Australia/Malaysia), MP Materials (United States), or Energy Fuels (United States) — requires requalification of the entire magnet production process. Different oxide purity profiles mean different sintering conditions, different heat treatments, different magnetic performance. Requalification takes quarters per production line. There is limited public evidence that the major Japanese producers have started.

The real non-Chinese-feedstock-qualified capacity today is closer to 3,000–4,000 MT: MP Materials, Neo Performance’s Estonia operation (sourcing from Lynas), and a handful of lines that have completed the switch. This means the supply gap during the 2027–2028 qualification window is roughly twice what consensus estimates suggest.

Every Western rare earth project in the last 15 years has been delayed 2–4 years from announced timelines. USA Rare Earth originally promised production in 2023; Phase 1a actually commissioned in March 2026. Lynas’s Texas heavy rare earth facility, funded by over $258 million in DoD support, is now described as “unlikely to proceed” by management. Per company and industry reporting, Iluka’s Eneabba refinery slipped from a 2026 commissioning target to 2027, with capital costs revised from an initial A$1.2 billion to A$1.7–1.8 billion; Arafura’s Nolans project reached Final Investment Decision on 21 May 2026 after years of delays, with first production now targeted for early-to-mid 2029.^[55][56][57]

The Federal Backstop (June 2026)

On June 3, 2026, the Department of Commerce signed definitive agreements with USA Rare Earth under the CHIPS Act, unlocking up to $1.6 billion: $277 million in federal funding plus $1.3 billion in senior secured loan capacity. USAR issued 16.1 million common shares and 17.6 million warrants to the US Government in exchange. Per USAR’s June 3, 2026 IR release, combined with the $1.5 billion PIPE closed in January 2026 and previous capital raises, total committed capital supporting USAR’s growth plan stands at approximately $3.5 billion.^[10][29] For comparison, Trump’s July 2025 equity stake in MP Materials was $400 million.^[11]

USAR — pre-commercial-revenue, mid-development at Round Top, and defendant in MP’s active trade-secret suit — just received roughly four times MP’s equity backing. Lawmaker pushback was reported before MP filed suit. The financing is milestone-tied and structured to reimburse capex, but the share-plus-warrant issuance is dilutive and large. Washington is not acting like this is a minor commodity shortage. It is funding pre-commercial magnet capacity like a strategic weapons system.

MP’s own production data confirms the floor: Q1 2026 NdPr production of 917 tonnes, up +63% year-over-year, against a DoD-locked $110/kg floor. The floor is now below market. The Pentagon’s pricing indicator tells you their private supply assessment: they expected the scarcity to persist.^[11]

This is not a permanent structural gap. It is a 2–3 year qualification and ramp bottleneck where demand is legally obligatory and supply is physically constrained.

Nameplate vs Qualified: The Producer Stack

Six Western producers appear in capacity tallies. Only MP Materials holds a locked, full-scale DoD offtake contract today — Neo, Noveon, and USAR are qualified or operational at smaller scale but constrained by feedstock allocation, defense contract status, and commissioning timeline. The table below carries every name, location, feedstock origin, qualification status, and timing chip in one frame.

How Defense Crowds Out Everyone Else

The Pentagon does not negotiate for magnet supply. It requisitions it. When qualified capacity is short and defense gets first call, every other buyer — EV manufacturers, wind turbine producers, robotics companies — goes to the back of the line.

MP Materials sued USA Rare Earth on May 22, 2026 over stolen heat-treatment recipes — specifically the grain-boundary diffusion (GBD) process that lets NdFeB magnets survive extreme temperatures. The lawsuit is years from resolution, but the filing itself tells you the moat is recipe-tight. Full case detail, GBD primer, and patent landscape →

The moat is real but narrow. Not unbreachable — litigated, recipe-tight, and held by a handful of producers.

Price Action

The price is already running. CIF Europe rare earth prices — assessed by Argus in mid-2025 and still elevated through 2026 as the control regime held^[37] — reflect a dual-market reality: neodymium oxide trades at $66–74/kg and neodymium metal at $81–86/kg — well above Chinese domestic benchmarks. Heavy rare earths are far more dramatic: dysprosium oxide at $650–850/kg and terbium oxide at $2,300–3,400/kg CIF Europe, with no further Chinese export licenses for heavy rare earth sales to Europe heard granted.^[54] Yttrium oxide — critical for laser crystals, radar ceramics, and jet-engine thermal barrier coatings — has moved to $50–70/kg CIF Europe, with traders reporting that business is no longer possible below $50/kg.^[54]

The DoD’s July 2025 deal with MP Materials locked in a $110/kg floor price for NdPr products — more than double the spot price at the time of signing. That floor is now below the market. The Pentagon’s own pricing indicator tells you their private supply assessment: they expected scarcity to persist.^[11]

Defense Absorbs Capacity

MP Materials’ new 10X Facility (commissioning 2028, ~10,000 MT capacity) has a 10-year DoD offtake covering 100% of production.^[11] Noveon, per company disclosures, holds defense grants and prioritizes DFARS-compliant output.

When defense buyers lock up qualified capacity at premium prices, everyone else — EV makers, wind turbine producers, industrial motor companies — goes to the back of the line.

This is not a short-the-primes thesis. Magnet cost increase per platform is < 0.5% of unit cost; the real cost to primes is compliance overhead. The trade is on the producers, not against the consumers. Cost-immateriality math →

Drones, Robots, and AI Add More Demand

The magnet ban does not exist in a vacuum. Four demand vectors — legacy defense, autonomous drones, missile defense, and humanoid robots — all hit the same scarce Western NdFeB stack at the same time. None of them can wait.

Four demand vectors hit the same Western NdFeB stack simultaneously: F-35 / submarine programs (legally mandated), Replicator drones (multiple thousands per year, net-new),^[14] Golden Dome interceptors (~$1.2T lifecycle, consumption-based — each one fired is gone),^[28] and humanoid robots (Optimus at 1M units would alone match the entire US defense magnet demand).^[16] See Exhibit 1 above. Per-program contractor list, BOM caveats, and Musk Q1 2025 earnings call detail →

China Still Controls the Choke

China is not flooding the market. It is restricting.

The architecture is layered. MOFCOM 18/2025 is the active baseline layer — not one of the suspended clocks — license-by-shipment controls on the Sm/Gd/Tb/Dy/Lu/Sc/Y elements plus NdFeB magnet materials. Heavy rare earth exports (yttrium, dysprosium, terbium) are still down roughly 50% vs the 12 months prior to the April 2025 controls, per Chinese customs data.^[37] The May 2026 customs release confirmed the pricing effect: aggregate rare earth exports surged 237% in dollar terms year-over-year while export volumes actually fell, confirming that China’s controlled scarcity is generating extraordinary pricing power on the materials it does allow out.^[61] Two additional MOFCOM suspension clocks — the extraterritorial 0.1% de minimis bundle (expires Nov 10 2026) and the US-specific Ga/Ge/Sb/super-hard-materials ban (expires Nov 27 2026) — sit inside a 17-day window. The first clock falls 52 days before NDAA § 854 binds; the second falls 35 days before. Both are one-year administrative suspensions that MOFCOM can reactivate without new legislation.^[17][62][63] The May 2026 Trump-Xi Beijing summit produced commitments on US agriculture purchases and Boeing aircraft but did not lift either MOFCOM suspension;^[18] snap-back should be treated as the base-case risk absent a new deal by October 2026.^[19]

Japan’s December 2025 cutoff precedent shows how this plays out: China used the active 18/2025 layer alone (no re-arming required) to target Shin-Etsu, TDK, Proterial, MHI, and Mitsubishi Materials — country-specific, license-issuance-driven, deniable. Japan’s response (Sojitz/JOGMEC locked ≥50% of Lynas LAMP HREO output through 2038) puts Tokyo in direct competition with the Pentagon’s March 2026 Lynas HRE purchase. Two sovereigns are now racing for the same scarce flow.

The pincer: dual MOFCOM suspensions expire Nov 10 and Nov 27 2026; NDAA § 854 binds Jan 1 2027. If the first clock snaps back, extraterritorial rare-earth processing restrictions reactivate 52 days before the procurement ban takes effect. If only the second snaps, the window shrinks to 35 days.^[20] A flood-the-commercial-channel counter-play remains a tail risk worth pricing — China has bankrupted Molycorp (2015), flooded solar (2012), and crushed lithium juniors (2023-24) using the same playbook — but the revealed strategy since 2023 is licensing leverage, not price destruction.

Japan December 2025 cutoff full chronology, MOFCOM license-regime mechanics, State Council Order 834-835 detail, and flooding-tail-risk historical analysis: Appendix · Precedent & Counterparty Detail →

How the West Tries to Escape

Three substitution paths are real, funded, and breaking ground. None of them reach defense-grade scale before the deadline.

Three substitution paths are building: Niron Magnetics (iron nitride Fe₁₆N₂, zero rare earths — 1,500 tpa Sartell MN plant under construction, Moog defense actuator partnership) ,^[22] ZF Friedrichshafen I2SM (inductively-fed synchronous EV traction motor with zero permanent magnets, removes ~3 kg NdFeB per EV motor),^[23] and Daido Steel (hot-deformed neodymium magnet with zero heavy rare earths — already shipping in Honda hybrid vehicle motors).^[24] See the bypass-paths embed above.

The three paths attack the problem at different depths. Niron eliminates rare earths entirely (longest qualification). ZF eliminates the magnet (mature engineering, EV-specific). Daido eliminates only the heavy rare earths (already shipping, limited by heat tolerance). None reach defense-grade scale before January 2027. The Pentagon’s priority allocation of Western magnet output remains the dominant scarcity force. Bypass paths relieve commercial pressure faster than defense pressure.

Niron Sartell facility detail, ZF I2SM motor architecture, Daido rapid-quench process, and acquisition/license optionality analysis: Appendix · Bypass Paths Detail →

The market counts capacity. Defense counts qualification.

What the Market Has Not Priced

Everything above is structural — statute, qualification clock, constrained capacity. The question is what it costs. Four layers, each grounded in ledger data: demand floor, supply gap, scarcity rent, and basket ramp. Every cell ties to a verbatim source quote.

1. Demand floor: $48.7B Pentagon Critical Minerals appropriation

The FY 2027 defense budget request pre-commits $48.675 billion to Critical Minerals as a single appropriation line, broken into three named carve-outs: $24.3 billion to the Industrial Base Analysis and Sustainment program (IBAS) — rare earths explicitly named — $6.4 billion in Defense Production Act Purchases (DPAP) for “mining, processing, metallization, and recycling,” and $18.0 billion to rebuild the National Defense Stockpile (NDS).^[42] An additional $17.5 billion Golden Dome program sits in the same budget — interceptor and radar production across Patriot PAC-3, SM-6, THAAD, and the new space-based layer all draw from the same high-performance magnet stack.^[42]

These are the appropriation lines, not the operational kilograms. The casefile does not convert dollars to magnet kilograms because no public-source BOM bridge ties them — per-platform magnet content is classified. What the appropriation establishes is the political commitment floor: $48.7 billion in pre-committed Critical Minerals dollars makes it structurally hard for the Pentagon to issue a class-wide §854 waiver without contradicting its own budget. The dollars themselves are the load-bearing fact.

2. Supply gap: definitional ambiguity, not authority dispute

Section 4 places defense-qualified Western capacity at 3,000–4,000 MT in 2027. A facility-by-facility rebuild using earnings disclosures and company filings produces a tighter band when the definition is restricted to operational capacity with traceable non-Chinese feedstock and named defense offtake agreements:

The 6× spread between definitions is the cascade’s structural mispricing. The market reads “30,000 MT of Western magnets” from press releases. The Pentagon binds against the bottom row.

3. Scarcity rent: bands only, no point forecasts

Per fund-grade discipline, pricing is presented as bands sourced to observed market data, not point forecasts. Three scenarios — baseline (current MOFCOM-suspended state extended), snap-back (Nov 10 / Nov 27 clocks reactivate without new deal), China-floods (Beijing deliberately collapses the merchant channel below Western production economics). The baseline bands are drawn from public Reuters / Argus / Benchmark reporting; the snap-back and China-floods edges are modeled scenarios disclosed as such.

4. Basket revenue ramp: ledger-grounded snapshots, outyears blank by design

Per fund-grade discipline, the table below contains only what management disclosed in Q1 2026 earnings calls and the Pentagon FY27 budget — no multiple framework, no market-cap targets, no per-name price targets. Outyear cells are blank where the ledger doesn’t support a dated milestone.

5. What the priced cascade does not answer

This exhibit is honest about its limits. It does not assert: (a) per-name market-cap targets — the multiple question (commodity-cycle vs scarcity-rent vs growth) is portfolio manager judgment; (b) the Tesla I2SM redesign trigger price for magnets — not publicly disclosed; (c) the Pentagon’s internal threshold for issuing a nonavailability determination — not disclosed; (d) the precise Lynas JARE floor price — Lynas will not disclose. Each absence is deliberate: declining to assert a number we cannot verify is the standard.

What the exhibit does establish: the demand floor ($48.7B Pentagon Critical Minerals + $17.5B Golden Dome) is appropriated, not aspirational. The defense-qualified Western supply at Jan 2027 is bounded by definition gap, not by analyst speculation. Scarcity rent is real, observed, and bracketed by Reuters/Argus/Benchmark data. The basket’s federal-backstop architecture is two distinct appropriation pools, not double-counted. Whether and how the cascade dollars accrete to per-name market caps is the reader’s call — we provide the source-grounded inputs.

Who Wins If This Is Right

If the cascade holds, three groups separate. Producers with qualified capacity and locked offtake benefit directly. Commercial buyers without alternative supply get squeezed on price and lead time. And a handful of substitution plays offer optionality on a longer timeline.

Near-Term Catalysts

What Would Prove This Wrong

Tells (Early Confirmation)

What Would Actually Break This

Confidence Assessment

Appendix

Receipts and mechanism detail behind the main-flow narrative. The Exhibits carry the load-bearing claims; this appendix carries the underwriting-grade detail.

Legal Detail

Statutory genealogy

The clause splits by date. Through December 31, 2026, contractors cannot deliver covered material that was “melted or produced” in covered countries (China, Russia, Iran, North Korea). Effective January 1, 2027, the prohibition expands to “mined, refined, separated, melted, or produced” — the full mine-to-magnet supply chain. The original 10 U.S.C. § 4872 was enacted via NDAA FY2019 § 871(a) (then codified as § 2533c) , restructured during the 2018 Title 10 recodification, and the mine-to-magnet concept was introduced in NDAA FY2021 § 844. NDAA FY2024 § 854 cemented the Jan 1 2027 cutoff into the implementing DFARS clause.^[3]

The § 857 mis-citation

COTS exemption mechanics

There is one exemption. Commercial off-the-shelf (COTS) items are exempt if the restricted material comprises less than 50% of the item by weight. NdFeB magnets are approximately 30% rare earth by weight (the balance is iron and boron). This means some COTS items containing NdFeB magnets may qualify for the exemption — but only if the magnet is embedded in a larger commercial product, not if the magnet itself is the procured item.^[6] Modeled

Waiver and nonavailability mechanism

The waiver mechanism exists but is designed to be narrow. The Under Secretary of Defense for Acquisition and Sustainment (USD A&S) can grant national security waivers under nonavailability determinations. Individual determinations are capped at 36 months per determination; in the case of an urgent national security requirement, certification to the defense committees may be provided up to 7 days after the determination is made. Class determinations — those applying to more than one contract — require USD(A&S) approval with at least 30 days advance Congressional notice to the maximum extent practicable. Contractors who promptly disclose noncompliant sources can apply.^[7] Individual or class nonavailability determinations can exempt specific contracts or material categories when compliant alternatives “cannot be procured as and when needed at a reasonable price.”^[8]

FY2026 NDAA and the Rare Earth Magnet Security Act of 2025

The FY2026 NDAA, signed into law in early 2026, did not push back the deadline. It tightened the restrictions — adding recycled samarium-cobalt provisions and expanding sourcing restrictions to critical minerals and batteries. The Rare Earth Magnet Security Act of 2025 (H.R. 1496) proposes $20–30/kg tax credits for domestic magnets, reinforcing the policy direction rather than softening it.^[9] Congress is moving harder, not lighter.

Federal Procurement Evidence

Active DoD and DOE contract awards related to rare earth supply chain buildout. Source: USASpending.gov, queried June 10, 2026.

DoD direct procurement

DPA Title III program management

KPMG holds a $16.4M Air Force contract for DPA Title III program development and support (awarded September 2020, ongoing). This is the administrative backbone of Defense Production Act investments in critical material supply chains.

What the money trail tells you

MP Materials is the only company with a direct DoD procurement contract for separated rare earth oxides — $11M via the Defense Logistics Agency, the same agency that procures fuel, food, and ammunition for the military. This is not R&D funding. It is a purchase order. Vulcan Elements is the only company with multiple DoD awards specifically for domestic magnet qualification. USA Rare Earth’s sole federal contract is $99K for an Army biotech research grant — two orders of magnitude below the procurement-grade contracts flowing to MP and Vulcan. The Air Force is also funding Niron Magnetics to develop rare-earth-free alternatives, a hedge that only makes sense if the DoD believes rare earth supply risk is structural, not transient.

Notable absence: no new DoD rare earth procurement or production contracts have appeared in 2026 despite the January 2027 deadline being six months away. The 2026 federal spending on critical minerals is concentrated in DOE consulting and program management (Energy Technology Alliance, Rack-Wildner, Accenture), not physical procurement. This is consistent with the managed-waiver-bridge scenario — the government is building the administrative infrastructure for the transition before the procurement ramp.

Strategic-Equity Doctrine — Cross-Vertical Pattern

Structurally adjacent to the magnet thesis but not load-bearing. Included for cross-vertical context.

The MP and USAR equity arrangements sit inside a broader pattern. In February 2025 Trump signed an executive order calling for the federal government to establish a sovereign wealth fund. Since then the administration has taken stakes in Intel, International Business Machines, and other quantum and critical-mineral companies.^[30] On June 5, 2026, CNBC confirmed that Sam Altman and the White House have been in talks for more than a year about a possible US government equity stake in OpenAI — potentially via OpenAI donating equity to seed a “Public Wealth Fund” the company outlined in its April 2026 policy proposal. The same week, Trump signed a directive instructing federal national security organizations to “accelerate AI adoption to meet surging demand” and to onboard the “most advanced AI models from multiple vendors” — plus an earlier executive order asking AI companies to voluntarily provide the government access to their models for up to 30 days before public release.^[30]

Same architecture, different industrial vertical. The rare-earth equity stakes are not a commodity-specific anomaly; they are the visible early instances of a Washington-wide strategic-equity doctrine.

Volume vs. Floor Architecture

Critics frequently point out that US military rare-earth procurement is an insubstantial fraction of market volume, concluding that defense demand cannot crowd out commercial buyers. This is empirically true but misinterprets the structural mechanics.

Per the U.S. Government Accountability Office (GAO-24-107176, September 2024), total DoD purchases of rare earths and other critical materials represent “less than 0.1 percent of the global demand for these materials.”^[52]

Section 854 does not require the Pentagon to compete on a volume basis; it forces the establishment of a price-insensitive political commitment floor for non-PRC supply chains. While China’s tightening export controls systematically choke commercial supply upstream, § 854 ensures Western producers have a guaranteed, high-margin baseline buyer to finance operations against. The defense floor and the commercial squeeze feed off the exact same scarce, qualified feedstock.

Supply-Chain Mechanics

Consolidated technical detail behind the nameplate-vs-qualified capacity claims above.

Feedstock Genealogy

The NDAA covered-materials chain is mine → concentrate → mixed rare earth carbonate (MREC) → separated oxides (NdPr, Dy, Tb) → metal → alloy strip → magnet powder → sintered or hot-deformed magnet. NDAA § 854 traces ALL of these to country of origin. A magnet sintered in Estonia or Vietnam using Chinese-separated dysprosium oxide is non-compliant. The supply chain knife cuts at the separation step — that is where China retains roughly 90% of global capacity for heavy rare earths (Dy, Tb).

Nameplate vs Defense-Qualified Capacity

The Western magnet capacity numbers reported in press releases are nameplate — design output of a facility once fully commissioned. Defense-qualified capacity is the subset that (a) has DFARS 252.225-7052 compliant feedstock traceability, (b) has completed prime-specific qualification testing for the relevant end-item, and (c) is contractually available to defense buyers without competing commercial obligations crowding it out.

Our 3,000–4,000 MT estimate for 2027 defense-qualified GBD-NdFeB capacity starts with nameplate and discounts for three factors: commissioning status (is the facility operational or still ramping?), feedstock traceability (does the facility use qualified non-Chinese feedstock, or is the supply chain uncommitted?), and defense allocation (is output locked to a defense offtake, or available on the open market?). The 30,000 MT nameplate aggregate becomes 3,000–4,000 MT once those three discounts are applied. This is estimated, not producer-audited. See Exhibit 4 above for the per-facility breakdown.

Grain-Boundary Diffusion (GBD) Primer

Standard sintered NdFeB magnets demagnetize above approximately 80°C. Defense actuators, EV traction motors, and aerospace turbomachinery operate at 150–200°C. The conventional fix is to alloy 5–10% dysprosium and terbium uniformly through the magnet during sintering — expensive (Dy/Tb are 10–100x the cost of Nd) and wasteful because the Dy/Tb only needs to be at the grain boundaries to do its work.

Grain-boundary diffusion (GBD) is the process of applying Dy/Tb only at the grain boundaries post-sintering — cutting heavy rare earth consumption by an estimated 30–50% while delivering the same thermal envelope. It is the dominant defense-grade process. The know-how is concentrated in a handful of Japanese, Chinese, and (newly) US operators. This is what the MP v USAR lawsuit is fighting over: whether GBD process knowledge can be litigated as trade-secret recipe.

If you remove GBD-NdFeB capacity from the Western stack, the remaining magnet supply cannot meet defense thermal specifications. That is why nameplate counts mislead — most Western nameplate is conventional sintered, not GBD.

Litigation & Moat Detail

MP v USA Rare Earth (May 22 2026) trade-secret lawsuit

Moat Stress Test: MP v USAR Trade Secret Lawsuit (May 22 2026)

On May 22, 2026, MP Materials sued USA Rare Earth in a Texas court for misappropriation of the formulations for grain boundary diffusion (GBD) — the specific process for forcing dysprosium and terbium into the boundaries of an NdFeB magnet to make it survive high-temperature operating envelopes. GBD is what lets an NdFeB magnet work inside an F-35 actuator at altitude, inside an EV traction motor above 180°C, and inside a humanoid BLDC motor under sustained load. It cuts heavy rare earth consumption by 30–50% versus uniformly-doped NdFeB. The technology was developed at Hitachi Metals (now Proterial), perfected at Shin-Etsu and a handful of Chinese producers, and independently rebuilt by MP. The complaint alleges USAR took MP’s specific recipes through an ex-employee and handed them to a third party.^[12]

The lawsuit is asymmetric information on the moat itself. If GBD were uncodifiable institutional muscle memory that takes 850 chemists and ten years to rebuild, no ex-employee could carry it out the door. The fact that MP thinks one did, and that MP is willing to litigate over it, means the chokepoint is something tighter and more transferable than “tacit knowledge.” It is a recipe — specific temperature profiles, timing, dopant ratios — held by a small number of operators, defended by trade secret law rather than physical impossibility. The Western capacity bottleneck is real. It is also one defection away from compressing.

Cost immateriality: why this is not a short-the-primes thesis

This is not a “short the defense primes” thesis. The magnet cost increase per platform is immaterial — less than 0.5% of unit cost. An F-35 costs approximately $90 million; even at aggressive DFARS-compliant magnet pricing, the incremental cost is perhaps $50,000–$100,000 per aircraft. For a Virginia-class submarine at approximately $4.5 billion per hull, the magnet cost increase is in the low single-digit millions.^[14]

The real cost to primes is compliance overhead: tracing every magnet to its origin, qualifying new suppliers, maintaining dual supply chains, managing the paperwork of waivers and nonavailability determinations. This is expensive but manageable for Tier 1 primes, and they will renegotiate or receive contract modifications. The investment thesis is on the producers, not against the consumers.

Demand-Side Detail

Underwriting-grade detail for each demand vector in the four-way collision — named contractors, BOM caveats, primary-source quotes. Exhibit 1 above encodes the load-bearing crosshatch; the prose below carries the verification.

Replicator: Thousands of Autonomous Drones

The Pentagon’s Replicator initiative aims to field “multiple thousands” of autonomous drones across all domains. Selected platforms include Anduril’s Altius-600, AeroVironment’s Switchblade 600, and the Ghost-X small UAS. Every electric drone motor contains NdFeB permanent magnets. At production volumes of thousands per year, this is net-new demand that did not exist when Section 854 was drafted in 2023.^[15]

Golden Dome — The Air-Defense Consumption Curve

The Golden Dome program (formerly Iron Dome for America, renamed May 2025 per Trump executive order) is the US homeland missile-defense build-out being ramped now — covering hypersonic, cruise, and ballistic threats. The Congressional Budget Office now estimates total lifecycle cost at ~$1.2 trillion over two decades — roughly seven times the initial $175B sticker. Acquisition costs alone exceed $1 trillion across interceptor layers and a space-based missile warning and tracking system. The CBO also warns the system could be overwhelmed by a full-scale peer-adversary attack — a real caveat worth pricing.^[28] Lead contractors include Lockheed Martin, RTX, Northrop Grumman, L3Harris, plus newer entrants like Anduril and Palantir. SDA, MDA, and the Space Force are running it. SpaceX and Lockheed Martin won up to $3.2B in space-based missile interceptor prototype contracts last month — that is real money already flowing.^[28] Interceptor and radar production across Patriot PAC-3 (LM), SM-6 (RTX), THAAD (LM), NASAMS (RTX/Kongsberg), and the new Golden Dome layer all pull from the same high-performance magnet stack used in radar actuators, guidance subsystems, and seeker servos. We don’t claim a specific component-level map of these BOMs — that data is classified or proprietary — but the demand-context point holds: this is consumption-based volume (each interceptor fired is gone) competing for the same qualified Western NdFeB output as F-35 actuators and submarine programs.

Humanoid Robots: The Demand Nobody Is Modeling

Each humanoid robot contains approximately 3.5 kg of NdFeB magnets across 28 or more brushless DC motors (BLDC). Tesla’s Optimus program, Figure AI’s deployments at BMW, and China’s AGIBOT shipping 10,000 units are all consuming the same magnet supply. At Tesla’s stated ambition of one million units per year, Optimus alone would require approximately 3,500 MT of NdFeB — equal to the entire US defense demand. Tesla CEO Elon Musk confirmed on the April 22, 2025 Q1 earnings call that China’s rare earth export restrictions had affected Optimus production.^[16]

Precedent & Counterparty Detail

MOFCOM 18/2025 permanent license-by-shipment baseline

The Permanent Baseline

MOFCOM Announcement 18/2025, active since April 4 2025, requires licenses for export of samarium, gadolinium, terbium, dysprosium, lutetium, scandium, yttrium, plus NdFeB magnet materials containing them. Issued by MOFCOM plus the General Administration of Customs. Heavy rare earth export volume (yttrium, dysprosium, terbium) is still down roughly 50% vs the 12 months prior to the April 2025 controls, per Chinese customs data.^[37] This layer is not on any expiry clock — it is the active baseline layer that licenses every shipment of those elements individually.^[17]

Suspension clocks 55-58/61/62 + State Council Orders 834-835

Two Suspension Clocks, 17 Days Apart

MOFCOM Announcements 55-58, 61, and 62 of 2025 — the extraterritorial 0.1% de minimis rule modeled on the US Foreign Direct Product Rule, plus a technology transfer ban covering separation, magnet manufacturing, and recycling — were suspended November 7–9 2025 at the Busan APEC sideline. The rare-earth extraterritorial bundle suspension expires November 10, 2026.

Separately, a US-specific ban on exports of gallium, germanium, antimony, and super-hard materials to the United States was suspended through November 27, 2026.

Both clocks are inside a 17-day window. Both are administrative suspensions that MOFCOM can reactivate without new legislation.^[62][63] Separately, State Council Decrees 834 and 835 (March–April 2026) added a broader legal framework for supply-chain security investigations and countermeasures against foreign extraterritorial restrictions.

The May 2026 Trump-Xi Beijing summit produced commitments from China to purchase at least $17 billion per year of US agricultural products plus an initial purchase of 200 American-made Boeing aircraft, plus a verbal commitment that “China will address US concerns regarding supply chain shortages related to rare earths.” The summit did not lift either MOFCOM suspension.^[18] Snap-back should be treated as the base-case risk absent a new deal by October 2026.^[19]

Japan December 2025 cutoff chronology + Sojitz/JOGMEC lock-up

The Precedent: Japan’s December 2025 Cutoff

In December 2025, China cut rare earth exports to Japan — specifically dysprosium, terbium, holmium, erbium, NdFeB alloys, gallium, germanium. Named target entities: Shin-Etsu, TDK, Proterial, Mitsubishi Heavy Industries, Mitsubishi Materials. The cutoff used the active 18/2025 layer alone. No re-arming of the suspended announcements was required. Country-by-country enforcement is the demonstrated operating mode.

Japan’s response: the Sojitz / JOGMEC consortium has locked up Lynas output under long-term arrangements running to 2038 — covering 5,000 tonnes per year of NdPr (the light rare earth magnet base) plus ≥50% of LAMP’s heavy rare earth oxide production (dysprosium and terbium, the GBD additives). That second number is the one that matters. NdPr is the bulk of the magnet, but it doesn’t survive heat without Dy/Tb diffusion. Japan didn’t just secure raw magnet volume — they secured the GBD additives that make their NdFeB magnets work in EV motors and defense actuators.

This puts Japan in direct competition with the Pentagon’s $96M Lynas heavy rare earth purchase agreement signed March 2026, drawing from the same nascent HREO output. LAMP’s HRE separation circuit is ramping, not at scale. Two sovereigns are now racing for the same scarce flow.

The pincer + flooding-tail-risk historical playbook

The Pincer

The dual MOFCOM suspensions expire November 10 and November 27, 2026. The NDAA § 854 mine-to-magnet ban takes effect January 1, 2027. If the first clock snaps back, extraterritorial restrictions on rare earth processing reactivate 52 days before the procurement ban takes effect. If only the second snaps, the window shrinks to 35 days. Either path simultaneously constrains supply and forces demand. The market is not pricing either interpretation.

A “China floods the market” scenario is not the base case because Beijing’s revealed strategy since 2023 has been licensing leverage, tracking, and selective scarcity — not price destruction.^[20] But it remains a real thesis-weakener and the historical playbook is well-documented. China bankrupted Molycorp (MP Materials’ predecessor) in 2015 by lifting quotas after the 2010 spike; flooded solar PV at predatory prices in the 2012 cycle; and crushed Western lithium juniors with oversupply in 2023–2024. The most lethal counter-play to the NDAA is not retaliation against the law — it is flooding the commercial channel below Western production economics, starving USAR, Energy Fuels, and Niron of the commercial volume they need to scale beyond the defense ration. Western buyers (Tesla, GM, GE Vernova) would object loudly to paying a 200–300% premium to subsidize domestic capacity, and the political coalition for sustained DPAS allocation would weaken. Calibrated easing remains a thesis-weakener; full flooding remains a tail risk worth pricing.

The NDAA ban is statutory — Congressional law, not an executive order. A trade deal cannot waive it. Trump cannot unilaterally override it. However, creative interpretation of nonavailability determinations, combined with liberal waivers, could functionally soften enforcement. This is a real but low-probability risk.

Bypass Paths Detail

Per-path facility, customer pipeline, and qualification-timeline detail behind the bypass-paths exhibit above.

Niron Magnetics — Iron Nitride Permanent Magnets

Iron nitride (Fe₁₆N₂) permanent magnets, zero rare earths. Founded 2013 with a US DOE grant, seeded after the 2010 China-Japan rare earth cutoff. University of Minnesota’s Jian-Ping Wang cracked the thin-film scaling problem using semiconductor manufacturing techniques. As of June 2026:

• Pilot plant in Minneapolis producing several tons per year
• 1,500 tpa Sartell, MN plant under construction (190,000 sqft, groundbreak Sept 26 2025, CEO target 1,500 tpa output by 2028)
• Second US plant in site selection at 10,000 tpa target (announced Mar 16 2026)
• 150-person organization, 150+ iron nitride patents (largest IP portfolio globally)
• Washington DC office opened Feb 11 2026; CEO Jonathan Rowntree testified to House Select Committee on Strategic Competition with the CCP on Nov 19 2025
• Named customer pipeline: Moog Inc. (NYSE: MOG.A) defense-actuator partnership (Dec 11 2025), MATTER variable-flux motor (Jan 2026), Bimotal motion, FaitalPRO loudspeakers (first commercial sales 2026)
• Stated end markets: AI Infrastructure, Aerospace & Defense, Robotics & Automation, Mobility, Industrial Motors^[22]

The honest caveat: Niron does not publicly disclose BHmax, max operating temperature, or coercivity vs. NdFeB. Some scientists remain skeptical of iron-nitride performance at high temperature under sustained mechanical load. Industrial qualification of a new magnet chemistry into a fielded weapons program (e.g. F-35 actuators via Moog) typically takes 18-36 months. Niron is a 2027-2028 relief path, not a Day 30 one. But it is building, not waiting.

ZF Friedrichshafen I2SM Motor — Magnet-Free EV Traction

A different solution to the same problem. ZF developed an inductively-fed synchronous EV traction motor that uses zero permanent magnets. BMW and Renault already ship a brushed wound-rotor analog; ZF’s I2SM eliminates the brushes via inductive coupling, producing a smaller, lighter motor with no rare earth dependence. ZF says ready for mass production. WSJ-cited analysts expect I2SM under the hood of an EV within a few years. If broadly adopted, this removes roughly 3 kg of NdFeB per EV traction motor from the demand pool — at the commercial layer where the bottleneck binds first.^[23]

Daido Steel — HRE-Free Rapid-Quench Magnet (Already Shipping at Honda)

Daido began R&D after the 2010 China-Japan REE cutoff. Daido Electronics uses a hot deformation method (distinct from the conventional sintering technique) that produces a fine-grain microstructure approximately ten times smaller than that of a sintered magnet — enabling high heat resistance without dysprosium or terbium. (Upstream rapid-quench magnetic powder is supplied by Canada’s Magnequench International.) The result is a permanent magnet with comparable performance to standard NdFeB but using zero heavy rare earths — no dysprosium, no terbium. Honda already uses this magnet in hybrid vehicle motors. Other named applications: semiconductor manufacturing equipment, MRI scanners, power steering, electric window motors. This is the most-relevant near-term relief path because it doesn’t replace NdFeB entirely — it just eliminates the Dy/Tb dependence that is the actual chokepoint inside the chokepoint. Light Nd from Australia and Mountain Pass stays available outside China.^[24]

The three paths attack the problem at different depths. Niron eliminates rare earths entirely (longest qualification, hardest physics). ZF eliminates the magnet (mature engineering, EV-specific). Daido eliminates only the heavy rare earths (already shipping at Honda, thermal-envelope-limited).

Neither Niron nor ZF nor Daido yet relieves the highest-performance defense applications, where the GBD-NdFeB thermal envelope still binds. The Pentagon’s DPAS lockout on Western GBD-NdFeB output remains the dominant scarcity force. But the commercial side has more relief in flight than the bear case credits. Under embargo, capital velocity changes — not deterministically, but materially. Optionality, not forecast: Tesla, GM, or GE Vernova facing multi-year commercial-channel supply starvation has a credible path to acquire or license one of the three bypass paths to brute-force qualification. The Minnesota workforce grant becoming a multi-billion-dollar emergency consortium build is one possible price-discovery scenario, not a base case.

Related Frameworks

How this casefile fits into the broader chokepoint taxonomy and adjacent statutory cascades.

Thesis Stress-Testing & Counter-Argument Response

Modeled against prevailing institutional bear cases on market volume, technical execution, and regulatory relief.

Methodology

The AXTI Pattern

ForcedAlpha’s pattern-match for small-cap structural-chokepoint re-rating — a template applied across casefiles, not specific to this trade. The fact pattern: an obscure small-cap controlling a critical bottleneck, asymmetric demand growth ahead of broad market recognition, and a re-rating event (capital, contract, or catalyst) that compresses the gap between fundamental position and market valuation.

Cascade construction methodology

This casefile was generated from the ForcedAlpha Supply Chain Intelligence Graph — a proprietary knowledge graph of 4,851 nodes and 21,311 edges mapping upstream material dependencies to downstream consumers across semiconductors, defense, energy, and critical minerals.

The cascade was surfaced by traversing outward from four seed nodes: the NDAA FY2024 Section 854 legislation node, the NdFeB permanent magnet material node, the China rare earth processing chokepoint, and the MOFCOM Announcement 61 legislation node. BFS traversal at depth 4 produced a subgraph of 55 nodes and 195 edges.

Load-bearing claims in this casefile are verified against primary sources; full claim-level validation of every statement is in progress. Primary sources include SEC filings (10-Ks, 10-Qs, 8-Ks, prospectuses, proxy statements), DFARS clause text, federal agency press releases (Commerce, DoD, DLA), company press releases and investor presentations, federal statutory and regulatory text, MOFCOM announcements, and contemporaneous reporting from named-byline business journalism (FT, WSJ, Bloomberg, Reuters, Nikkei).

The casefile carries claim-level confidence indicators throughout (Verified, Derived, Modeled). Where a claim is a synthetic estimate rather than a verified data point, this is labeled explicitly — for example, the 3,000–4,000 MT qualified-Western-capacity figure is a modeled synthesis, not the output of a producer survey.

The casefile is continuously monitored for status changes in its load-bearing claims. When evidence shifts, this page is updated and the dateModified metadata is bumped. Falsifiable conditions and enforcement regime probabilities are listed in the Tells & Falsifiers section.