LEU$203.63+0.8%Cap: $4.0BP/E: 52.152w: [====|------](Apr 19)
Widely-distributed syntheses this year have established the structural thesis. 60 USGS-designated critical minerals organized into seven end-use baskets — Battery, Magnet, Power Grid, Chip & Display, Defense, Agriculture, Industrial. China controls ≈85% of processing. Mining a mineral domestically does not safeguard national security if the United States remains dependent on a foreign country for the processing of that mineral — the exact language of Presidential Proclamation 10858 (January 14, 2026, Section 232). The framework is settled.
This piece is the execution layer beneath that framework. An intelligence report, not a synthesis. Six parts:
- Part 1 walks the physical supply chain by basket — named mines, named processors, named finished-product producers. Tesla's Robstown alkaline-leaching refinery is a vertical case study in Battery. Chinese refiners are listed by ticker (Shanghai / Shenzhen / Hong Kong), not by aggregate percentage. Japanese and Korean operators get equal time.
- Part 2 decomposes demand with the actual DOE 2022 Supply Chain Deep Dive numbers (Table 3: global NdFeB end-use percentages; Table 4: NdFeB grade × dysprosium content × max operating temperature). Orthogonal factor decomposition per basket.
- Part 3 maps ten U.S.-listed survivors against basket, supply-chain stage, and demand factor, with verbatim quotes from the relevant 10-K, 20-F, 6-K, and 8-K filings.
- Part 4 documents the allied midstream playbook — the Japan Australia Rare Earth (JARE) / Lynas structure running since 2011 that Pentagon copied for the MP Materials package in 2025. Korea Zinc Onsan. Mitsubishi Materials + H.C. Starck. Sumitomo Philippines HPAL. Proterial's Dy-free sintered NdFeB. This is the context the U.S.-centric framing omits.
- Part 5 covers private pipeline, M&A consolidation patterns, and the technical ceiling on Niron iron-nitride that bounds the NdFeB substitution risk to a specific ≈37-49% segment of demand — insulating the Dy/Tb heavy-REE thesis that drives the real supply-demand gap.
- Part 6 is the honesty section — what we cannot verify, what sits behind paywalls, what baskets we have no edge in, and the full verified China export control timeline as an appendix.
Sources throughout are tiered. Tier 1: SEC filings, Federal Register notices, Presidential Proclamations, USGS Mineral Commodity Summaries 2025, DOE supply chain reports, IEA Global Critical Minerals Outlook 2025, CRS reports. Tier 2: Industry analyst publications (Adamas Intelligence, Benchmark Mineral Intelligence, TrendForce, Project Blue — many fragments only; paywalls noted). Tier 3: Reputable trade press (Electrek, Nikkei Asia, Korea Herald, Inside Defense, Mining.com). Where a specific claim cannot be verified primary-source, the uncertainty is stated. Industry estimates are labeled as such.
One framework governs the execution layer: Paleologo's rule that a name with idiosyncratic variance below 75% on factor regression is sector beta in a critical-minerals costume. ETF is cheaper than active management fees for sector beta. Every name discussed in Part 3 either clears the 75% filter or is flagged where it does not.
Factor model methodology
The idiosyncratic variance percentages cited throughout this piece are computed from a trailing-60-day multi-factor daily-return OLS regression. Factor basket per name is SPY (market) + the most relevant sector ETF (REMX for rare earths; URNM for uranium; XME for broad mining; SMH or SOXX for semiconductors where applicable) + a style factor where it loads materially (HML/SMB on occasion). Regression specification: r_i,t = α + β_m · r_SPY,t + β_s · r_sector,t + (β_d · r_style,t if applicable) + ε_t. Idiosyncratic variance share is σ²_ε / σ²_total.
Two explicit caveats the reader should hold while evaluating the idio percentages:
-
Values above 100% are model noise, not literal precision. NB at 107% and USAR at 122% occur when a low-beta or near-zero-beta small-cap has trailing volatility that the factor model explains less than zero of — mathematically the residual variance exceeds total variance, which the OLS form tolerates but which has to be interpreted as "idio-dominant" rather than "107% idio." Pre-revenue names (USAR, NMG pre-FID), merger-arbitrage names (UUUU pending ASM close), and tiny-cap names with wide bid-ask spreads all tend to trip this. Treat any reported idio above ≈85% as "clearly idio-driven" rather than as a precise ratio.
-
Trailing 60 days is short enough to miss regime shifts. A name that prints 80% idio in April 2026 under one tape regime could print 55% in August 2026 if a sector-wide rally pulls correlations up. The filter is not a one-time classification; it needs re-running quarterly against current factor loadings. The specific values in this piece are April 18, 2026 snapshots.
Readers with access to Python + yfinance can replicate in ≈20 lines using statsmodels.OLS on daily log returns. Readers without: the directional conclusion (Defense-basket names, NB, USAR, LEU are idio-dominant; PESI, ULBI, LYSDY, IPX, MTRN are sector-mixed) is robust to model-specification choices.
Executive summary — ten-name screen
For readers who prefer the conclusion before the basket walk:
Ticker Basket Stage Primary thesis Idio Entry
(60d) discipline
LEU Defense + Midstream HALEU monopoly; $3.8B 86.6% Pullback to
Grid (enrichment) backlog to 2040; Fluor $170-180
EPC $24M cancel fee
NMG Battery Midstream BINDING Panasonic 13k 98% Pre-FID lottery
(anode AAM) tpa offtake Oct 2025; only; size small
pre-FID
PESI Defense Downstream Only US pure-play nuc 57% Below target;
(waste proc) waste; Hanford DFLAW size-down for
ramp; 2 analysts sector beta mix
ULBI Defense Downstream $110M backlog = 57.7% 54% Below target;
(defense batt) of revenue; Army Leader L3Harris 27%
Radio + FOSOV concentration
risk
ALM Defense Midstream Tungsten Sangdong 94% Post-Sangdong
(W mining/ commissioned Mar 2026; ramp confirmation
processing) $268M cash
IPX Defense + Midstream Ti sponge $160M non- 75% Right at filter
Aerospace (Ti processing) dilutive US gov; Titan threshold;
TN project watch scale-up
USAR Magnet + Downstream Stillwater Phase 1a Q1 122% Pre-first-shipment
Defense (magnet mfr) 2026; DFARS 225.7018 binary; size as
Jan 2027 tailwind option
LYSDY Magnet ADR Upstream + Only ex-China separator 64% Below target;
Midstream Dy/Tb; JARE 2038 PPA factor-sensitive
to China policy
NB Defense + Upstream Niobium 75% first-10yr 107% β = -0.04 counter-
Grid (multi-comm.) off-taken; β=-0.04 cyclical
diversifier
EU Defense Upstream US utility term contr; 76% Clean US uranium
(U ISR mining) 22.5% gross margin ISR
Three structural anchors the ten-name screen rests on: (1) MP Materials' DoW package is the best-structured binding contract in the US universe (covered in Part 5.1); (2) the allied midstream is a Japanese-Korean industrial network in 1:3 ratio to US-listed names (Part 4); (3) the Dy/Tb heavy-REE thesis is structurally insulated from Niron iron-nitride substitution because Fe16N2 physics cannot enter the 150-220°C magnet regime (Parts 1.2 and 6 Closing).
Exclusions documented in Part 3: WWR (SK On terminated Mar 2026), UUUU (late-phase tape, $0 REE revenue), CENX (insider distribution), ELBM (LGES non-binding per 20-F), SGML (hidden China default), LAC (factor bet on lithium), ATI/CRS/HWM/LMT/RTX (sector β mega-caps), ASX Pink Limited universe.
Part 1 — Supply Chain Intelligence by Basket
1.1 Battery Basket — Li, Co, Ni (Class 1), graphite, Mn
Minerals: lithium hydroxide (LiOH) and lithium carbonate (Li2CO3), cobalt sulfate, Class 1 nickel sulfate, natural flake graphite, synthetic graphite, high-purity manganese sulfate (HPMSM).
Mining geography, 2024 (USGS MCS2025, Benchmark Mineral Intelligence fragments):
Mineral Leading producers US share
Lithium Australia (Greenbushes, Mt Cattlin, Pilgan- <1%
goora, Wodgina) ≈50% | Chile (Salar de Ata- Silver Peak NV only
cama) + Argentina (Salar de Olaroz, Caucha-
rí, Pastos Grandes) ≈35% | China (Sichuan,
Jiangxi hard rock + Qinghai brine) ≈15%
Cobalt DRC ≈76% (Glencore Katanga, CMOC Tenke Fu- 0%
gurume, ERG, Eurasian Resources) | Indonesia
≈10% (HPAL co-product) | Russia (Norilsk)
Nickel (Class 1 Indonesia (laterite, growing very fast, ≈1% (Eagle MI)
for battery) Sulawesi + Halmahera) | Russia (Norilsk)
| Canada (Vale Voisey's Bay, Glencore
Raglan) | Australia (IGO Nova, BHP Mt Keith)
Natural graphite China (Heilongjiang + Shandong) ≈78% | 0%
Mozambique (Syrah Balama) ≈7% | Madagascar,
Tanzania, Brazil
Synthetic graphite China ≈85% Small
(coal-tar + needle coke feedstock)
Manganese South Africa ≈30% | Gabon (Comilog) ≈20% 0% since 1970
Australia ≈15% | China ≈12% | Brazil ≈8%
Processing geography, global shares of refined / battery-grade output:
Refined Li (LiOH + Li2CO3) China 61% | Chile 12% | Argentina 12%
| Australia ≈8% | US ≈0% pre-Tesla
Battery-grade Co sulfate China 71% | Finland (Umicore
Kokkola) ≈10% | Canada (Electra
BM, under construction)
Class 1 Ni sulfate China on-soil only 6% (Indonesia
dominates via Chinese-JV HPAL —
see 1.1.3)
Spherical purified graphite China 90%+ | Syrah Vidalia LA
(emerging, 11,250 tpa)
HPMSM (battery-grade Mn) China 97%
Source for refining shares: Benchmark Mineral Intelligence + IntelMarketResearch compiled chart widely circulated. The underlying company tonnage data is behind BMI paid subscription. USGS MCS2025 per-mineral chapters corroborate the directional shares.
Finished-product geography: Lithium-ion cell production is approximately 70% China, 15% Korea (LG Energy Solution, SK On, Samsung SDI), 7% Japan (Panasonic Energy), remainder Europe + North America. Cathode active material (CAM) and precursor (pCAM) concentrate further in China. Outside China: POSCO Future M, EcoPro BM, L&F (Korea); Umicore and BASF (Europe); Redwood Materials and Mitra Chem (private, US emerging).
1.1.1 Named Chinese lithium refiners
Ganfeng Lithium SHE: 002460 / HKEX: 01772
- 2024 production: 130,253 t LCE (world's largest by volume)
- HQ Xinyu, Jiangxi. Target 600,000 t/yr LCE by 2030
- Products: LiOH, Li2CO3, LiF, LiCl, Li metal, Li foil, Cs/Rb
compounds, organolithium
- Vertically integrated: resources in Australia, Argentina,
Mexico, China spodumene + brine → refining → precursor →
cell → recycling
Tianqi Lithium SHE: 002466 / HKEX: 09696
- 25.86% stake in Greenbushes via Windfield Holdings JV with
Albemarle (world's largest spodumene mine)
- Shehong refinery (Sichuan): 14,500 t LiCO3 + 5,000 t LiOH +
4,500 t LiCl + 200 t Li metal
- Kwinana WA refinery: 24,000 tpa LiOH Phase I ramped; Phase II
planned 24,000 tpa. TLEA JV 51% Tianqi / 49% IGO
IGO publicly stated "no path forward" mid-2024 — Phase II
operationally stalled
Sichuan Yahua SHE: 002497
- Historic primary LiOH supplier to Tesla
- Capacity 30,000 t/yr (2022) → 73,000 t (Feb 2023) → target
130,000 t/yr end-2025
- HQ Yibin, Sichuan. Also supplies LGC, Panasonic, LGES, SK On,
BYD
Source: Ganfeng 2024 annual report (publicly available on Ganfeng IR site); Tianqi 2024 annual report; Yahua official capacity disclosures via SMM and Moomoo financial data.
1.1.2 U.S. and allied Battery-basket additions 2024-2026
Tesla Robstown TX — operational January 2026. See the dedicated vertical case study in section 1.1.4.
Lithium Americas Corp (LAC) — Thacker Pass, NV. DOE ATVM loan of $2.26B with $867M drawn as of February 24, 2026 — first $435M drawn October 2025 (8-K 2025-10-01 Item 8.01), second $432M drawn February 2026 per 10-K 2026-03. Construction 93% engineering complete, 60% procurement complete per FY2025 10-K. GM holds a 20-year offtake for 100% of Phase 1 (40,000 tpa LCE). Trump administration restructured the DOE position October 2025: DOE received a 5% LAC penny warrant + 5% JV stake in exchange for $182M debt service deferral over the first 5 years. LAC is actively running a $250M ATM — sold 68M shares in FY2025 at an average of $5.98 and 32.5M shares in January 2026 at average $5.92; shares outstanding grew from 247M to 347M+ over 12 months, a 40% dilution. A subscriber looking at LAC is buying a lithium-price recovery trade into known future dilution, not a clean binding-contract idea.
Syrah Resources Vidalia, LA. 11,250 tpa spherical purified graphite anode material. DOE loan-funded construction. Binding Tesla offtake agreement. Feedstock from Syrah's Balama Mozambique flake mine. Syrah itself is ASX-listed as SYA.AX (Pink Limited on OTC — fails the OTCQX minimum filter for most U.S. investors).
Nouveau Monde Graphite (NMG) — Matawinie mine + Bécancour Battery Material Plant, Quebec. Binding Panasonic Offtake Agreement dated October 31, 2025, for 13,000 tonnes per year of active anode material (6-K filed 2026-04-14, line 932-934, verbatim: "'Panasonic Offtake Agreement' means the offtake agreement with Panasonic for the supply of active anode material from the 13,000 tons per year from the Battery Material Plants dated as of October 31, 2025."). GM is an equity investor (Investor Rights Agreement February 28, 2024), not a product offtake. This distinction matters — the critical-minerals press often conflates the two. NMG has not reached FID: the company closed a US$96.5M bought-deal equity offering on April 16, 2026, at US$1.84 per subscription receipt, and expects conditional access to US$335M senior project debt facilities, making the Phase-2 Matawinie financing stack roughly US$644M. Subscription receipts are escrowed pending shareholder approval by July 31, 2026.
Westwater Resources (WWR) — Kellyton, AL. Synthetic graphite anode facility under construction. SK On Products Procurement Agreement TERMINATED effective March 31, 2026 (8-K 2026-04-01 Item 1.02). Anchor customer removed; the Stellantis offtake had already been canceled November 2025. Kellyton construction is >50% complete with approximately $117M remaining capex and no committed replacement customer. The facility is structurally stranded pending either a new offtake or an asset sale.
Electra Battery Materials (ELBM) — Temiskaming Shores, Ontario, Canada. Cobalt sulfate refinery under construction. Feedstock binding: ERG cobalt-hydroxide agreement, Swiss/DRC origin, 3,000 tpa. The market-narrative LGES Cobalt Supply Agreement is EXPLICITLY non-binding per ELBM 20-F filed March 30, 2026: "no enforceable or binding obligation on either party." The same 20-F discloses a minimum-cash covenant on the October 2025 term loan — a distress signal. A name priced on the LGES headline is priced against a filing that explicitly disowns the headline.
South32 Hermosa, Arizona. Zinc-manganese-silver project. $166M DOE Battery Materials Development Grant + $20M Defense Production Act (DPA) Title III grant. HPMSM potential (battery-grade manganese sulfate). South32 trades as SZAAF on OTC but this is a Pink Limited tier and structurally does not support investable position sizes.
American Battery Technology Company (ABAT) — McCarran, NV. Operating Li-ion battery recycling hydrometallurgical facility. Active $144M DOE grant, of which only 1.4% ($2M) has been drawn as of December 31, 2025. A separate $57.7M DOE lithium-hydroxide refinery grant was TERMINATED October 2025 with only 10% ($5.9M) drawn — appeal filed October 10, 2025. Binding customer contract: Moss Landing Veolia agreement (8-K 2025-11-06 Item 1.01) for 3,126 metric tons (6.9 million pounds) of LG NMC battery material from the BESS fire site; time-of-essence clause.
1.1.3 Indonesian HPAL — the Chinese-JV chokepoint
Indonesia supplied ≈56% of global mined nickel in 2024, projected ≈2.4 Mt in 2025 (+7.7% YoY), ≈60% by 2025 full-year, ≈75% by 2030. The "China dominates battery nickel" story is not on Chinese soil — it is in Sulawesi and Halmahera, where Chinese operators hold majority ownership of essentially every operating high-pressure acid leach (HPAL) facility that converts Indonesian laterite ore to Class 1 battery-grade nickel sulfate.
Indonesia Morowali Industrial Park (IMIP) — Sulawesi. Tsingshan Group owns ≈49.7% + additional shares via a second subsidiary; remainder Bintang Delapan (Indonesian consortium with retired-military / bureaucrat leadership). Combined capacity 500-700 kt Ni/yr with additional 300-500 kt/yr planned. 53 RKEF lines with ≈4.2 Mt NPI + 4 Mt stainless billet + 7 Mt carbon steel installed.
Key IMIP HPAL tenants:
- PT Huayue Nickel Cobalt (HNC): Huayou Cobalt (SHA: 603799)-led JV with Tsingshan. 60,000 tpa Ni in MHP, expanding toward 90,000 tpa. First Indonesian HPAL, operational mid-2022.
- PT QMB New Energy Materials: GEM Co (SHE: 002340) + Tsingshan + Brunp (CATL affiliate). ≈50,000 tpa Ni in MHP.
- PT Excelsior Nickel Cobalt (ENC): Shanghai Decent (Tsingshan) lead, Nickel Industries (ASX: NIC) acquiring 55%. 72,000 tpa Ni across MHP + sulfate + cathode. First HPAL autoclave commissioned October 2025.
- PT Sulawesi Nickel Cobalt (SLNC): Huayou-led JV. Construction from January 2025, mid-2026 operational target. ≈60-90 kt/yr planned.
Indonesia Weda Bay Industrial Park (IWIP) — Halmahera. Projected ≈550,000 t Ni metal equivalent in 2025 — Indonesia's largest single nickel hub, surpassing IMIP. PT Weda Bay Nickel ore concession: Eramet 43% / Tsingshan 57% / PT Antam 10%. Park operator is Tsingshan with Huayou and Eramet partners.
Critical development: Indonesian government cut PT Weda Bay Nickel's 2026 RKAB (mine plan) quota to just 12 million wet metric tons from 42 Mwmt in 2025 — a 71% cut, reported via Metalnomist April 2026. This structurally tightens Indonesian laterite feed into IWIP and provides support for global nickel pricing into 2027.
Pomalaa HPAL — the FEOC test case. PT Vale Indonesia (IDX: INCO) site. Three-party equity: PT Vale + Huayou Cobalt + Ford Motor Co. Total investment IDR 67.5T (≈$4.5B). Target 120,000 t MHP/yr. As of December 2025: first two autoclaves arrived at site; three more expected. Commissioning pushed into 2027 (from 2026). FEOC (Foreign Entity of Concern) eligibility contingent on Chinese ownership remaining below 25% of the JV — not yet resolved with Treasury/DOE. Adjacent Sorowako HPAL (same Vale-Huayou structure, 60,000 tpa) construction expected to start 2026, identical FEOC contingency.
PT Halmahera Persada Lygend — Obi Island, North Maluku. Harita Group (Indonesian) + Ningbo Lygend (Chinese) joint venture. 60 kt Ni MHP currently, expanding to 120 kt. 52 kt Ni in sulfate + 6 kt Co in sulfate via downstream. The market claim: "world's biggest nickel sulfate factory."
GEM Co Indonesia — full nameplate 150,000 t Ni in MHP, the largest single-operator HPAL output. Q3 2025 shipments 35,939 t Ni (≈12,000 t/month run-rate). Cobalt co-product: 6,534 t in 9M 2025. September 2025: GEM signed with Indonesia's Danantara sovereign wealth fund for $1.4B HPAL smelter (planned 66,000 tpa Ni MHP). Chinese technology + capital + Indonesian state partner rather than private Indonesian oligarch. Signal: Indonesian political establishment moving into direct participation, which complicates future FEOC-compliant restructuring.
Tsingshan Holding Group. Private, founded 1988 by Xiang Guangda in Wenzhou. FY2024 group revenue ≈$56B (up from $53B 2022). Estimated ≈22% of global nickel output in 2021; share likely higher in 2025. The 2022 LME nickel squeeze context matters: Tsingshan held massive short positions hedging matte; price went $25k → >$100k/t in two days March 2022; LME suspended trading and canceled trades — an institutional intervention that effectively rescued Tsingshan and revealed that the Chinese nickel complex had crossed from economic to strategic concern for Western policymakers.
FEOC collision. Under Internal Revenue Code §30D (finalized December 2023, expanded by OBBBA July 2025), China, Russia, North Korea, and Iran are FEOCs. EV buyers claiming the $7,500 tax credit cannot source critical minerals (extracted, processed, or recycled) from a FEOC entity starting 2025, defined as ≥25% ownership, control, board seats, voting rights, or equity. Every operating Indonesian HPAL (Huayue, QMB, Lygend, GEM, Excelsior) is FEOC-excluded. Pomalaa-Ford-Huayou-Vale is the Pomalaa test case for staying under 25% Chinese control but unresolved. Translation: the wall of Indonesian nickel "flowing into" the IRA-subsidized US EV supply chain is structurally blocked. Two pricing regimes emerging — IRA-compliant nickel (sulfide + restructured Indonesian trickle) at a premium versus Chinese-JV Indonesian at global market.
Indonesian HPAL cobalt co-product: ≈8-12% ratio to Ni. Total Indonesian HPAL cobalt output: ≈25-35 kt Co in 2025, forecast ≈50 kt by 2027. Indonesia is becoming a structurally important cobalt supplier behind DRC — partially breaking DRC-cobalt dependence — but through the same Chinese-JV chokepoint. FEOC propagates.
Western sulfide alternatives (the "break-glass" option if FEOC-strict interpretation holds):
- Vale Voisey's Bay, NL Canada: underground expansion complete, ramping to 45 ktpy Ni in concentrate, full ramp H2 2026. Expansion cost $2.94B.
- Glencore Raglan, Quebec: ≈40 ktpy Ni in concentrate from three underground mines, 1.5 Mt ore/yr.
- BHP Mt Keith + Leinster, WA: ≈80-90 ktpy combined, but BHP Nickel West under strategic review with several assets placed on care-and-maintenance in 2024.
- IGO (ASX: IGO): Nova + Forrestania + Tawana JV; modest ≈30-40 ktpy, Western-compliant.
- Talon Metals (TSX: TLO / OTC: TLOFF): Tamarack MN pre-development, Tesla offtake for 75,000 t Ni over 6 years (≈$1.5B implied), contingent on commercial operations by January 1, 2026 (extendable one year, then Tesla can terminate). Not in commercial operation. A 34.9m massive sulfide intercept at 12.65% Ni announced May 2025 provides resource upside.
Global Class 1 sulfide production sits ≈600-700 ktpy. The U.S. EV market alone could need 400-500 ktpy by 2030. Sulfide cannot scale fast enough. Policy realistically lands on a minority-Chinese tolerated structure rather than pure exclusion.
1.1.4 Case study: Tesla Robstown alkaline-leaching refinery
The Tesla lithium hydroxide refinery at Robstown, Texas (near Corpus Christi) merits its own section because it changes the Western LiOH refining cost curve.
The facility. Groundbreaking May 2023 with Governor Greg Abbott and Elon Musk. Construction completed early 2025. First lithium hydroxide production late 2025. Fully operational January 2026 per Governor Abbott's office and Tesla's own confirmations via Electrek, Teslarati, and Spectrum Local News. Target throughput ≈50 GWh-equivalent battery-grade LiOH per year, approximately one million vehicles. First large-scale spodumene-to-LiOH refinery in North America at commercial scale.
The process innovation: alkaline leaching rather than sulfuric acid. Traditional Western LiOH refining uses sulfuric acid roasting — spodumene concentrate + H2SO4 yields lithium sulfate + waste, causticized to LiOH via lime + soda. Eight to twelve sequential unit operations. Large sodium sulfate (Na2SO4) waste stream per unit LiOH — a significant environmental liability and disposal-cost line that has made replicating Chinese refining economics in Western jurisdictions difficult.
Tesla's innovation is the elimination of sulfuric acid. Public description is an alkaline-route process: spodumene calcined in a kiln (standard α-to-β conversion at approximately 1000-1100°C), then alkaline leaching (caustic system), then purification to battery-grade LiOH. Six stages rather than eight to twelve. The byproduct is analcime (NaAlSi2O6·H2O), a sand-like sodium aluminosilicate that can be sold into cement and construction markets rather than disposed as hazardous waste. Two cost lines change simultaneously: reagent consumption and waste-handling capex/opex.
The reagent chemistry is not disclosed. Candidate flowsheets from underlying thermodynamics and from the January 15, 2025 TCEQ TPDES permit filing:
- NaOH high-pressure alkaline leach — most likely primary route. At elevated temperature, NaOH disrupts the β-spodumene lattice and liberates soluble sodium-lithium aluminosilicate species.
- Na2CO3 + Ca(OH)2 combination ("soda autoclave") — classical research route producing Li2CO3 first, causticized to LiOH.
- Pressure alkaline leach using NaOH/Na2CO3 mixture at ≈200°C.
Tesla has not published the specific reagent system; inference from the chemistry plus TCEQ permit language is analytical, not primary-source.
The sulfate tension deserves weight, not a footnote. The TCEQ TPDES permit (filed January 15, 2025) regulates discharge up to 231,000 gallons/day of treated wastewater to an unnamed ditch flowing to Petronila Creek. The permit's monitored analyte list explicitly includes sulfates as a regulated discharge parameter — which is unexpected for an "acid-free" process. Three plausible explanations, in decreasing order of innocence:
- Feedstock impurity carryover. Spodumene concentrate from Liontown, NAL, and Core contains trace sulfide minerals; sulfur reports to process water even in an alkaline flowsheet. Regulatable but not alarming.
- Cooling-tower chemistry. Industrial cooling systems use sulfate-based corrosion inhibitors independent of the main process. Normal discharge source; reportable even in a fully acid-free main process.
- Incomplete process changeover. The main alkaline route may be supplemented with sulfuric-acid unit operations at bottleneck steps (e.g., impurity precipitation, specific separations) that Tesla has not publicly described. The "acid-free" public framing would then be partially marketing — correct about the primary leach step, incomplete about the full flowsheet.
Inside Climate News reported in March 2026 that Tesla was discharging lithium refinery wastewater into a local ditch and that South Texas officials were unaware until the reporting surfaced. The company has not publicly reconciled the sulfate-monitoring line with the acid-free messaging. Resolution of this tension is material for the replicator thesis — if Western LiOH refining truly eliminates sulfuric acid at scale, the capex line for sulfuric acid plant + Na2SO4 disposal drops out of greenfield project pro formas. If the "acid-free" description is partially marketing, bankability of Nemaska / Piedmont / Covalent / other replicator projects at Western capex levels is less improved than it appears. Watch for Tesla's first 10-K / 10-Q disclosure of Robstown unit economics, any TCEQ enforcement action, and Nemaska's own flowsheet technical disclosures when/if they reach the FID stage.
Feedstock. Tesla's original 2022-2023 spodumene supply chain included Core Lithium (Finniss, NT Australia) + Liontown Resources (Kathleen Valley, WA) + Piedmont Lithium / Sayona Mining North American Lithium JV (Québec). Core went to care-and-maintenance January 2024 but secured $205M restart funding February 2026 ($70M Glencore + InfraVia convertible note, $50M Nebari senior secured, A$120M equity) — not distressed out of the chain. Tesla's current publicly-reported feedstock is primarily:
- North American Lithium (NAL), Québec, Canada — Piedmont Lithium (PLL) / Sayona Mining JV. Operating. Nearest-shore supply.
- Liontown Resources Kathleen Valley, WA (ASX: LTR) — operating since July 2024, 5-year binding agreement commencing 2024, 500,000 t/yr spodumene concentrate Phase I scaling to 700,000 t/yr by year 6.
- Restarting tertiary supply from Core Finniss (Glencore bought stockpile February 2026 per Bloomberg).
Hard facts verified: the facility is operational; the process is described as "acid-free" by Tesla and corroborated by trade press; throughput targets are stated but not yet validated against actual run-rate disclosures. Unit economics, yield, reagent consumption per kg LiOH, and the full environmental balance are undisclosed.
Structural implication. If the alkaline route scales at Robstown, the cost curve of ex-China LiOH refining compresses materially. This helps upstream lithium producers with binding Western offtakes (LAC Thacker Pass with GM, NMG with Panasonic for anode) because the midstream they must pass through gets cheaper to build. It mixed for mid-stage Chinese converters because it narrows their cost advantage. There is no public vehicle for direct exposure to Tesla's refining economics — TSLA equity is dominated by auto and energy segments. The second-order read is that Western LiOH refining capex becomes more bankable. That helps Nemaska Lithium (private, Quebec), Piedmont Lithium (PLL, North Carolina), Covalent Lithium (private, Western Australia). Watch Robstown quarterly run-rate disclosures and any replicator announcements.
1.2 Magnet Basket — NdFeB and SmCo
Minerals: neodymium (Nd), praseodymium (Pr), dysprosium (Dy), terbium (Tb), samarium (Sm), cobalt (Co). Plus gallium as a small additive in high-temperature NdFeB grades, and boron and iron as structural inputs (not supply-constrained). The distinction that actually prices the thesis: light rare earths (Nd, Pr) are the bulk of magnet mass; heavy rare earths (Dy, Tb) are additives required for high-temperature performance — and not substitutable at current specs.
Mining geography:
NdPr (light REE) China 86% (Bayan Obo + Maoming) | US 6%
(Mountain Pass, MP) | Australia 7% (Mt
Weld, Lynas) | rest ≈1%
Dy / Tb (heavy REE) China ≈85% | Myanmar (Kachin State, illegal
flow via Yunnan crossings) ≈10% | scattered
remainder. No US primary production.
Sm China-dominant, small global volumes.
Myanmar heavy REE flow matters. Myanmar was the world's third-largest REE producer as of 2023. Tonnage trajectory: 19,500 t HREE oxides pre-coup (2021) → 12,000-14,000 t post-coup disruption (2022) → 41,700 t HREE oxides valued $1.4B in 2023 as China diversified away from Chinese mine consolidation toward Kachin sourcing. ≈31,000 t total REE in 2024. Active sites grew from 130 (2020) to 370+ (end-2024). Cumulative 2017-2024: >290,000 t, >$4.2B value, with 85% of that value post-coup. In October 2024 the Kachin Independence Army (KIA) seized Chipwi and Pangwa — central HREE mining towns. Chinese buyers subsequently negotiated an arrangement with KIA: ore exports at 35,000 yuan/ton + 20% tax. Transit runs via Ruili and Tengchong (Yunnan–Myanmar crossings) into Ganzhou separators. Up to two-thirds of global heavy rare earths now originate in Myanmar and pass through Yunnan. Sources: Stimson, Jamestown Foundation, New Security Beat, Asia Times, Asia Financial.
Separation and processing — post-2021 Chinese consolidation:
Critically, the Chinese REE separation industry consolidated in December 2021. By 2024-2025, all REE mining and smelting-separation quotas go to just two entities:
China Northern Rare Earth Group (High-Tech) SSE: 600111
- LREE monopoly; Baotou-centered (Bayan Obo iron ore co-mined
REE deposit, world's largest)
- Controlled by Baogang Group
- Baotou Huamei subsidiary: smelting-separation capacity rising
to 106,000 t/yr; total Baotou city separation capacity target
234,000 t/yr end-2024
- 2021 mining and smelting quota ≈60%
- Building 4,000 t/yr NdFeB scrap recycling (Baotou)
China Rare Earth Group Co., Ltd. Unlisted SOE
- Formed December 2021 by merger of China Minmetals RE,
Chinalco RE & Metals, China Southern RE, Ganzhou Zhonglan,
Jiangxi Ganzhou RM Exchange. Xiamen Tungsten's RE quota
folded in via Sept 2023 JV
- HREE consolidator; ≈30% of total Chinese REE; 60-70% of
national HREE supply (higher post-2024 rollups)
- Feedstock: ion-adsorption clay from Jiangxi, Guangdong,
Fujian, Hunan + Myanmar imports via Yunnan
Shenghe Resources Holding SSE: 600392
- Sichuan-based LREE separation + NdPr metal specialist
- Holds ≈25% of Chinese domestic separation quota
- Leshan Shenghe: bastnäsite processing ≈5,500 tpa REO
- Ganzhou Chenguang subsidiary: acquired Peak Rare Earths
(Tanzania Ngualla) in 2025 — international expansion
- HISTORICAL NOTE: Shenghe has tolled MP Materials' Mountain
Pass concentrate. MP ceased all China product sales July 2025
under DoW agreement covenants.
Grirem Advanced Materials Subsidiary of GRINM (SOE)
- State-owned via GRINM (China National Research Institute of
Nonferrous Metals)
- Claims >65% of global REE smelted/separated using Grirem-
licensed process technology — IP footprint enormous, branded
tonnage smaller than SOE majors
Xiamen Tungsten (RE segment folded into CRE) SSE: 600549
- Integrated tungsten/molybdenum/RE; RE segment folded into CRE
Group 2024 via September 2023 JV
Western policy papers referencing "multiple Chinese producers" are structurally out of date. Two entities receive the quotas.
Sintered NdFeB magnet production, 2024 (IEA Global Critical Minerals Outlook 2025, p.171-172):
- China produces ≈94% of sintered NdFeB. Bonded NdFeB ≈80%.
- Total global NdFeB magnet consumption 2024: ≈259,000 tonnes (Adamas Intelligence public commentary).
- Total magnet REE demand 2024: 90 kilotonnes REO (IEA, p.170).
- Growth rate 2023: +13.3% YoY NdFeB consumption.
- Forecast CAGR 2024-2040: 8.7% (Adamas).
Named Chinese sintered NdFeB producers:
JL MAG Rare-Earth SHE: 300748 / HKEX: 06680
- 2024 production: 29,300 t blanks (+39% YoY); 20,900 t
finished (+38% YoY)
- Capacity target 40,000 tpa end-2025; 60,000 tpa by 2027
- 4 China sites (Ganzhou HQ 284 acres + 15,000 tpa integrated
line, Baotou, Ningbo) + Mexico plant under construction
- Single most important Chinese magnet producer for Western EV
OEM supply chains
Zhong Ke San Huan SZSE: 000970
- Founded 1999 by Beijing Sanhuan New Materials (Chinese
Academy of Sciences affiliation)
- JVs: SANVAC with VAC Germany, Seiko-Epson (bonded Japan)
- Plants: Ningbo Konit, Tianjin Sanhuan Lucky (>12,000 t/yr
alone), SANVAC Beijing, Ganzhou, Shanghai San Huan
- Announced Dy-free high-performance NdFeB R&D — potential
workaround if HREE stays constrained
Ningbo Yunsheng SSE: 600366
- 21,000 t/yr as of 2024 (up from 15,000 t/yr 2022)
- Baotou Phase II: +15,000 t/yr end-2024
Earth-Panda Advanced Magnetic SSE STAR: 688077
- Founded 2003. Sintered NdFeB + SmCo (rare in China) + magnetic
components
- 24,000 t/yr end-2022; 36,000 t/yr target end-2026
- 4 China sites + Germany
- SmCo capability matters for high-temp defense
Ex-China named sintered NdFeB producers (the allied backbone):
- Shin-Etsu Chemical (TYO: 4063). Takefu Plant (Fukui, acquired 1945) for primary sintering. Shin-Etsu Magnetic Materials Vietnam (SMMV) at Hai Phong: 2,200 tpa (doubled from 1,100 via ¥5B investment completed 2018). Industry estimate: Takefu 2,500-3,500 tpa, Vietnam 2,200 tpa, total ≈5,000 tpa — mark as estimate; Shin-Etsu does not disclose NdFeB capacity discretely.
- Proterial Ltd (formerly Hitachi Metals, NEOMAX brand). DELISTED March 1, 2023 via Bain Capital + Japan Industrial Partners + Japan Industrial Solutions consortium tender offer at ¥816.8B ($5.6B initial, $7.5B TEV). Rebranded January 3, 2023. Origin of the Sagawa NdFeB patent family (1982). July 22, 2025: Proterial disclosed heavy-rare-earth-free sintered NdFeB magnets (NMX-F1SH-HF and NMX-G1NH-HF) with Br=1.40 T, HcJ≥1,671 kA/m, rated to a 180°C operating window — competitive with Dy/Tb-containing grades in temperature-critical EV drive motor applications. Mass-production plant accepting pre-production sample requests from April 2026. Pre-delisting Hitachi Metals capacity was in the 1,500-2,000 tpa sintered range; directionally stable post-delisting.
- TDK Corporation (TYO: 6762). NdFeB manufacturing offshored to China via Guangdong Rising TDK Rare Earth Advanced Material Co. JV (TDK ≈59%, $33M initial investment). 1,500 tpa design, Chinese-jurisdictional — counts against allied supply for DFARS 225.7018 purposes.
- Daido Steel / Daido Electronics (TYO: 5471). Hot-deformed NdFeB (not sintered) branded HYQNEO at Nakatsugawa, Gifu. Hot-deformed uses nanometer-scale crystal alignment enabling heavy-rare-earth-free magnets. First commercial deployment: 2016 Honda Freed hybrid (world-first HRE-free drive motor magnet adoption), METI-subsidized. Low hundreds tpa range.
- Neo Performance Materials (TSX: NEO / OTCQX: NOPMF). Narva, Estonia sintered magnet plant grand-opened September 19-22, 2025. Phase 1A design capacity 2,000 tpa. Magnequench (subsidiary) holds ≈80% global bonded NdFeB powder share, operating from Korat Thailand + China. 30% of Neo's revenue remains China-exposed; Russian feedstock historically at Sillamae Estonia rare metals facility. Cumulative worldview log Bayes +2.02 bullish on operating-name idiosyncratic alpha.
- Vacuumschmelze (VAC) (Germany). Private (Apollo Global Management), Hanau. Neorem (Finland) subsidiary. Traditional European allied NdFeB.
- USA Rare Earth (USAR) (NASDAQ). Stillwater OK Phase 1a commissioned Q1 2026 per 10-K filed March 30, 2026 (line 551-557). First customer shipments targeted Q2 2026. Full capacity 600 MTPA for Phase 1a plus 600 MTPA for Phase 1b = 1,200 MTPA total. Pre-revenue for magnets — FY2025 revenue of $1.643M is sourced entirely from Less Common Metals UK metal-making operations acquired November 2025 per 10-K line 1975.
- MP Materials (MP). Fort Worth TX Independence facility — 1,000 tpa commissioning 2025. 10X Facility project under DoW package: 3,000 → 10,000 tpa by 2028.
- Noveon Magnetics (private, San Marcos TX). $215M Series C January 2026 (OneIM lead $200M, with LG Electronics + Kangwon Energy as Korean strategic partners). Producing DFARS-compliant sintered NdFeB at commercial scale with ABB under a binding Long-Term Agreement. This is the incumbent US commercial sintered NdFeB producer — a fact that tempers the "USAR is the only US NdFeB vehicle" narrative. USAR is the only public US NdFeB vehicle.
- Niron Magnetics (private, Sartell MN). Iron-nitride (Fe16N2) rare-earth-free magnets. Full technical analysis in section 1.2 below. Sartell facility groundbreaking September 29, 2025; operational target early 2027. $214M cumulative funding across 8 rounds from 18 investors (Samsung Ventures, GM Ventures, Stellantis Ventures, Volvo Cars Tech Fund, Allison Transmission, Magna, Shakopee Mdewakanton Sioux Community, University of Minnesota). Selected under Section 48C tax-credit allocation in 2025.
The Niron physics ceiling. Iron nitride is a genuine rare-earth-free magnet chemistry, but the substitution envelope is bounded by physics, not engineering. Per peer-reviewed literature and Niron's then-CEO Andy Blackburn (E-Mobility Engineering interview; current CEO is Jonathan Rowntree as of 2025):
Property Theoretical Niron Gen 2 Niron 2025 N42 N42EH
ceiling target achieved (baseline) (EV
traction)
BHmax (MGOe) 135-143 36 ≈10 42 42
Br (T) 2.39 1.5 ≈1.0 1.30 1.28
Hci (kOe) 2.5-10 3.75 2.1 ≥12 ≥30
Max op T (°C) — NOT DISCLOSED 82-150 80 200
(inferred)
Dy content (%) 0 0 0 <0.5 8.5-11
Three tiers of numbers. Niron's production-achieved BHmax is below commodity N35 NdFeB (≈35 MGOe); the 2025-achieved product resembles AlNiCo and hard ferrite more than modern NdFeB. Blackburn publicly stated in the E-Mobility Engineering interview: "We think the maximum possible coercivity is between 4000 and 5000 oersteds, compared with around 12,000 oersteds or more for neodymium." He further acknowledged that iron nitride coercivity "drops off as temperature rises," reinforcing the thermal ceiling argument below. This is the company's own public CEO statement of the hard physics limit — anisotropy field Ha for Fe16N2 is ≈10 kOe versus Nd2Fe14B's 76 kOe.
Thermal stability from peer-reviewed primary sources (Bao et al. 2019, JMMM 2020, MDPI Metals 2024, Dirba UK Magnetics Society 2025): Fe16N2 oxidation onset 160°C in air, decomposition onset 200°C in inert atmosphere, must be kept below 82°C for 99% stability over 100 years (activation energy 199 kJ/mol). 82°C is below M-grade NdFeB (100°C). The max operating temperature Niron has NOT publicly disclosed is the decisive gate for EV motor qualification; the production-grade service temperature is inferred <150°C from decomposition kinetics.
Practical segment-level assessment (demand decomposition in Part 2):
NdFeB end-use Fe16N2 risk
------------------------------------------------------------
Consumer electronics HIGH — matches <150°C M-grade demand
Bonded magnets HIGH — same customers as Magnequench
Non-drivetrain auto HIGH — <120°C typical
Industrial motors MEDIUM — duty-cycle dependent
Other sintered (e-bikes) HIGH — low-temp
EV traction motors VERY LOW — 180-220°C EH/AH grade,
15-year warranty, Dy-doped
Offshore wind direct- LOW — 150°C SH-grade, 25-year life,
drive marine environment, 4% Dy
Niron threatens approximately 37-49% of NdFeB demand (consumer electronics + bonded + non-drivetrain auto + other sintered) on a 5-10 year horizon if Gen 2 ships and scales past the Sartell 1,500 tpa nameplate. Niron does NOT threaten the 20-30% of demand driving growth (EV traction + offshore wind direct-drive + SH-grade industrial). The heavy REE (Dy/Tb) supply thesis is structurally insulated from Fe16N2 substitution because Dy/Tb consumption is concentrated in the 150-220°C magnet regime that Fe16N2 physics cannot enter.
Scale reality: Sartell's 1,500 tpa nameplate = 0.58% of 2024 global NdFeB or 0.37% of 2030 global. Even a 10x expansion to 15,000 tpa by 2035 is 3-4% of the market.
The real Dy demand pressure is Chinese grain-boundary diffusion (GBD), not Niron. GBD reduces Dy loading 30-50% in M/H/SH grades at Chinese and Japanese manufacturers 2022-2030. Toyota's La/Ce substitution work. Dual-alloy magnets. These are the actual Dy-demand pressures that matter. And Proterial's July 2025 Dy-free sintered announcement (Br=1.40 T, HcJ≥1,671 kA/m at 180°C) is a more credible credible threat to Dy demand than Niron because Proterial's chemistry is sintered-class, EV-motor temperature-qualified.
The policy variable: DFARS 225.7018. Final rule, DFARS Case 2021-D015, published May 30, 2024 Federal Register. Effective January 1, 2027:
- DoD prohibited from acquiring NdFeB or SmCo magnets with any part of the supply chain (mining → finished magnet) sourced from Iran, North Korea, Russia, or China.
- Through December 31, 2026 the prohibition applies only to "melting and producing" — China-mined material legally usable by DoD contractors until year-end 2026.
- COTS exception: previously exempt unless ≥50% tungsten by weight; effective January 1, 2027 narrows to exempt only when ≥50% of any covered material by weight.
The regulation is final, not aspirational. Enforcement mechanism: DFARS 252.225-7052 is the corresponding contract clause incorporated into individual DoD contracts — non-compliance is a contract default event, not merely a policy violation. The open question is whether DoD grants Domestic Non-Availability Determinations (DNADs) or programmatic waivers under the framework of FY07 NDAA Section 842 (which authorizes COTS exceptions to specialty metals provisions). Historical specialty metals compliance under the parallel DFARS 252.225-7009 clause has included multiple program-specific DNADs for items where domestic qualified sources did not exist — individual determinations appear in the Federal Register and program-specific notices but GAO has not published a cumulative waiver database. The most plausible 2027 outcome is DNADs for specific defense programs where US + allied magnet supply is insufficient (Columbia SSBN magnet content, early Sentinel deliveries), coupled with binding supply expansion at domestic producers (MP 10X Facility, USAR Stillwater, Noveon) — mixed enforcement with program carve-outs rather than absolute prohibition. Monitor the first post-January-2027 defense contracts for specific DNAD patterns.
China's April 2025 counter-move: MOFCOM + GAC Announcement No. 18. Seven medium and heavy rare earth elements (samarium, gadolinium, terbium, dysprosium, lutetium, scandium, yttrium) plus their alloys, oxides, compounds, mixtures, and permanent magnet materials added to China's dual-use export control list. October 2025: Announcements Nos. 55-62 added 12 additional categories (five magnet types, select graphite, battery CAM equipment). November 7-9, 2025: Announcement No. 72 suspended Article 2 of No. 46 until November 27, 2026 — Ga/Ge/Sb/superhard to US restored under general licensing; six October 9 announcements suspended until November 10, 2026. Direct implication: the April 2025 HREE inclusion is still operative for Dy/Tb and the other five HREE.
1.3 Defense Basket — W, Sb, Ti sponge, Ta, Nb, PGMs, specialty alloys
The Defense basket is structurally different from the other baskets: demand is budget-driven, not price-driven. Procurement is on multi-year program-of-record cadences. Price inelasticity is high.
Minerals and military applications:
Tungsten Kinetic penetrators (APFSDS), shaped-charge liners,
armor, radiation shielding, cutting tools. China 83%
of mined output; US zero since 2015.
Antimony Flame retardants in munitions + cabling; alloying
agent in primer compositions (lead-antimony); sulfide
form for PV cells and IR detectors. China ≈60% mine;
December 2024 full export ban to US; Nov 2025
suspension of that ban until Nov 27, 2026.
Titanium sponge Aerospace engines, airframes, naval alloys, medical.
Japan ≈40% (Toho, Osaka), Kazakhstan (UKTMP), China,
Russia (VSMPO-AVISMA, sanctions risk). US effectively
zero primary since 2020 (Henderson NV idled, Rowley
UT idled 2016).
Tantalum Missile guidance + radar + avionics capacitors +
superalloy additive. DRC + Rwanda + Australia mining;
refining concentrated in China + H.C. Starck (now
Mitsubishi Materials, Germany). US zero since 1959.
Niobium HSLA steel for pipelines, naval hulls, auto structure;
superalloys. Brazil (CBMM) ≈92%.
PGMs Autocatalyst + electronics + specialty chemicals +
H2 electrolysis. South Africa 70%+, Russia (Norilsk)
≈25%. US: Sibanye-Stillwater (SBSW) Stillwater MT,
sole US producer.
Specialty Superalloys for jet engines, rocket motors, high-
Ni/Fe temperature defense applications. ATI, Carpenter
superalloys Technology (CRS), Howmet Aerospace (HWM). Haynes
International acquired by Acerinox November 2024.
Universal Stainless acquired by Aperam January 2025.
1.3.1 The federal demand signal — appropriation, not press release
Via the One Big Beautiful Bill Act (OBBBA) Section 20004, Congress appropriated $7.5 billion to critical minerals programs: $2 billion to the National Defense Stockpile (NDS), $5 billion to the DoD Industrial Base Fund, $500 million to a DoD critical minerals loan program supporting up to $100 billion in loans/guarantees. DoD communicated intent to acquire up to $1 billion of stockpile materials in 2025. Disclosed allocations:
- Cobalt: up to $500M
- Antimony: $245M
- Tantalum: $100M
- Scandium: $45M
- Tungsten, bismuth, indium: DLA Strategic Materials RFIs issued; specific dollar quantities undisclosed publicly.
But the acquisition ceilings are in the Annual Materials Plan, not in the press releases. The FY25 AMP is published at Federal Register 88 FR 60633 (09/05/2023, Docket BIS-2023-0018). FY26 AMP at 89 FR 70166 (08/29/2024, Docket BIS-2024-0030). Actual authorized-acquisition ceilings FY25 → FY26:
Material FY25 AMP FY26 AMP Change
Antimony 700 MT 700 MT flat
Tantalum 64,500 lbs 64,500 lbs flat
Tungsten 4,500,500 lbs 587,000 lbs -87%
Titanium 15,000 ST 13,608 MT re-expressed
Nd-Pr oxide 300 MT 300 MT flat
NdFeB magnet block 450 MT 450 MT flat
Samarium-cobalt alloy 60 MT 60 MT flat
Lanthanum 1,100 MT 1,100 MT flat
Ferroniobium 300,000 lbs 300,000 lbs flat
Iso-molded graphite 1,700 MT 1,700 MT flat
Electrolytic manganese metal 5,000 MT 5,000 MT flat
Magnesium 3,500 MT 3,500 MT flat
High-purity aluminum 1,700 MT 1,700 MT flat
Zirconium / Zr-Hf 2,300 MT 2,300 MT flat
Grain-oriented electrical steel 3,200 MT 3,200 MT flat
Seven observations from the AMP data that the press headlines obscure:
- $245M antimony at AMP ceiling (700 MT/yr × ≈$40K/MT ≈ $28M/yr) is a 9-year spend at AMP pace. $245M of antimony ≈ 6,125 MT. Multi-year buildout constrained by AMP cadence, not a single-year event.
- Tungsten ceiling collapsed -87% FY25 → FY26. Either FY25 partially executed, or DLA signals that Almonty (Sangdong) + Perpetua + 6K Additive + Fireweed (Mactung) domestic onshoring will meet demand. Parallel expansion of tungsten-rhenium Conversion (5,000 kg, new line) supports the pivot from ore stockpiling to specialty-alloy reprocessing.
- Stockpile universe grew 57 → 61 materials between FY25 and FY26. Four new materials added in 12 months.
- Scandium is NOT listed in either year's AMP Attachment 1. The $45M "scandium" headline is DPA Title III / DIB Fund programmatic, not NDS acquisition. Most likely beneficiary: NioCorp (NB) via the December 4, 2025 FEA Materials acquisition that produces Al-Sc master alloy.
- Cobalt acquisitions are ZERO in both AMPs. The only cobalt line is Recovery from Government Sources: 50,000 lbs (FY25) → 500 MT (FY26), a 22× increase but recycling, not new procurement. The "$500M cobalt" press funds DPA Title III awards, not NDS purchases.
- Rare earth ceilings are small. Nd-Pr oxide 300 MT/yr; NdFeB magnet block 450 MT/yr; Sm-Co alloy 60 MT/yr; lanthanum 1,100 MT/yr. Government demand alone cannot sustain a domestic rare earth industry; commercial offtake (USAR, Lynas, MP Materials) must drive the economics with DLA as purchaser of last resort.
- Dysprosium and terbium oxide are NOT listed at all in either AMP despite China export controls. No DLA stockpile acquisition authority for heavy rare earths. Vehicle has to be DPA Title III — or the Lynas Seadrift $258M commitment. Per Lynas CEO Amanda Lacaze on the August 2025 FY2025 results call, there is "significant uncertainty as to whether construction of the Seadrift processing facility will proceed, and if so, in what form." Subscribers reading this as "unlikely to proceed" should know the primary-source language is "might not proceed."
Germanium appears on three of four ledger sides (Disposal 5,000 kg, Conversion 5,000 kg, Recovery 5,000 kg) but NOT as an Acquisition line. Rotation-only. DLA is managing inventory, not net-accumulating.
Per-mineral confirmed DPA Title III awards (separate from NDS):
Company Mineral Amount Date
Nova Minerals Antimony (Alaska Range) $43.4M 2026
US Antimony (UAMY) Antimony extraction / processing $27.0M 2025
Fireweed Tungsten (Mactung) $15.8M 2026
Guardian Metal Tungsten (Pilot Mountain) $6.2M 07/2025
6K Additive Ti/Nb/Ni/W reprocessing undisclosed 2025
South Star Battery Graphite (Alabama) $3.2M 2025
Perpetua Resources Antimony + gold (Stibnite) $80M+ grants
+ $2.7B EXIM ongoing
1.3.2 The demand math from platforms
CRS R41744 (Rare Earth Elements in National Defense, September 17, 2013, Valerie Bailey Grasso) is the foundational primary source, reproduced in USGS MCS2025:
Platform Rare Earth Content
F-35 Lightning II 920 lbs (417 kg) per airframe
DDG-51 Arleigh Burke 5,200 lbs (2,359 kg) per hull
SSN-774 Virginia 9,200 lbs (4,173 kg) per submarine
These figures are from 2013, pre-F-35 Block 4, pre-Flight III DDG with SPY-6 GaN radar, pre-Block V Virginia with Virginia Payload Module. Subscribers should apply +25-50% uplift for post-2020 configurations; 2013 figures are defensible floor, not current-generation.
F-35 program annual pull-through: 150 aircraft/year × 920 lbs = 62.6 metric tons/year REE — a single platform at 1.5-2.1% of total DoD annual permanent magnet demand.
F-135 engine per unit content (Pratt & Whitney, ≈1,701 kg dry weight): Cobalt 5-12% in superalloy → 100-200 kg Co per engine; rhenium 3-6% in single-crystal HPT blades → 2-5 kg per engine; tantalum 3-8% → 5-15 kg per engine; hafnium 0.1-1.5% → 0.1-1 kg per engine; titanium ≈500-700 kg; nickel ≈850-1,020 kg. F-35 production of 150 airframes/year × 1 engine = a single engine program consuming ≈22 tonnes Co/year.
F-47 NGAD (Boeing): Contract awarded March 21, 2025, $20B to Boeing. First sixth-generation fighter, replaces F-22. Fleet plan 185 units. Per SFA Oxford: REE content comparable to F-35 baseline, likely higher for additional sensor/EW load. Fleet program REE total ≈170,000 lbs (77 metric tons). First deliveries realistic mid-2027.
Virginia-class SSN-774: S9G reactor (BWXT sole manufacturer) pressurized water, ≈210 MWt, 40,000 shp. Fuel: ≈93% HEU, life-of-ship 33-year core, no refueling. Virginia does NOT use HALEU — naval propulsion is one of the last remaining HEU applications globally. Each core estimated 100-500 kg HEU (exact classified). Production: 2/year target, currently ≈1.3/year actual; AUKUS adds 3-5 SSNs for Australia over the 2030s. Pull-through 2/year × 9,200 lbs = 18,400 lbs = 8.3 tonnes REE/year.
Columbia-class SSBN: S1B reactor (≈220 MWt), life-of-ship 42-year core. Columbia displaces 20,815 tons. Scaling from Virginia (9,200 lbs at 7,800 tons) is non-linear; defensible range 15,000-25,000 lbs per SSBN. 12 SSBNs total; approximately one per year 2028-2042. Columbia is the first SSBN with integrated turbo-electric drive (versus Ohio-class mechanical), so additional permanent-magnet content. Program midpoint REE ≈240,000 lbs = 108 metric tons.
Ford-class carrier: Two A1B reactors per hull. EMALS (General Atomics) uses linear induction motors — not permanent magnets — but the kinetic-energy storage flywheel motor-generator requires substantial NdFeB or SmCo. Four flywheels per carrier, each storing 121 MJ.
DDG-51 Flight III (HII Ingalls + GD BIW): The critical step change is the radar. Flight IIA has AN/SPY-1D PESA with GaAs T/R modules. Flight III (DDG-125 USS Jack H. Lucas onward) uses AN/SPY-6(V)1 AMDR with GaN T/R modules, generating 35× more radar power at 2× electrical input with 1,000× more power efficiency. 37 Radar Module Assemblies × 4 faces = 148 RMAs per ship. Flight III procurement ties directly to gallium demand — and China has restricted gallium exports since December 2023, with the November 2025 suspension of Announcement 46 Article 2 holding only until November 27, 2026. Production: 2-3 DDGs/year. Pull-through 3/year × 5,200 lbs = 15.6k lbs = 7.1 tonnes REE/year.
Constellation-class FFG-62: Program cut — 4 of 6 hulls CANCELLED November 2025. Only FFG-62 and FFG-63 proceed. AN/SPY-6(V)3 EASR on survivors — 3-face GaN radar.
Missile replenishment cadence:
-
PAC-3 MSE (Lockheed Martin): FY25 $9.8B for 1,970 PAC-3 MSE (September 2025). FY26 framework: tripling from 500/year to 2,000/year by end-2030 (Pentagon/Lockheed joint press January 2026). 2025 deliveries: 600+ (20% YoY). Per round: 5-10 kg W Hit-to-Kill warhead + several hundred grams NdFeB in 40+ attitude control motors + Ta capacitors throughout Ka-band seeker + Ti airframe. At 2,000/year steady-state: 15-20 tonnes W + 1-2 tonnes NdFeB + hundreds of kg Ta/year from a single program. Single biggest near-term missile-program rare-earth pull-through.
-
Javelin FGM-148 (Lockheed + Raytheon JV): FY24 $1.3B contract (August 29, 2024). Production ramp to 3,960 AUR/year by late 2026. ≈4 tonnes W + ≈4-6 tonnes Ti + ≈200-400 kg NdFeB annually at full rate. 4,000+ for Ukraine replenishment.
-
Stinger FIM-92 (RTX): September 2024 $126.2M, September 2025 $579M follow-on. Dual Detector Assembly (DDA) seeker redesign required — the HgCdTe/InSb industry largely moved offshore. Production ramp ≈1,000-1,100/year.
-
SM-3 / SM-6 (RTX): SM-3 Block IIA kill vehicle LEAP uses tungsten alloy kinetic warhead (10-ton-truck-at-600-mph kinetic equivalent, throwbody ≈20-30 kg with high W density fraction). SM-6 uses C-103 Nb nozzles, Re throat liners. Production: ≈125 SM-6/year, ≈45 SM-3/year.
-
Sentinel LGM-35A ICBM (Northrop Grumman): Replaces Minuteman III in 450 silos. Program cost $141B through FY29 per GAO April 2025 (81% over initial estimate). 634 missiles planned. Public disclosure "advanced composite materials" for booster segments. Mk 21A RV kinetic penetrator 5-15 kg W per RV. C-103 niobium alloy (89% Nb, 10% Hf, 1% Ti) per stage nozzle throat/exit cone, ≈10-50 kg C-103 per stage.
-
Hypersonic glide vehicles (LRHW "Dark Eagle," AGM-183A ARRW revived FY26, HACM): Tungsten carbide + Ta4HfC5 leading edges (operate to 3,000°C). C-103 niobium alloy control surface edges, upper-stage nozzles. Rhenium-molybdenum rocket nozzle throats ≈3-5 kg per throat. UHTC (ZrB2, HfB2) zirconium and hafnium dependent. ≈20-50 kg specialty refractory metals per vehicle. Annual burn small in absolute terms but concentrated on thin supply: global Re ≈50 tonnes/year, Hf ≈80 tonnes/year. LRHW + ARRW + HACM + Dark Eagle all compete for the same narrow supply.
-
155mm artillery ramp: 2022 ≈14,000/month → 2025 40,000/month → target 100,000/month by mid-2026 (per Inside Defense January 2026; delayed from late-2025). Per round primer (MIL-L-16355E composition): lead styphnate 25-55%, barium nitrate 24-25%, antimony sulfide 0-10%, tetracene + calcium silicide + glass powder. Primer mass ≈15-30 g/round → ≈1-3 g Sb sulfide/round. Annual Sb demand at 1.2M rounds/year from 155mm alone ≈ 1.7 tonnes Sb. Small arms at 1-2B rounds/year × 0.05-0.15 g Sb/round = 100-300 tonnes Sb/year. 120mm tank rounds add. Total US defense antimony pull-through from primers is in the low hundreds of tonnes annually. Makes the $245M Sb AMP line make sense.
-
Soldier-portable batteries: Ultralife (ULBI) BA-5390 (Li/MnO2) delivers 50% more energy than incumbent BA-5590 (Li/SO2) at half size/weight; used in ≈60 military applications. Ultralife FY25 10-K (March 23, 2026): "Backlog orders for Battery & Energy Products were approximately $102,000" (thousands; so $102M) as of December 31, 2025; total backlog $110.2M = 57.7% of 2025 revenues, +22.1% over Q3 2025. Key programs: Army Leader Radio Program, Family of Special Operations Vehicles (FOSOV). Customer concentration: L3Harris 27% of 2025 revenues, up from 23% in 2024. DLA contracts: September 2025 $5.2M BA-5390 (through early 2027), December 2021 $9.9M IDIQ.
Germanium damage already at the P&L line. Leonardo DRS 10-K FY2025 (filed February 27, 2026) explicitly cites approximately 2% revenue cost overruns in infrared optics (Advanced Sensing & Computing segment), directly attributable to China's December 2023 germanium export controls. Safety stock deployed; mitigation expected in 2026 through recycling and redesign for alternative materials. This is the first primary-source confirmation that Chinese Ge restrictions are constraining US defense IR programs in real time. SBIRS / Next-Gen OPIR missile warning uses HgCdTe focal plane arrays (RTX payload delivered August 2024 after 13-month delay per GAO) that are 3× more sensitive than SBIRS for hypersonic tracking — larger arrays → more Hg/Cd/Te consumption → germanium optics for primary telescopes.
Stockpile dollars versus demand dollars — the reality check:
Stockpile line $ headline Tonnage Coverage vs pure Coverage at
equivalent defense consumption AMP acquisition
(≈300-500 t Sb/yr; cadence
US commercial Co demand)
Cobalt $500M ≈15,000 t ≈1.5 years N/A — $500M is
at $33/kg (incl. commercial) DPA Title III,
not NDS purchase
Antimony $245M ≈8,000 t ≈15-25 years ≈9 years
at $30/kg (pure defense flow) (AMP 700 MT/yr ×
≈$40K/MT =
$28M/yr spend)
Tantalum $100M Per DLA Multiple years ≈$5M/yr AMP
Scandium $45M N/A N/A Not on AMP;
DPA Title III
The two antimony denominators resolve differently: $245M envelope versus actual US defense antimony flow (300-500 t/yr) = 15-25 years of pure defense demand coverage; $245M at the AMP acquisition rate of 700 MT/yr = a 9-year spend schedule. Both are true; the first frames the envelope relative to consumption, the second relative to procurement cadence.
Stockpile is a crisis bridge, not a structural fix. The AMP acquisition ceilings (700 MT Sb/yr, 64.5k lbs Ta/yr, 587k lbs W/yr FY26) are small relative to annual defense flow. The binding demand function is platform production cadence, not stockpile appropriation. When PAC-3 MSE tripling 500 → 2,000/year hits full rate, annual W+NdFeB+Ta pull-through from one program alone approaches the stockpile numbers. When 155mm hits 100k/month, annual Sb primer demand is 100-300 tonnes/year from small arms + artillery combined. Track platform production releases, not mineral spot headlines.
Platform schedule realism. The targets cited above — PAC-3 MSE 2,000/year by 2030, 155mm 100k/month by mid-2026, Javelin 3,960/year by late 2026, Columbia first-of-class delivery by 2028, Sentinel IOC, F-47 first deliveries mid-2027 — are publicly stated Pentagon and prime contractor guidance. The historical pattern is that complex weapons-system ramps slip. 155mm production has already slipped once (from late-2025 to mid-2026). Sentinel program cost is already 81% over initial estimate per GAO April 2025. Columbia first-of-class is behind schedule. The demand-side conclusions about pull-through magnitudes are directionally right. The timing — and the year-by-year mineral consumption curve — will slip relative to the stated targets. Subscribers pricing specific platform-cadence dates should apply ≈12-24 months of slippage risk across most of these programs.
Section 232 Critical Minerals Proclamation (Proclamation 10858, "Adjusting Imports of Processed Critical Minerals and Their Derivative Products into the United States," signed January 14, 2026; published Federal Register Vol. 91, No. 12, January 20, 2026): Does NOT impose tariffs. It is a negotiation trigger. The verbatim language: "Mining a mineral domestically does not safeguard the national security of the United States if the United States remains dependent on a foreign country for the processing of that mineral. The Secretary and the Trade Representative shall jointly pursue negotiation of agreements to address the threatened impairment of the national security... The Secretary and the Trade Representative should consider price floors for trade in critical minerals and other trade-restricting measures. The Secretary and the Trade Representative shall provide one of these updates within 180 days of the date of this proclamation." The Proclamation reserves fallback authority: "depending on the status or outcome of those negotiations, the President may take other measures to adjust the imports of PCMDPs to address the threat to the national security." Binding operational deadline: July 13, 2026. Subscribers who sold imported-mineral-derivative stocks on January 14 on "tariff news" were wrong — the Proclamation is a negotiation trigger with an optional tariff fallback, not a tariff itself.
Outcome distribution at July 13, 2026 deadline is uncertain but the plausible branches are narrow: (1) trade agreements negotiated with one or more processing partners that include price floors (the stated preference in the Proclamation itself) — helps allied processors (Japan, Korea, Australia) and US-listed midstream pure-plays indirectly; (2) selective tariffs on Chinese-origin processed material only — helps US + allied producers, risks CATL / cathode / anode import cost pass-through; (3) deadline slips and no concrete action is taken — reverts to the pre-Proclamation baseline. No primary-source document specifies which branch is targeted; the Proclamation language explicitly allows price floors, tariffs, "other trade-restricting measures," or negotiated agreements. Watch July 13, 2026 deliverables, plus interim disclosures from USTR + Commerce Secretary in May-June 2026.
1.4 Chip & Display Basket — Ga, Ge, In, Hf, Sb, Te, As, heavy REE, Si, Ag, Sn, Bi
Eighteen minerals, approximately $50B annual production value. This is where Chinese export controls have bitten hardest and fastest. Verified positions per USGS MCS2025 and the Benchmark Mineral Intelligence / IntelMarketResearch compiled data:
Mineral Mining Refining US import
dominance dominance reliance
Gallium China 99% China 99% 100%
Germanium Zinc China dominant >50%
byproduct
Indium China ≈70% China dominant 100%
(Zn byproduct)
Hafnium Zr byproduct Germany ≈50% N/A
China ≈20%
Antimony China ≈60% China ≈65% 85%
Tellurium China 76% China ≈75% <25%
(Cu byproduct) (net exporter
2024)
Arsenic Peru ≈47% Peru ≈45% 100%
Heavy REE China 69% China 91% 80%
Silicon China 84% China 70% <50%
Tin China ≈23% China 53% 73%
Indonesia, Myanmar
Bismuth China ≈81% China ≈81% 89%
Silver Mexico ≈25% China ≈70% 64%
This basket has the thinnest US-listed pure-play exposure. Most of these minerals flow as byproducts of primary mining (zinc, copper, lead, zircon). Direct exposure is through primary metal producers or downstream constraint owners:
- Downstream constraint owners who capture part of the scarcity rent: First Solar (FSLR) for tellurium (CdTe thin-film PV monopoly consumes ≈10-15% of FSLR COGS); Coherent (COHR) for germanium via laser optics; Qorvo (QRVO) and MACOM (MTSI) for GaN/GaAs in RF devices.
- Specialty refiners: 5N Plus (VNPLF, Canadian OTC) — closest thing to a public pure-play on Ge/Te/Bi/In refining, not US-primary-listed.
- Indirect via zinc smelting: Teck Resources (TECK) Trail BC cross-border (Ge byproduct of Zn + Pb); Nyrstar (private since 2022 Trafigura acquisition).
- Korea Zinc (KRX: 010130) Onsan smelter, Ulsan: world's largest refined zinc producer + world's largest non-Chinese indium refiner at ≈150 tpa @99.999% purity (≈11% of ex-China global supply). Antimony: ≈10 tons/day (≈3,650 tpa) defense-critical output. Germanium: 10 tpa announced via 140B won investment. Gallium: pilot operations expected H1 2028, full capacity 15.5 tpa with >16 tpa incremental indium recovery as byproduct. Produces 11 of USGS critical minerals. Plus bismuth, tin, lead, copper, gold, silver.
Korea Zinc Tennessee megaproject — Crucible Metals. $7.4-7.8B Korean-US integrated smelter at Clarksville, Tennessee. Design capacity ~half of Onsan's ("mini-integrated smelter"). 13 non-ferrous metals including 11 critical minerals plus sulfuric acid for semiconductor supply. CHIPS Act incentives awarded by US Department of Commerce to Crucible Metals subsidiary December 2025. Target trial operations 2029. Equity + US private capital + CHIPS. This is the first attempt to onshore the Korea Zinc byproduct capability to the US.
Named Chinese producers in the Chip & Display basket:
Gallium (99% China)
Zhuhai Fangyuan Primary + high-purity, bauxite
Beijing Jiya Semiconductor High-purity Ga, GaN precursors
Aluminum Corp of China Primary (via Chalco subsidiaries)
East Hope Group Primary (bauxite Jinzhong, Henan, Shanxi)
Hangzhou Jinjiang Group Primary (bauxite)
Germanium (≈60-70% China)
Yunnan Lincang Xinyuan Ge SHE: 002428 — ≈280 tpa products (2022),
≈20% of China's output
Yunnan Chihong Zinc & Ge SSE: 600497 — 60 tpa Ge H1 2024,
historical ≈200 tpa
Tungsten (83% China)
Xiamen Tungsten SSE: 600549 — integrated
concentrate + APT + oxide + powder + carbide
Jiangxi Tungsten Industry Unlisted parent, largest group
China Minmetals Tungsten SOE
Antimony (≈48-50% China)
Hunan Gold SHE: 002155 — 20,000 tpa ingot +
40,000 tpa products, ≈50% of China's
2024 national output; single
counterparty for US-bound antimony
Hsikwangshan Twinkling Star Xikuangshan mine (world's largest Sb
deposit), 10,000 tpa Sb smelting
Anhua Zhazixi Mining >6,000 tpa refined Sb
Graphite anode (90.6% China)
BTR New Energy Materials Unlisted (CNPC Battery spinoff) —
575,000 tpa capacity, 22-23% global
share, Indonesia 80,000 tpa Phase I 2024
Shanshan Corp SSE: 600884 — 21% of 2024 artificial
graphite anode; Anning Yunnan 300,000
tpa nameplate; Finland €1.28B plant
Shanghai Putailai SSE: 603659 — Sweden $1.4B plant
announced
Shinzoom 213,000 t shipments 2024 (#3 China)
Shangtai Technology 208,000 t shipments 2024
Kaijin New Energy 158,000 t 2024; 590,000 tpa target
end-2025
The US-investable exposure in this basket is structurally thin. Private market and M&A are the likely vehicles for restoration.
1.5 Power Grid Basket — Cu, Al, electrical steel, SiC/GaN
Minerals: copper (transmission, distribution, transformer windings, datacenter interconnect); aluminum (transmission conductor, ACSR); electrical steel (Fe + 3% Si grain-oriented for transformer cores, non-grain-oriented for motor laminations); silicon carbide and gallium nitride wide-bandgap semiconductors for grid inverters and EV drivetrain.
AI-datacenter-driven capex has moved utility transformer order books from ≈1-year to 3-5-year delivery times (GE Vernova, Hitachi Energy, Siemens Energy earnings commentary 2024-2025). Datacenter load is a material new grid demand factor on top of EV charging and broader electrification.
The mineral-to-equipment value capture in this basket is lowest. Copper and aluminum are fungible commodities with deep futures markets; equity exposure to mineral price runs through FCX (Freeport) or SCCO (Southern Copper), both mega-cap sector-β names without idiosyncratic alpha. Value capture is richer downstream at transformer, switchgear, and power-electronics level — where idiosyncratic alpha lives (GEV, POWL, SXI, ACA in the "Feed Chain" thesis covered elsewhere).
Wide-bandgap semiconductors — the inflection zone 2026-2028:
SiC substrates consumed 40%+ China share in 2025 and rising; TrendForce May 2025 data:
Producer Substrate share 2024
Wolfspeed (WOLF) 33.7%
TanKeBlue (China) 17.3%
SICC (China, SSE) 17.1%
Coherent (COHR) 13.9%
Others (ST, Rohm, 17.9%
SK Siltron,
Synlight, smaller
Chinese)
Chinese 6-inch SiC wafers at $400-500 vs Western $700-900 historically. The Chinese price competition drove Wolfspeed's June 30, 2025 prepackaged Chapter 11 filing in S.D. Texas. Wolfspeed emerged September 29, 2025 after 91 days — eliminated ≈$4.6B of $6.7B debt (≈70% reduction). Mohawk Valley NY (200mm SiC device fab, world's first at 200mm) stayed operational; equity substantially diluted but not zeroed. Post-restructuring Wolfspeed has a clean capital structure and is the flagship US WBG vehicle, but faces Chinese-substrate pricing pressure plus onsemi/Infineon/ST closing the 200mm device gap.
NVIDIA 800VDC architecture — October 2025 OCP Global Summit. NVIDIA formalized reference architecture for AI datacenter 800VDC power conversion, eliminating the traditional 54V intermediate bus. Direct 800V → 12V or 6V at point of load. Reduces copper ≈70% at rack level, collapses conversion stages, enables MW-scale single-rack compute. Architecture layers:
- Grid → 800VDC bus: SiC 1200-1700V solid-state transformers and rectifiers. Infineon, STM, onsemi, Wolfspeed.
- 800VDC → 12V/6V: High-voltage GaN or SiC. Navitas 100V GaN FETs + 650V GaN + high-voltage SiC supplying NVIDIA 800VDC architecture launching 2027 (per Navitas press release 2025). POWI 1250V + 1700V PowiGaN. ST GaN LLC.
- 12V/6V → core voltages: Low-voltage Si or GaN. TI, Monolithic Power (MPWR), Vicor (VICR).
TI unveiled complete 800VDC power architecture with NVIDIA on March 16, 2026 — 800V to 6V DC/DC bus converter with integrated GaN power stages, 97.6% peak efficiency, >2000 W/in³ density. STM announced 800VDC → 12V/6V board families March 17, 2026 at 1 MHz, 98% efficiency, 2,600 W/in³ density.
Power Integrations (POWI): AC/DC integration + PowiGaN. FY2025 revenue $443.5M (+6% YoY). PowiGaN revenue +40% in 2025. InnoSwitch cumulative shipments >1 billion units. 1700V Switcher IC design wins in 800V BEVs. Production on a GaN-based InnoSwitch auxiliary power design for a US cloud services provider started Q4 2025.
Navitas (NVTS): Pure-play GaN. Q2 2025 $14.5M (-29% YoY), trough driven by mobile handset softness. Q3 2025 guide $10M ±$0.5M, non-GAAP GM ≈38.5%. Re-rated sharply on NVIDIA partnership through 2027 launch. Navitas estimates $2.6B GaN+SiC AI datacenter market by 2030.
Infineon (FRA: IFX / OTC: IFNNY): CoolSiC brand. Kulim Malaysia Module 3 — Phase 1 opened August 2024 (€2B, 900 jobs); Phase 2 €5B targeting "world's largest and most efficient 200mm SiC fab" by 2030. FY2025 SiC revenue guidance €1B+. Hyundai 800V E-GMP customer. Acquired GaN Systems March 2023 for $830M — absorbed the Canadian GaN pure-play.
STMicroelectronics (STM): Catania SiC Valley, Italy — 200mm fully integrated SiC fab, production 4Q2025, full build-out to 2033, up to 15,000 wafers/week. ~€5B total including €2B Italian state support under EU Chips Act. Sanan JV Chongqing: 200mm SiC for Chinese market, production 4Q2025, $3.2B total.
onsemi (ON): SiC built via Cree acquisition (2021). $4B+ committed SiC revenue over 3 years via LTSAs. Bucheon South Korea 200mm SiC line, 1M+ 200mm wafers/yr at full capacity.
Gallium chokepoint impact on GaN power: minimal. Entire global GaN power device output consumes less than 10 tonnes Ga/yr. The August 2023 Chinese export controls initially drove procurement friction, but by mid-2024 Western fabricators had pivoted to Japanese, German, and Canadian Ga suppliers. Ga recovery from residual alumina processing scaled 15-20% in 2024-25.
Tesla SiC content evolution: Model 3 Gen 1 inverter used ≈48 SiC MOSFETs at ≈$1,300/vehicle. March 2023 Tesla announcement: next-generation platform reduces SiC content 75% via chip-level integration and topology. This is efficiency of deployment, not abandonment. Other 800V platforms (Hyundai E-GMP, Porsche Taycan, Lucid Air, Audi e-tron GT, BMW Neue Klasse, GM Ultium-later) use 20-60 SiC devices / $400-1,000 SiC content per vehicle.
800VDC datacenter is the most concentrated, highest-conviction WBG demand inflection 2026-2028. Selection now locks architectural incumbency for 2027-2030 AI datacenter rollout. Churn cycle for datacenter power conversion is 3-5 years; EV platforms amortize 8-10. In dollar terms, near-term datacenter outruns EV traction.
Near-term cyclical caveat. The 2025 EV slowdown compressed SiC revenue across the coverage universe. STMicroelectronics (STM) guided 2025 SiC revenue down versus 2024 with recovery to 2024 levels not expected until 2027 per CEO Jean-Marc Chery commentary. onsemi (ON) ran at reduced 150mm SiC capacity utilization through H1 2025 while converting Bucheon to 200mm. Wolfspeed filed prepackaged Chapter 11 June 30, 2025 under exactly these pressures. Navitas (NVTS) Q2 2025 revenue -29% YoY. The 800VDC thesis is real and starting to drive design-win announcements (March 2026 TI + STM NVIDIA disclosures), but the revenue inflection for most names is a 2026-2027 story, not an H2 2025 story. Pair the WBG thesis with explicit near-term revenue-timing discipline.
1.6 Agriculture Basket — honest gap
Potash, phosphate, sulfur, nitrogen (ammonia). Mega-cap integrated producers — Nutrien (NTR), Mosaic (MOS), CF Industries (CF), ICL Israel, Uralkali (Russia, sanctions-impacted), K+S (Germany). Any idiosyncratic alpha at retail-investor position sizing is consumed by analyst coverage and scale. Food-security supply-chain analysis has its own specialist universe. A thorough treatment is outside our competence. We flag the omission rather than fabricate coverage.
1.7 Industrial Basket — catalysts, refractories, specialty alloys
The Industrial basket is residual: catalyst metals (Pt, Pd, V, Ni for chemical processing), refractories (Cr, Mg, Al oxides), specialty alloys (Co superalloys, Ni superalloys, W carbide tooling), and industrial-use specialty materials (Be, Mo). Most overlaps with Magnet and Defense baskets at the mineral level — cobalt appears in Battery, Magnet, Defense, and Industrial. Antimony appears in Chip & Display AND Defense. A single mineral constraint (Chinese export control on Ga, for instance) propagates across multiple demand systems.
Materion (MTRN) — the Industrial Basket specialty case. MTRN operates the world's largest bertrandite (beryllium ore) mine and refinery at Delta, UT — the only Western vertically integrated beryllium producer. Beryllium is non-substitutable in nuclear weapons primaries, nuclear reactor reflectors, satellite mirrors, X-ray windows, spacecraft structures. MTRN is not a processing pure-play (integrated miner + refiner + downstream across multiple segments) and its trailing factor regression shows approximately 38.5% idiosyncratic variance — partly sector beta. But the beryllium monopoly is real.
Mitsubishi Materials (TYO: 5711) — the allied tungsten transformation. December 2024: acquired H.C. Starck (Germany) GmbH. Post-acquisition tungsten supply capacity approximately 15,000 tpa, nearly 6× prior, making Mitsubishi Materials one of the largest non-Chinese tungsten refiners globally. H.C. Starck brings ammonium paratungstate (APT) and downstream tungsten carbide / tungsten powder refining capability at Goslar, Germany. This is the most strategically consequential Japanese critical-minerals M&A of the past two years. Same parent operates the Naoshima Smelter & Refinery (Kagawa) for integrated copper + PGM recycling, plus Materials Eco-Refining Co. Ltd. subsidiary for PGM intermediates.
Part 2 — Demand Factor Decomposition
Each basket's demand is driven by a distinct set of orthogonal factors. The same mineral can appear in multiple baskets (cobalt, antimony, heavy REE), so mineral-level supply constraint propagates unevenly across end-uses.
2.1 Battery Basket
Driver Approx. share Primary macro 2026-2030
of demand factor CAGR range
EV traction 65-75% Global auto × EV 15-25%
batteries share × kWh/vehicle
BESS (stationary, 10-15% Utility capex × 30-40%
2-8h duration) firming require- (fastest-growing)
ment × LCOE vs gas
Consumer electronics 10-15% Mature replacement 3-5%
+ medical + indl. cycle
Heavy-duty vehicles + 2-5% DoD + commercial 20-30%
marine + air mobility aviation electrify (small base)
Orthogonality check: EV and BESS are partially orthogonal — BESS correlates with utility capex and peaker economics, not with auto sales. A weak EV quarter can coexist with strong BESS deployment. Lithium and cobalt are disproportionately EV-exposed; lithium and manganese are disproportionately BESS-exposed (LFP chemistry dominates BESS, NMC dominates EV).
Substitution: Na-ion batteries have moved from lab to early commercial (CATL, BYD, Natron Energy, Faradion). Credible forecasts put Na-ion at 5-10% of stationary storage by 2030, aggressive upside 20% in 4-hour-duration BESS. Na-ion displaces lithium, not cobalt/nickel/manganese. LFP share of battery chemistry moved from ≈30% (2022) to 40%+ (2025 industry estimates), reducing cobalt demand per kWh at the margin. Solid-state remains commercial 2027-2030+ and does not affect near-term supply-demand math.
2.2 Magnet Basket
Source: DOE "Rare Earth Permanent Magnets: Supply Chain Deep Dive Assessment" (February 2022, Table 3, p. 17). The canonical primary-source table. Global shares for 2020, updated where possible:
Application Global 2020 US 2020 Implied 2024 tonnage
(260 kt base)
Consumer electronics 29% 45% ≈75 kt
(HDDs, phones, speakers)
Industrial motors 30% 30% ≈78 kt
Offshore wind turbines 14% 0% ≈36 kt
(direct-drive PMSG)
Non-drivetrain auto motors 8% 9% ≈21 kt
EV traction motors 6% (2020) 11% ≈16 kt in 2020,
→ 10-12% ≈28-32 kt 2024
(2024)
Bonded magnets (Magnequench 7% 4% ≈18 kt
NOPMF dominant)
Other sintered (e-bikes, 5% 1% ≈13 kt
specialty)
IEA STEPS forecast (Global Critical Minerals Outlook 2025, p. 170):
Year Magnet REE demand NdFeB implied EV motor share Wind share
(kt REO) (kt) of magnet REE
2024 91 ≈303 9% ≈11%
2030 123 ≈410 16% 14%
2040 150 ≈500 20% 14%
2050 180 ≈600 22% ≈12%
Adamas Intelligence undersupply projections (May 2023):
NdFeB alloy: 60 kt undersupply by 2030 → 246 kt by 2040
(≈ current annual production shortfall)
NdPr oxide: 19 kt undersupply by 2030 → 90 kt by 2040
Dy oxide: 1,800 tonnes undersupply by 2040
(= current total world production)
Tb oxide: 450 tonnes undersupply by 2040
(= current total world production)
Heavy REE loading per kg NdFeB (DOE 2022 Table 4, p. 18):
Grade suffix Max op T Dy content Applications
(none) 80°C <0.5% Toys, advertising
M 100°C 1.4% HDDs, CD/DVD, MRI, sensors
H 120°C 2.8% Gauges, magnetic separation
SH 150°C 4.2% Wind direct-drive, industrial, e-bikes
UH 180°C 6.5% Commercial/industrial generators
EH / AH 200/220°C 8.5-11% Hybrid/EV traction drives
Per-BEV traction motor: 1.5-2.5 kg NdFeB (EH/AH grade). At 9% Dy loading = 135-225 g Dy per BEV. Large offshore wind turbine: 650 kg NdFeB × 4% Dy = 26 kg Dy per turbine.
Grain-boundary diffusion (GBD) reduces Dy below Table 4 values — manufacturers achieve 2-4% Dy in EH grades with GBD. This is the real continuing threat to Dy demand, not Fe16N2 substitution.
Orthogonality check: EV/PHEV demand and wind demand are partially correlated through energy-transition capex cycles but decouple at quarterly horizons. Robotics demand (humanoid + industrial) is a cleaner orthogonal driver — labor-cost macro and automation penetration, not auto production. Defense demand is orthogonal to all of the above and accelerates under DFARS enforcement.
2.3 Defense Basket
Demand is budget-driven, not price-driven. Program cadence IS the demand function.
Mineral DoD demand vector FY25-FY26 budget signal
Tungsten Kinetic penetrators, shaped-charge AMP 587k lbs FY26 (-87% vs FY25),
liners, armor, radiation shield, DLA RFI issued, PAC-3 MSE
cutting tools tripling pull-through
Antimony Munition primers (1-3 g Sb sulfide/ AMP 700 MT flat; $245M
155mm round), small arms primers headline = 15-25 yr pure
(1-2B rounds/yr), night vision defense demand coverage
(Sb-doped HgCdTe)
Titanium Aerospace engines, airframes, AMP 13,608 MT FY26; $160M
sponge naval structures, IPX $160M non-dilutive to IPX verified
Tantalum Missile guidance, radar, avionics AMP 64,500 lbs flat; $100M
capacitors headline
Niobium HSLA steel, superalloys AMP 300k lbs; NB Elk Creek
75% first-10-year off-taken
per USGS MCS2025
Cobalt Superalloys + battery + magnet AMP ZERO acquisitions;
$500M = DPA Title III, not NDS
Scandium Al-Sc alloys, SOFC Not on AMP; $45M = DPA
Title III programmatic
HALEU U Advanced reactors (Oklo, TerraPower, LEU $3.8B firm backlog
X-energy, Kairos, USNC, Radiant) + 2040, $2.3B contingent LEU
naval reactors (Virginia, Columbia, sales
Ford use HEU not HALEU)
PGM Autocatalyst + electronics + Diversified; not pure
specialty chemicals + H2 defense driver
Beryllium Nuclear weapons primaries, reactor DoD + DOE integrated
reflectors, satellite mirrors, demand; MTRN Delta UT
X-ray windows monopoly
Defense basket demand is orthogonal to all other baskets. A recession that compresses EV demand (Battery, Magnet) does not compress munitions replenishment. Enforcement risk (DFARS waivers) is the main structural variable; if DoD waives specialty-metals content restrictions for specific programs to avoid cost overruns, the demand floor erodes.
2.4 Chip & Display Basket
Driver Approx. share Primary macro 2026-2030
of basket factor CAGR range
Semiconductor fab capex 40-50% WFE cycle × AI 10-15%
(logic + memory) accelerator ×
advanced node
AI datacenter buildout 10-15% Hyperscaler capex 20-30%
(compute-adjacent) ($300B+ 2025E for
top 4)
Fiber optics + 5G/6G 10-15% Telecom capex + 5-10%
infrastructure government
broadband
Display (TV, smartphone, 15-20% Consumer electronics 3-5%
automotive) replacement + auto
PV (thin-film CdTe 5-10% Utility solar × thin-10-15%
via FSLR) film share
Military IR / night 5-10% DoD budget × sensor 10-15%
vision / RF modernization
AI datacenter buildout is the strongest 2025-2028 vector. Hyperscaler capex from Microsoft, Google, Amazon, Meta totaled $300B+ expected 2025, with AI accelerator purchases driving materials demand for Ga (GaN RF + power), Ge (substrate + fiber optics), Te (CdTe solar for power sourcing), and heavy REE (magnets in cooling fans).
2.5 Cross-basket correlation — categorical
Not a statistical fit. Shared-driver analysis based on the underlying demand factors in 2.1-2.4. The numeric correlation table was pulled because the decimal estimates implied precision the shared-driver analysis does not have.
Basket pair Shared drivers Correlation tier
Battery × Magnet EV cycle (both) HIGH
Battery × Grid Utility capex (BESS) MEDIUM
Battery × Chip Weak — consumer electronics only LOW
Battery × Defense Weak — soldier batteries only LOW
Magnet × Grid Wind direct-drive, AI datacenter MEDIUM
cooling-fan magnets
Magnet × Chip Weak — HDDs declining LOW
Magnet × Defense DFARS 225.7018 + platform magnets MEDIUM
Grid × Chip AI datacenter power conversion + HIGH
fab expansion
Grid × Defense Minimal LOW
Chip × Defense Minimal LOW
Defense basket demand is the most orthogonal to EV-cycle-exposed Battery + Magnet baskets. Chip × Grid share the AI-datacenter driver. For orthogonal diversification within a critical-minerals sleeve, Defense is the cleanest incremental exposure.
Part 3 — US-Listed Exposure Map
Ten names that clear a four-filter screen: binding contract evidence; idiosyncratic variance ≥75% on factor regression (or explicit flag where below); US-investable at OTCQX minimum; market cap preferred <$2B (flexibility to $3B for near-binding candidates). Per-name briefs below include verbatim quotes from the relevant 10-K, 20-F, 6-K, and 8-K filings with line citations.
On sizing. This piece does not recommend position sizes — it maps names that survive the filter. For the reader who uses proportional sizing (Paleologo): weight each name by |α| / Σ|α|, not by conviction. Names with idio variance above 75% carry the clean factor-regression signature; names flagged below 75% (PESI 57%, ULBI 54%, LYSDY 64%, IPX 75%) are sector-beta-mixed and should size down accordingly or be paired with a sector hedge. Pre-revenue names (NMG pre-FID, USAR pre-first-shipment, IPX pre-commercial-scale) are binary on commissioning; size as options, not positions. Entry discipline matters more than name selection across this tape — most of the ten names were RSI 70-90 mid-April 2026, and the 2025 EV slowdown pressured battery-adjacent names into tight financing windows that the equity market will continue to price.
The ten survivors
Ticker Basket Stage Primary demand Idio % Mkt cap
LEU Defense + Midstream Advanced reactors 86.6% ≈$3.0-3.5B
(Grid adjacent) (HALEU + naval + DFARS
enrichment)
NMG Battery Midstream Panasonic binding 98% ≈$350M
(anode AAM) offtake 13,000 tpa
Oct 31, 2025
PESI Defense Downstream DFLAW Hanford 57% ≈$150M
(nuclear waste) ramp (NO medical
processing) isotope pivot)
ULBI Defense Downstream Army Leader Radio 54% ≈$130M
(defense + FOSOV; backlog
battery) 57.7% of revenue;
L3Harris 27%
ALM Defense Midstream Tungsten Sangdong 94% ≈$1.9B
(W mining + (Korea) + Gentung
processing) (MT); DoD
stockpile pull-
through
IPX Defense + Midstream Ti sponge + $160M 75% ≈$2.0-2.3B
Aerospace (Ti processing) non-dilutive US
Government + Titan
Project TN
USAR Magnet + Downstream Stillwater NdFeB 122% ≈$1.2B
Defense (magnet mfr) Phase 1a Q1 2026 +
DFARS 225.7018
Jan 1, 2027
LYSDY Magnet Upstream + Japan + Blue Line 64% ≈$9.0B
(ADR) Midstream JV + NdPr + Dy/Tb
circuit 2026
NB Defense + Upstream Niobium 75% first- 107% ≈$400M
Grid (multi- 10-year off-taken (β=-0.04)
commodity) + scandium + β=-0.04
EU Defense Upstream US utility term 76% ≈$400M
(U ISR mining) contracts, 655k
lbs 2025 @
$51.09/lb cost
vs $74-77 spot
Per-name briefs
LEU (Centrus Energy). Sole US NRC-licensed commercial HALEU enrichment facility at Piketon, OH. 10-K filed February 11, 2026 (line 624-625): "The Company's backlog across both segments was $3.8 billion and $3.7 billion as of December 31, 2025 and 2024, respectively, and extends to 2040." Of that $3.8B firm backlog, $2.3B is specifically contingent LEU sales per line 581-582: "Centrus plans to leverage its multi-billion-dollar uranium enrichment expansion to meet its growing backlog of $2.3 billion in contingent LEU sales." Three DOE IDIQ contracts executed: HALEU Production (October 2024), HALEU Deconversion (October 2024), LEU Production (December 2024). Fluor EPC contract signed February 9, 2026, with a declining cancellation fee — line 3440-3443: "cancellation fee in the initial amount of $24.0 million on the contract effective date, subsequently decreasing by $2.0 million per month until reaching $0." Real credibility commitment. January 5, 2026 DOE $900M HALEU task order awarded but subject to negotiation of a definitive agreement — "no guarantees about whether or when funding by the DOE for such expansion would be awarded" per line 965-970. CEO disclosed ≈3-year moat to first new commercial HALEU centrifuges from any US competitor. 22.8% short interest reflects bull/bear split. Entry discipline: pullback to $170-180 (200DMA).
NMG (Nouveau Monde Graphite). Only NYSE-listed graphite anode with binding Panasonic offtake. Pre-FID. $96.5M bought-deal closed April 16, 2026; concurrent $213M PIPE conditional on shareholder approval by July 31, 2026; $335M senior debt commitment. Phase-2 total ≈$644M. GM is equity, not product offtake. Idio 98%. Size as pre-FID lottery.
PESI (Perma-Fix Environmental Services). Treatment Segment $45.1M = 73.1% of 2025 revenue (+29% YoY) per 10-K filed March 24, 2026 (line 312-313). PFNWR permit renewed December 2025; liquid mixed waste capacity approximately tripled to 1.2M gallons/yr + 175,000 tons/yr macroencapsulation (line 247-253). DFLAW hot commissioning began October 2025; operational phase anticipated 2026. Only US pure-play nuclear/hazardous waste processor. Scope is Hanford DFLAW + PFAS destruction patent. Idio 57.3% — below the 75% target because Services Segment adds government-contract beta noise. 2 analysts. Under-covered.
ULBI (Ultralife). $102M Battery & Energy Products backlog Dec 2025 (all-time high); total backlog $110.2M = 57.7% of 2025 revenues, +22.1% QoQ. BA-5390 50-100% more energy than BA-5590 in ≈60 military applications. Army Leader Radio Program + FOSOV. L3Harris customer concentration 27% of 2025 revenues (up from 23% in 2024) — single-customer concentration risk is material. Director Bradford Whitmore bought $1.2M single-day March 16, 2026; $2.2M cumulative 2025-2026 insider buying.
ALM (Almonty Industries). Commercial tungsten mining started at Sangdong (Korea) December 2025; formal commissioning March 2026. Gentung MT acquired November 2025 (100% owned, H2 2026 production). Nasdaq uplist July 14, 2025. $268M cash year-end 2025 after $129M December 2025 follow-on. April 13, 2026 6-K: HQ relocating Toronto → Dillon, Montana. Kennametal, Plansee, Global Tungsten & Powders customer relationships disclosed in March 17, 2026 commissioning press release (non-SEC). DoD FY25 tungsten AMP ceiling cut -87% — read as pivoting from ore stockpiling to specialty-alloy reprocessing rather than pure demand drop.
IPX (IperionX). $160M non-dilutive US Government commitment verbatim in 6-K filed March 19, 2026: "The U.S. Government has committed ~US$160 million in non-dilutive funding to accelerate the re-shoring of a domestic titanium supply chain. Beyond being our largest non-equity financier, the U.S. government is poised to become a major customer." Four US-located facilities (Virginia Titanium Campus, Utah R&D, Virginia R&D, Tennessee Titan Critical Minerals Project). US net import reliance >95% since 2021.
USAR (USA Rare Earth). Stillwater OK Phase 1a commissioned Q1 2026 per 10-K filed March 30, 2026 (line 551-557). Phase 1a + 1b at full capacity = 1,200 MTPA sintered NdFeB magnet blocks = ≈2.7× DLA AMP NdFeB ceiling (450 MTPA). CHIPS Act LOI $1.6B ($277M direct grants + $1.3B 15-year loans) is non-binding with six conditions pending per 10-K line 4297-4307: (i) $500M non-federal sources [satisfied via $1.5B PIPE]; (ii) two semiconductor-end-user MOUs; (iii) NdPr oxide + MREC feedstock supply agreements through 2027; (iv) Texas GLO surface purchase option; (v) third-party validation of Colorado Facility nuclear material licensing; (vi) Stillwater power infrastructure plan. Carester SAS transaction April 9, 2026 is a term sheet under Item 7.01 (Regulation FD), NOT Item 1.01 — "subject to customary conditions, including confirmatory due diligence and negotiation and execution of definitive documentation." USAR + InfraVia each to take ≈12.5% minority if closed. FY2025 revenue $1.643M is all from Less Common Metals UK metal-making operations acquired November 2025 (line 1975). Pre-revenue for magnets. Pro-forma cash ≈$1.86B after January 28, 2026 $1.5B PIPE. CEO Barbara Humpton joined October 2025 from Siemens USA; previous CEO Joshua Ballard received $5.5M FY25 compensation (including $526,879 "All Other Compensation," suggesting severance). Round Top is exploration stage, no proven reserves under S-K 1300, PFS expected H1 2026. Noveon Magnetics (private, San Marcos TX) is the incumbent US commercial sintered NdFeB producer with ABB LTA — USAR is the only public US NdFeB vehicle.
LYSDY (Lynas Rare Earths ADR). OTCQX Best Market + Billion+ Index, ≈$4.6M/day ADV. Mt Weld mining + Kuantan Malaysia separation. First separated Dy and Tb production at Kuantan achieved Q2 2025; FY26 H1 production 35 tonnes combined Dy+Tb at "strategic premium to spot." Only producer of separated heavy rare earths outside China. Seadrift TX: CEO Amanda Lacaze on August 2025 FY2025 results call said there is "significant uncertainty as to whether construction of the Seadrift processing facility will proceed, and if so, in what form." Real factor sensitivity to China policy (January 7, 2026 China-Japan magnet export ban headline moved LYSDY +16% in single day).
NB (NioCorp Developments). Elk Creek, NE multi-commodity (niobium + scandium + titanium + REE). 10-Q filed February 6, 2026 (line 1601-1603): "As of December 31, 2025, the Company had cash of $306.4 million and working capital of $297.9 million, compared to cash of $25.6 million and working capital of $24.8 million on June 30, 2025." 12× cash increase in six months from $139.1M October 2025 offering + warrant exercises. $44.6M Mine Portal Project approved December 22, 2025 (Q1 2026 start) per line 1394-1407. EXIM Financing up to $800M per line 1637-1641 — "subject to, among other matters, the satisfactory completion of due diligence, the negotiation and settlement of final terms, and the negotiation of definitive documentation" — pending review since June 6, 2023. December 4, 2025: $8.4M cash acquisition of FEA Materials LLC for Al-Sc master alloy production. 75% of first 10-year ferroniobium production contracted per USGS MCS2025. Beta = -0.04 on factor regression — counter-cyclical diversifier within the basket.
EU (enCore Energy). 10-K filed March 31, 2026 (line 6959-6965): 655,000 lbs U3O8 sold in 2025 at average realized $65.89/lb against weighted-average cost $51.09/lb = 22.5% gross margin. South Texas ISR (Rosita, Alta Mesa) + Upper Spring Creek + Gas Hills WY. Great Plains New Nuclear Consortium MOU with Nebraska PPD, Omaha PPD, Lincoln Electric System, Grand River Dam Authority (OK) — MOU language, not binding supply agreement. Cash $52.4M, working capital $96.1M. No S-K 1300 reserves (exploration stage resources only).
Excluded names (and why)
- WWR: SK On terminated March 31, 2026 per 8-K 2026-04-01 Item 1.02. Stranded.
- UUUU: +313% 1Y. $0 REE revenue 2024 AND 2025. Phase 2 commissioning Q1 2029. $700M convertible dilution cap at ≈$20.34. Wait for ASM/KMP close June 2026 window.
- CENX: Glencore distributed $327M March 4, 2026. Six C-suite + officers sold $14M+ Jan-Mar 2026. Zero insider buying same window. +299% 1Y. Real DOE Coop $500M binding, but trade in late-phase distribution.
- ELBM: LGES Cobalt Supply Agreement explicitly non-binding per 20-F March 30, 2026. Minimum-cash covenant distress.
- SGML: 40-F discloses international sales via sea freight without named non-China offtakes. Hidden-China default.
- LAC: ATM active, 40% share growth 12 months, buying into dilution. Factor bet on lithium recovery, not clean idio alpha. Size separately.
- ATI / CRS / HWM / LMT / RTX / GD / NOC: Idio 25-45% = sector β. Good businesses, not idiosyncratic critical-minerals alpha at mega-cap.
- Pink Limited ASX (ARAFF / ILKAF / ILKAY / ASMMF / PILBF / SYAAF / SRHYY / VULNF / TLOFF): Below OTCQX minimum filter. ILKAY 924 shares/day ADTV is structurally uninvestable.
Part 4 — The Allied Midstream Playbook
The U.S.-centric framing of critical-minerals policy omits a structural fact: the allied midstream operational backbone is Japanese-Korean, not American. Japan built its rare-earth lifeline in 2011 in response to the 2010 Senkaku Islands crisis. Korea built its zinc-byproduct critical metals capability at Korea Zinc's Onsan over 50 years. The Pentagon's 2025 MP Materials package is a direct derivative of that Japanese template — not an innovation, but a catch-up.
4.1 JARE + Lynas — the 2011-2026 template
March 2011: Sojitz Corporation + JOGMEC (Japan Organization for Metals and Energy Security) deployed US$250M combined (loan + equity) to Lynas via the Japan Australia Rare Earths (JARE) special-purpose company. This was Japan's structural response to the 2010 Senkaku Islands crisis where China cut rare-earth exports.
September 2022 + March 2023: JARE committed an additional AUD 200M equity, explicitly for LRE capacity expansion AND to commence Dy/Tb separation using Mt. Weld feedstock. The 2022-2023 terms gave Sojitz and JOGMEC supply of up to 65% of Lynas heavy rare earth output.
March 11, 2026: Revised agreement extends through 2038 (Lynas + JARE joint press; corroborated by Metal Tech News, Argus Media, Benchmark Minerals, Japan Times, Sojitz corporate). Key terms:
- 5,000 tonnes per annum of NdPr from Lynas through 2038.
- Price floor US$110/kg NdPr (JARE pays the shortfall if market VWP is below $110/kg).
- Profit-share above US$150/kg: if market NdPr price exceeds $150/kg, Lynas shares 30% of the additional revenue with JARE, capped at $10 million annually.
- Heavy rare earths: JARE to purchase at least half of all Dy/Tb produced by Lynas (revised down from the 2022-2023 "up to 65%" framing; the 50% floor is the March 2026 contractual minimum).
The MP Materials / DoW structure announced July 2025 replicates the core mechanism in US form: $400M Series A preferred + $150M Samarium Project Loan + 10-year NdPr price protection at US$110/kg + $140M/yr EBITDA guarantee on magnet offtake. The NdPr price-floor number is identical to JARE's — the Pentagon picked the same $110/kg figure Japan had already been paying Lynas for years. The profit-share mechanisms are structurally similar but not identical: JARE shares 30% of revenue above $150/kg, capped at $10M/year; DoD takes 30% of MP's NdPr sales price above $110/kg with no stated annual cap. Both contracts front-run China-origin material risk through a fixed downside + shared upside structure. The JARE agreement (2011 origin, March 2026 revision) is the ≈15-year precedent.
JOGMEC's mandate extends beyond Lynas — the agency also backs lithium, nickel, copper, graphite across multiple jurisdictions. But Lynas is the cleanest disclosed case and the closest analog to US-listed critical-minerals equity.
4.2 Korea Zinc Onsan — the allied byproduct critical metals node
Korea Zinc (KRX: 010130) operates the Onsan smelter, Ulsan, on the Onsan National Industrial Complex. Site: 1.41 million m² — approximately 18× the Seoul World Cup Stadium. The facility produces 11 of the USGS critical minerals:
Mineral Korea Zinc position
Zinc World's largest refined zinc producer
Indium ≈150 tpa @ 99.999% purity; ≈11% of ex-China global
supply; world's largest non-Chinese indium refiner
Antimony ≈10 tons/day (≈3,650 tpa) defense-critical output
Germanium 10 tpa capacity announced (140B won investment)
Gallium Pilot ops expected H1 2028; full 15.5 tpa with
>16 tpa incremental indium recovery as byproduct
Bismuth, tin, lead, copper, gold, silver — additional outputs
Crucible Metals Tennessee megaproject. $7.4-7.8B Korean-US integrated smelter at Clarksville, TN. Design capacity ~half of Onsan. 13 non-ferrous metals including 11 critical + sulfuric acid for semiconductor supply. CHIPS Act incentives awarded to Crucible Metals subsidiary by US Department of Commerce December 2025. Target trial operations 2029. This is the first attempt to onshore Korea Zinc's byproduct capability to the US.
Governance overhang: Young Poong (founding-family ally-turned-adversary) + MBK Partners launched a hostile tender offer September 2024. Korea Zinc Chair Choi Yun-beom + allies hold 33.9% vs Young Poong + MBK bloc 49.6%. Seoul Court rejected Young Poong/MBK injunction December 2025 to block the $1.94B share sale funding the Tennessee project. Proxy battle ongoing. Material overhang on Korea Zinc equity but does not change Onsan's strategic value as allied supply-chain node.
4.3 Sumitomo Metal Mining Philippines — the HPAL original
Before Tsingshan scaled Indonesian HPAL starting 2018, Sumitomo Metal Mining (TYO: 5713) operated the proven non-Chinese HPAL blueprint. Coral Bay Nickel Corp. (Palawan, Philippines, operational since 2005) and Taganito HPAL Nickel Corp. (Caraga region, Philippines, operational since 2013). Both plants produce mixed sulfide (MSP) shipped to Niihama refinery (Ehime) for refining to electrolytic Ni + Co.
Niihama is the only plant in Japan producing electrolytic nickel and electrolytic cobalt via proprietary MCLE (Matte Chlorine Leach Electrowinning) technology. Product mix includes electrolytic nickel, nickel chloride, nickel sulfate, electrolytic cobalt, ferronickel, and scandium oxide.
Taganito HPAL commenced scandium recovery in 2018, chromite recovery from 2021. SMM produces scandium oxide as a disclosed standard product — tonnage undisclosed but historically single-digit tpa, material to non-Chinese supply.
4.4 Mitsubishi Materials + H.C. Starck — the 2024 tungsten consolidation
Mitsubishi Materials (TYO: 5711) acquired H.C. Starck Holding (Germany) GmbH in December 2024. Post-acquisition tungsten supply capacity approximately 15,000 tpa, nearly 6× prior, making Mitsubishi Materials one of the largest non-Chinese tungsten refiners globally. H.C. Starck brings ammonium paratungstate (APT) and downstream tungsten carbide / tungsten powder refining at Goslar, Germany. This is the single most strategically consequential Japanese critical-minerals M&A of the past two years.
4.5 Toho + Osaka Titanium — the aerospace Ti floor
Toho Titanium (TYO: 5727): ≈40% global Ti sponge share historically. Chigasaki (Kanagawa) + Wakamatsu/Kitakyushu (Fukuoka) sponge plants. Saudi Arabia JV (TOHO TITANIUM METAL Co., with Advanced Metal Industries Cluster / Saudi ARAMCO). Current expansion +3,000 tpa by 2026 at Wakamatsu. Toho actively evaluating a new Ti sponge facility — US, Saudi Arabia, or Japan — as part of capital deployment to absorb Boeing's post-VSMPO reallocated demand.
Osaka Titanium Technologies (TYO: 5726): World's second-largest Ti sponge producer after VSMPO. Post-Russia-sanctions, arguably the largest effective non-Russian supplier. Amagasaki facility. ¥30B (~US$191M) commitment to new sponge plant at Amagasaki lifting capacity from 40,000 to 50,000 tpa. Incremental tonnage 2028-2029.
Combined Japanese sponge capacity is approximately 2× VSMPO's capacity as of 2024, subject to the caveat that aerospace-grade delivery is constrained by PMA (Part Manufacturer Approval) qualification timelines at airframer level.
4.6 POSCO Future M + EcoPro BM + L&F + Lotte Energy — the Korean cathode/anode/copper foil stack
POSCO Future M (KRX: 003670). Most vertically integrated Korean battery-material company. Steelmaking parent POSCO Holdings provides coal-tar feedstock that uniquely enables domestic synthetic graphite. Cathode: combined Pohang + Gwangyang capacity 150,000 tpa domestic as of 2025. Target 2026: 248,500 tpa total; full run-rate 395,000 tpa. Customers include Samsung SDI (January 2023 ≈40 trillion won NCA cathode supply agreement), GM via Ultium CAM JV Quebec (30,000 tpa), and a major Japanese battery maker. pCAM: 45,000 tpa Gwangyang precursor plant operational. Anode: Sejong Plant 1 (24,000 tpa natural) + Sejong Plant 2 Phase 1 (20,000 tpa natural) + Pohang synthetic using POSCO coal-tar byproduct — only such facility in Korea. Synthetic capacity to exceed 18,000 tpa end 2025. October 2025: $470M anode materials supply deal with unnamed US carmaker — largest Korean anode deal in industry history. Target full phase-out of Chinese feedstock by 2027. March 2024: signed Syrah Resources (Australia / Balama Mozambique) for African natural graphite.
EcoPro BM (KOSDAQ: 247540). NCA + CSG cathode. HQ Cheongju, principal operations Ochang. Capacity roadmap target: 710,000 tpa by 2027 (Hungary plant 108,000 tpa NCA design). Samsung SDI ≈$34B multi-year deal. TTM revenue $1.77B as of September 2025.
L&F (KOSDAQ: 066970). Cathode specialist, HQ Daegu. High-nickel NCA focus. February 2023: US$2.91B two-year Tesla supply agreement. The contract was subsequently amended down 99% — from 3.83 trillion won ($2.67B) to 9.73 million won. Cause: Cybertruck production challenges and 4680 ramp issues. Major demand signal: 4680 architecture underperformed vs cathode offtake commitments. LG Energy Solution selected L&F over parent LG Chem for Tesla battery cathodes — unusual intra-chaebol routing indicating L&F technical advantage.
Lotte Energy Materials (KOSPI: 020150, formerly Iljin Materials). Lotte Chemical acquired controlling stake 2022 for 2.7 trillion won. Copper foil (elecfoil) for battery anode current collectors. 45,000 tpa 2022 capacity. Samsung SDI selected Lotte Energy as exclusive copper foil supplier for Samsung's first US battery plant. Spain 30,000 tpa plant — Europe's first copper foil factory. Customers: LG Energy Solution, Samsung SDI, BYD.
SK IE Technology (KRX: 361610). Only listed pure-play battery separator producer in Korea. HQ Seoul. Film-based and ceramic-coated lithium-ion battery separators. Capacity planned at 2.73 billion m² by 2024, Korea domestic 530 million m². Customers: SK On, LG Energy Solution, EU cell makers.
4.7 Proterial Dy-free sintered — the real Dy substitution threat
Proterial Ltd (formerly Hitachi Metals, delisted March 2023) disclosed on July 22, 2025 the development of heavy-rare-earth-free sintered NdFeB magnets (NMX-F1SH-HF and NMX-G1NH-HF) with Br = 1.40 T, HcJ ≥ 1,671 kA/m — competitive with Dy/Tb-containing grades in a 180°C operating window. Mass-production plant accepting pre-production sample requests April 2026. This is a far more credible threat to Dy demand than Niron iron nitride because Proterial's chemistry is sintered-class and EV-motor-temperature-qualified. Fe16N2's physics ceiling caps it at <150°C service temperature; Proterial's chemistry operates at the 150-220°C regime where Dy/Tb loading actually matters.
If Proterial's Dy-free sintered qualifies at automotive customers over 2026-2028 and achieves meaningful production scale, the Dy loading per kg NdFeB in the high-volume EV-motor segment falls — cutting into the 8.5-11% Dy content of EH/AH grades (DOE 2022 Table 4). The magnitude depends on how fast Proterial ramps versus the concurrent Chinese GBD reduction and the ongoing Adamas Dy undersupply forecast (1,800 tonnes by 2040). Range of outcomes is wide; this is an analytical estimate pending Proterial's own capacity disclosure, which has not been made. Watch the April 2026 pre-production sample requests and first disclosed automotive qualification wins as near-term catalysts.
4.8 Energy Fuels acquiring ASM Korean Metals Plant
Australian Strategic Materials (ASX: ASM) operates the Korean Metals Plant (KMP) at Ochang, South Korea — the only facility outside China producing REE metals and alloys (NdPr metal, NdFeB strip alloy, DyFe). Energy Fuels (UUUU) + ASM scheme of arrangement announced January 20, 2026, amended March 13, 2026, expected to close June 2026. Consideration: 0.053 UUUU CDIs + A$0.13 cash per ASM share. Implied ASM equity A$447M; US EV ≈$299M. Post-close, the non-China REE alloy capacity moves to US ownership. This is material strategic consolidation.
4.9 US-investable allied universe — the reality check
The direct US-listed universe in allied midstream is thin. Where it exists, it is via:
- ADR shadows — Shin-Etsu (OTC: SHECY / SHECF), Sumitomo Metal Mining (OTC: STMNF / SMMYY). Thinly traded, OTC Pink, not suitable for concentrated sizing.
- Upstream miners and concentrators — MP, Perpetua, NioCorp, IperionX, Almonty, Energy Fuels, Centrus.
- Neo Performance Materials (TSX: NEO / OTCQX: NOPMF) — only dual-listed midstream magnet/REE processor with reasonable US liquidity. Narva, Estonia sintered magnet plant opened September 19-22, 2025, Phase 1A 2,000 tpa design, targeting up to 15% of EU demand.
Count approximately 10 US-listed names in Part 3 against the 25-30+ Japanese + Korean + European + Indonesian operators catalogued in Sections 4.1-4.8 (Shin-Etsu, Proterial, TDK, Daido, Mitsubishi Materials / H.C. Starck, Sumitomo Metal Mining + Coral Bay + Taganito, Toho, Osaka Titanium, POSCO Future M, EcoPro BM, L&F, Lotte Energy Materials, SK IE Technology, Korea Zinc, Vacuumschmelze, Lynas Kuantan, Tsingshan IMIP + IWIP, Huayou, GEM, Harita / Lygend, Eramet, Vale Pomalaa, JL MAG, Ningbo Yunsheng, Zhong Ke San Huan, Earth-Panda, Shenghe, Ganfeng, Tianqi, Yahua, Xiamen Tungsten, Hunan Gold). The ratio is approximately 1:3. US subscribers building allied critical-minerals supply-chain exposure through US tickers alone will miss most of the operational footprint. This is a real structural constraint, not a modeling error.
Part 5 — Private Pipeline, M&A Patterns, Multiples
5.1 The MP Materials DoW package — structural template
The July 2025 MP Materials / Department of War public-private partnership is now a reproducible federal-equity financing template (8-K filed July 10, 2025, the largest single-transaction 8-K in the critical-minerals universe). The structure verbatim:
- $400M Series A Cumulative Perpetual Convertible Preferred Stock. Stated value $1,000/share. 7.0% PIK dividend compounding quarterly. Conversion price $30.03 (closing price + NYSE "Minimum Price" requirement). Convertible after 45 days. Beneficial ownership cap 19.9%. Company can force convert if common stock ≥ 200% of conversion price for 20 of 30 trading days.
- $350M additional Series A commitment at Company election (or equivalent under JPMorgan/Goldman Sachs committed debt).
- $150M heavy rare-earth separation loan from DoD within 30 days of closing.
- 10-year NdPr Price Protection Agreement (PPA). Quarterly payment per kg NdPr equal to shortfall between $110/kg and benchmark VWP. After 10X Facility reaches full production, if benchmark VWP exceeds $110/kg, DoD takes 30% of sales price above $110/kg (profit-share mechanism). PPA applies to concentrate + oxide + metal.
- 10-year Magnet Offtake Agreement. Special purpose vehicle Project Company sells entire 10X Facility magnet production to DoD. DoD may syndicate up to 100% to commercial buyers. Minimum $140M/yr EBITDA at full production, inflation-adjusted. DoD pays 25% of Threshold EBITDA as quarterly magnet facilitation payments pre-COD, capped at $30M/yr.
- Warrant for 15% common stock at closing (combined with Series A), capped 19.9% beneficial.
- $1B JPMorgan + Goldman Sachs committed secured debt; $650M allocated to 10X construction. Total 10X budget $1.25B (balance $600M from MP's cash).
- Covenant: MP ceased all China product sales July 2025.
The structure is equity + preferred + loan + offtake + price floor + profit share + warrant. It transforms an otherwise-uneconomic project into a floored cashflow stream. Pre-deal MP EV ≈$2.5B; post-deal EV ≈$10B by April 2026 — MP stock 4-5x on the restructuring alone.
Trilogy Metals precedent: $35.6M DoW stake @ $2.17/unit October 2025 for NOVA copper-zinc project — smaller MP-template applied to copper-zinc. Establishes the mechanism works for minerals beyond rare earth.
The structure is reproducible for NioCorp (NB), IperionX (IPX), Almonty (ALM), Perpetua (PPTA), Energy Fuels (UUUU) — any US critical-minerals pure-play with cash-flow pre-requirement of federal price support. Watch DPA Title III award announcements and DoD Industrial Base Fund allocations 2026-2028 for replication signals.
5.2 Aerospace specialty alloy consolidation — the 10-12× EBITDA template
Two completed deals established a tight comp set for aerospace specialty alloys acquired by European stainless majors:
- Haynes International (HAYN) / Acerinox — closed November 21, 2024. $970M EV, $61.00/share cash (22% premium to 6-month VWAP). 12.3× FY2023 EBITDA ($78.9M implied). Haynes revenue FY2023 $590M with $290.4M aerospace (+26.3% YoY) and backlog $460.4M (+23.2% YoY).
- Universal Stainless (USAP) / Aperam — closed January 23, 2025. $539M EV, $45.00/share cash (19% premium to 3-month VWAP). 10.6× TTM / 7.7× forward EBITDA. USAP FY2023 revenue $285.9M; 2024E ≈$330M; 2024E adjusted EBITDA ≈$70M.
Implied comp math for remaining names:
- ATI Inc (ATI) at $3.2B 2024 revenue, ≈$450M EBITDA → 10× = $4.5B EV, 12× = $5.4B EV. Currently trades ≈$7-8B EV — already above specialty-alloy comp.
- Carpenter Technology (CRS) similar scale; already trading through specialty-alloy multiples.
Neither ATI nor CRS is a distressed-multiple takeout candidate at current pricing. The HAYN/USAP acquirers (Acerinox, Aperam) have consolidated what they could at sensible multiples; further European stainless acquirers (Outokumpu, Kobe Steel, Daido) would need valuation reset to match the HAYN/USAP comp.
5.3 Tier 1 miner consolidation at sub-cycle prices
- Rio Tinto / Arcadium Lithium — closed March 5, 2025. $6.7B all-cash, $5.85/share (≈90% premium to undisturbed). ≈4-5× 2024 revenue; EBITDA not meaningful at sub-cycle lithium. Integrated upstream (brine + spodumene) + midstream (LiOH Bessemer City NC, Naraha Japan). Rio financed with up to $9B new US bond issuance.
- BHP + Lundin Mining / Filo Corp (Vicuña Corp 50/50 JV) — closed January 2025. C$4.1B, C$33.00 per share cash. Argentina-Chile Filo del Sol + Josemaria copper combine, >2Bt inferred resource.
- Newmont / Newcrest — closed November 6, 2023. US$16.8B. Stock exchange 0.400 NEM per NCM. World's leading gold miner with Cadia-Telfer copper byproduct exposure.
Tier 1 majors pay premiums at sub-cycle when Tier 2s cannot finance. Albemarle (ALB), SQM, Pilbara (PLS), Sigma Lithium (SGML) are next-tier candidates if/when lithium price floor finds real support.
5.4 Battery recycling and anode distress cycle
- Jervois Global Chapter 11 — filed January 2025, 37-day prepack. Millstreet + Penderfund take-private with $145M cash injection + $100M IDA bonds debt-to-equity. DoD DPA Title III grant survived. Equity wiped. Idaho Cobalt Ops (only US cobalt mine) operational.
- Li-Cycle Holdings (LICY) — CCAA + Chapter 15 May 2025. Glencore stalking horse $40M credit bid + assumption, closed August 5, 2025. Rochester Hub + Germany/AZ/AL/NY/Ontario spokes + IP. Creditors receive proceeds; equity impaired.
- Westwater Resources (WWR) — SK On termination March 31, 2026. Next distress candidate. Pred-709tiu 55% distressed sale by end-2026.
Glencore is the serial acquirer of battery-recycling distress assets at pennies-on-the-dollar-of-capex multiples.
5.5 Private pipeline IPO candidates 2026-2027
Most-likely to file S-1 within 18-24 months, grouped by thesis:
Magnets (ex-NdFeB supply chain)
- Noveon Magnetics (San Marcos, TX). $215M Series C January 2026. US-operational sintered NdFeB at commercial scale. Binding ABB Long-Term Agreement. GM + Lynas + Solvay strategic partners. The incumbent US commercial producer. Likely S-1 mid-2026 window.
- Niron Magnetics (Sartell, MN). $214M cumulative across 8 rounds, 18 investors. Sartell operational target early 2027. Differentiated rare-earth-free Fe16N2. IPO window post-commissioning.
Battery materials (recycling + cathode + anode) 3. Redwood Materials. $350M Series E October 2025 at ≈$6B post-money. NVIDIA NVentures on cap table. Sparks NV + Charleston SC. Integrated battery recycling + cathode + anode copper foil. 4. Ascend Elements. $542M equity + $480M DOE grants. Hopkinsville KY pCAM manufacturing + Covington GA recycling. Multiple named offtake contracts. 5. Cirba Solutions. $50M Series D Feb 2023 + $200M DOE grant Sept 2024. Wixom MI HQ + Columbia SC hub. Battery recycling specialist. 6. Anovion Technologies. $117M DOE grant Oct 2022 + $400M EXIM letter of interest Nov 2024. Bainbridge GA synthetic graphite anode (40k tpa design at full build). $800M total facility. 7. Mitra Chem. $60M Series B August 2023 (GM-led). Iron-phosphate LMFP cathode specialist, Silicon Valley. China LFP substitution thesis. 8. Our Next Energy (ONE). $300M Series B Feb 2023 at $1.2B post-money. LFP cells, Van Buren MI. Staff cuts during 2024 EV slowdown; track capacity actualization before IPO. 9. Lyten. $200M Series B September 2023 + $200M 2025 + Cuberg asset acquisition Nov 2024. $650M EXIM letter of interest. Lithium-sulfur battery + advanced materials. San Jose HQ + Nevada planned.
Exploration technology / critical-minerals mining 10. KoBold Metals. $537M Series C January 2025 at $2.96B post-money. Bezos + Gates + Andreessen + T. Rowe + Durable capital. AI-driven critical-mineral exploration. Mingomba (Zambia) copper discovery = revenue proof-of-concept. Most mature cap table after Redwood.
Two gating observations: (1) none of these has filed an S-1 as of April 18, 2026; timing is speculative. (2) EV slowdown 2024-2025 pressured several battery-material private names to delay. A sustained lithium price recovery is the likeliest 2026-2027 IPO-window opener for Ascend / Cirba / ONE / Lyten.
5.6 Pending allied consolidation
Energy Fuels + ASM KMP closes June 2026, US EV ≈$299M. Transfers non-China REE alloy capacity to US ownership.
MP Materials downstream — DoW capital unlocked 10X Facility construction; opportunistic bolt-ons 2026-2028. Northlake TX selected as 10X site.
Part 6 — Honest Gaps and Policy Appendix
6.1 What we cannot verify
- Chinese on-soil tonnage flows. The refining-share chart (Benchmark Mineral Intelligence / IntelMarketResearch) aggregates data from industry-paid sources. Specific tonnage by Chinese smelter by year is behind paywall. We cite aggregate shares; we do not claim smelter-level precision.
- Private processor capacity. Heraeus, Umicore, Johnson Matthey autocatalyst PGM recycling capacity by facility; Noveon NdFeB tonnage shipped under ABB LTA; Niron Sartell commissioning date and nameplate; Nemaska LiOH status. Private-market facts that surface only partially through customer filings.
- Adamas Intelligence end-use percentage breakdowns. The 2024 figures are in the paid "Rare Earth Magnet Market Outlook to 2040" report. We cite public fragments (8.7% CAGR 2024-2040; +13.3% consumption growth 2023; 259 kt 2024 consumption) and use the DOE 2022 Table 3 canonical percentages as the public baseline.
- Tesla Robstown unit economics. Reagent cost per kg LiOH; yield; waste-water profile; actual current throughput vs 50 GWh target; specific alkaline reagent system (inferred NaOH-dominant pressure leach, not primary-source). The Inside Climate News March 2026 wastewater story suggests the environmental balance is still being resolved.
- DoD stockpile-to-contract conversion cadence. FY25 AMP ceilings are published; contract-specific DLA procurement notices appear individually in the Federal Register. Multi-year buildout in progress.
- DFARS 225.7018 enforcement track record. The regulation is finalized and effective January 1, 2027. How aggressively DoD will enforce it (waivers for specific programs to avoid cost overruns, COTS interpretation) is forward-looking.
- Niron Fe16N2 max operating temperature at production scale. Not disclosed. Inferred <150°C from peer-reviewed decomposition kinetics.
- MOFCOM license issuance counts. Aggregate approved/denied for Ga/Ge/Sb exports post-August 2023 are not disclosed publicly. Pillsbury, White & Case, DLA Piper track the policy changes; numeric counts require S&P Global Commodity Insights / Argus / Fastmarkets subscriptions or Chinese Customs Statistical Yearbook (12-18-month lag).
- Myanmar HREE flow attribution to specific Yunnan smelters. Open-source reporting redacts the specific Chinese entity names. Global Witness, Jamestown Foundation, Stimson Center document flow; named Yunnan receiving points are not in public primary sources.
6.2 Baskets we do not cover
Agriculture basket (potash, phosphate, sulfur, nitrogen) — outside our competence, dominated by mega-cap integrated producers. Residual Industrial catalysts and refractories (Cr, Mg oxide, Mo, V apart from grid-storage context) — exposure at commodity level is mega-cap sector beta.
6.3 China export control timeline (verified from public record)
Date Event
2023 Jul 3 PRC announces gallium + germanium license requirement
2023 Aug 1 Ga/Ge licensing effective
2023 Oct 20 Graphite licensing announced
2023 Dec Graphite licensing effective
2024 Aug 15 Antimony export restrictions announced
2024 Sep 15 Antimony licensing effective
2024 Dec 3 MOFCOM Announcement No. 46: full US export ban on
Ga, Ge, Sb, superhard materials
2025 Feb 4 Tungsten + molybdenum + tellurium + indium + bismuth
added to dual-use list
2025 Apr 4 MOFCOM + GAC Announcement No. 18: seven medium and
heavy REE (Sm, Gd, Tb, Dy, Lu, Sc, Y) + alloys +
oxides + permanent magnet materials added to export
control
2025 Oct 9 Announcements Nos. 55-58, 61, 62: 12 additional
categories (superhard, heavy RE equipment, Li
battery, artificial graphite negative-electrode
materials)
2025 Nov 7 Announcement No. 72 suspends six Oct 9 announcements
until Nov 10, 2026
2025 Nov 9 Article 2 of No. 46 suspended until Nov 27, 2026
(Ga/Ge/Sb/superhard to US restored under general
licensing)
Sources: ORF America, Global Trade Alert, Pillsbury client advisories, Clark Hill advisory, CSET Georgetown translations, China MOFCOM/GAC primary announcements.
6.4 US regulatory + funding stack
- DFARS 225.7018 (final rule, effective January 1, 2027): NdFeB + SmCo magnet supply chain prohibition covering Iran, North Korea, Russia, China. Pre-2027: limited to "melting and producing."
- DFARS 252.225-7052 (related final rule): restriction on acquisition of certain magnets, tantalum, tungsten.
- DPA Title III (Defense Production Act) programmatic awards: MP ($400M Series A pref), LAC (DOE ATVM $2.26B), IPX ($160M), NB (DOE + DoD), Nova Minerals ($43.4M), UAMY ($27M), Fireweed ($15.8M), Guardian Metal ($6.2M), South Star Battery ($3.2M).
- IRA Section 45X advanced-manufacturing credit: CMC recognized $53.8M H1 FY26; ENR $9.7M Q1 FY26; ULBI $1.4M FY25.
- IRA Section 48C advanced-energy project credit: $6B second-round including Niron Sartell.
- EXIM Bank direct loans: NB Elk Creek up to $800M pending since 2023; PPTA Stibnite $2.7B congressional notification.
- OBBBA Section 20004: $2B NDS + $5B DIB Fund + $500M loan program = $7.5B total.
- Section 232 Critical Minerals Proclamation (January 14, 2026): 180-day negotiation deadline July 13, 2026. Optional tariff fallback.
The stack is real and material. The operative question is conversion speed from appropriation to binding contract. FY25 execution through Q1 2026 has been slower than appropriation implies. Watch FY26 OBBBA continuation + DLA procurement cadence as the near-term signal.
Closing
The heavy-REE (Dy/Tb) supply thesis is the structurally cleanest long exposure in this basket, because the two most-cited magnet-side substitution cases do not touch it. Niron iron nitride's physics ceiling at <150°C service temperature limits it to ≈37-49% of NdFeB end-use — consumer electronics, bonded powder, non-drivetrain auto, other sintered — and does not reach EV traction or offshore wind direct-drive. Proterial's July 2025 Dy-free sintered NdFeB (rated to 180°C) is a more credible Dy-demand compressor than Niron, but only qualifies at automotive customers with commissioning cadence and disclosure that have not yet happened. Chinese grain-boundary diffusion is actively reducing Dy loading per kg NdFeB 30-50% in M/H/SH grades. All three pressures combine to compress Dy demand growth — the 150-220°C regime where Dy matters remains underserved by non-China supply (Lynas Kuantan Dy/Tb circuit 2025-2026 is the only operating ex-China separation at scale). That's the structurally insulated corner of the thesis.
The structural thesis beyond that corner is established: 60 USGS critical minerals, China ≈85% processing, seven end-use baskets, federal policy stack of DFARS + DPA Title III + 45X + 48C + EXIM + OBBBA + Section 232 stacking toward July 13, 2026 deadline + January 1, 2027 DFARS 225.7018 effective date. The execution layer is name-by-name primary-source work on a tape that is late-cycle extended.
Ten US-listed names survive the filter as of April 18, 2026. Nine minerals (Ge, Ga, Te, Bi, Co non-Congo, Ni sulfide, Hf, Ta, HPMSM Mn) have no US-listed pure-play under $2B — structural gaps that likely produce M&A targets and private IPOs over 24 months.
The best-structured binding contract in the US critical-minerals universe is MP Materials' Department of War package — $400M Series A preferred + $150M loan + $110/kg NdPr price floor + 30% profit share above $110/kg + $140M/yr EBITDA guarantee. Anyone actively following the theme for 12+ months is already anchored to MP. That is a starting point, not an undiscovered idea.
The allied midstream is a Japanese-Korean industrial network. Japan built its rare-earth lifeline in 2011 with JARE + Lynas. Korea built Onsan over 50 years. Indonesian HPAL is Chinese-controlled. Japanese titanium sponge capacity is ≈2× VSMPO. Mitsubishi Materials + H.C. Starck now operates ≈15,000 tpa tungsten refining, one of the largest non-China tungsten capacities globally. The US-investable pure-play universe is approximately one-third of the actual allied operational footprint catalogued in Part 4.
A framework synthesis establishes the thesis. A primary-source filter establishes the position. Corrections, additional primary-source citations, and better tonnage data welcome.
// comments (0)