Pre-registration: four tests for a non-Chinese rare earth supply chain (Lynas, Iluka, MP Materials), with dated forward predictions

# Pre-registration: four tests for a non-Chinese rare earth supply chain (Lynas, Iluka, MP Materials), with dated forward predictions Version 1.0.0 · 28 October 2026 · Author: Jorge Anibal Restrepo-Morales ## Purpose This deposit fixes, before the data exist, the forward predictions that will be used to evaluate a planned Perspectives article on whether a rare earth supply chain outside China has passed four tests of viability. It also states the scoring rules and the baseline figures read from company reports on 28 October 2026. ## Question, cases and window Which tests of viability has each case passed between July 2025 and June 2026, and which has it not? Cases: Lynas Rare Earths (Australia, with separation in Malaysia), Iluka Resources (Eneabba refinery) and MP Materials (United States, comparator). The subject is the three companies, not a country. Quantitative window: 1 July 2025 to 30 June 2026 (Lynas fiscal year 2026). Events after the window are reported only as dated events, up to 31 October 2026. ## Four tests and scoring rules (denominators declared by the companies) - P1, light oxides: NdPr produced in the window divided by the nominal capacity target declared by the company (Lynas: 12,000 t per year, FY25 results, 28 Aug 2025). At least 90 %: pass; 50 to 90 %: pass with conditions; below 50 %: fail. - P2, commercial viability without state support: share of new capacity or revenue that depends on a public price floor or public financing. Above 50 %: no evidence of viability without support; 10 % or less: pass; in between: pass with conditions. - P3, heavy rare earths: Dy+Tb produced divided by the nominal target of the new heavy-separation plant (300 t per year, Lynas announcement of 29 Oct 2025, plant not yet built). At least 50 %: pass; 10 to 50 %: pass with conditions; below 10 %: fail at scale. - P4, downstream: magnet product sold by allied plants linked to the case. Sold at commercial scale: pass; manufactured without a declaration of commercial production: pass with conditions; only agreements or plants not yet built: fail. The scoring of the window is retrospective: the rules were written after the figures were seen. Only the forward predictions below are ex ante. ## Baseline figures (primary sources listed in the supplementary file) Lynas FY26 NdPr production 7,260 t (quarters between 1,404 and 2,003 t); Dy+Tb 62 t; Iluka Eneabba start-up mid-2027 with a state loan of A$1.65 billion; MP Materials Q2 2026 NdPr 840 t and US$17.6 million of price-floor income against US$108.5 million of revenue. Figures not confirmed in a primary source are marked as such in the supplementary file. ## Forward predictions (fixed on 28 October 2026) They will be evaluated with the Lynas quarterly report for the September 2026 quarter, the Lynas report for the December 2026 quarter, and the MP Materials Q3 2026 release. They will be reported whether they hold or not. - F1. Lynas NdPr production in the September 2026 quarter lies between 1,404 and 2,003 t. It fails if it falls outside that range. - F2. Lynas Dy+Tb production in the September 2026 quarter is 30 t or less. It fails if it exceeds 30 t. - F3. MP Materials reports price-floor income above US$10 million in Q3 2026. It fails if it is US$10 million or less. - F4. Neither Noveon nor Lynas announces a definitive (binding) magnet or supply agreement with the US Department of War before 31 December 2026. It fails if either does. ## Outcome of earlier predictions (plan v1, same day) Spot NdPr below US$110/kg in most quarters: confirmed with reservations (2 of 4 quarters with data). Dy+Tb nominal capacity above 620 t per year: failed (300 t per year). Heavy rare earths of Lynas processed in China: no source found. Ore-grade comparison: not evaluable and omitted. ## Files VERIFICACION_PRIMARIA_2026-10-28.md (sources and verification status of each figure, in Spanish), SHA256SUMS.txt. Artificial intelligence statement: drafting and source reading were assisted by an AI system under the specification and verification of the author.