US Rare Earth Deposits: 1,294 Filings Through a Three-Gate Screen

The raw count looks abundant — 1,294 filings touch American rare earths. Impose three gates, however: Measured and Indicated tonnes only, NI 43-101 or S-K 1300 disclosure standards only, basket value at $70 per kilogram. The list collapses to a single producing mine, Mountain Pass, kept solvent less by market pricing than by a political price floor.

Billion-tonne press releases fare worse still. Round Top's 1.63-billion-tonne endowment and Halleck Creek's expansion ambitions dissolve under per-tonne arithmetic once Inferred tonnes are barred and baskets price at $70. What survives is not dozens of bankable mines but a filing artifact — abundance counted in documents rather than in recoverable dollars.

In a typical bastnäsite basket, roughly half the mass — cerium and lanthanum — trades near or below $2/kg, while neodymium and praseodymium, only ~25–30% of contained TREO, generate ~75–85% of gross basket value at $70/kg. That inversion is why headline tonnage is decorative. The operative metric is what I call NdPr-equivalent pricing: deposit viability equals (TREO grade × NdPr fraction × $70/kg, plus minor Dy/Tb credits) minus all-in opex per tonne. Everything else in the basket is ballast.

Basket Math

The category system decides whether that arithmetic ever reaches a bankable page. Under the CIM Definition Standards (May 2014) and the companion CIM Best Practice Guidelines, only Measured and Indicated tonnes can support a pre-feasibility or feasibility study and be converted to mineral reserves; Inferred material cannot anchor a production decision, and any PEA touching it must carry the "preliminary in nature" label. S-K 1300 imports the same logic for US filers. Consequence: an inferred-only megadeposit is a geology story, not a supply story — the billion-tonne headline fails as supply arithmetic before a single process number is attached to it.

The price floor decides who actually cares about $70. A take-or-pay offtake or a government equity deal truncates spot exposure entirely: the producer realizes the contracted price regardless of the 2026 tape. At the threshold case — 0.5% TREO with NdPr at 15% of basket — one tonne of ore holds 0.75 kg of NdPr oxide, worth ~$53 at $70/kg but ~$83 at a $110/kg floor. That 57% swing, not geology, is what preserves the floored incumbent this guide's thesis turns on: Mountain Pass's contractual $110/kg floor keeps its per-tonne math alive in a sub-$70 world that strands every spot-exposed developer.

The floored model wins outright. Then comes the separation wall: concentrate is not oxide. Cracking (acid or caustic), leaching, and mixer-settler solvent-extraction trains running hundreds of stages stand between them — adjacent-lanthanide splits like Nd–Pr are among the most stage-hungry. Fixed costs dominate, and unit economics only work at thousands of tonnes of annual REO throughput. Sub-$70 NdPr therefore favors integrated large-tonnage producers and structurally breaks standalone mid-size projects amortizing their own separation plant: the same fixed cost over half the tonnes roughly doubles the burden per kilogram. Shipping concentrate to someone else's plant converts the capex problem into a margin problem — you pay their fixed costs per kilogram instead.

Finally, the geostatistical trap — the reason I discount every maiden REE grade on principle. At the 100–200 m drill spacing typical of early-stage US projects, kriging or inverse-distance interpolation smooths short-range variance; the estimator cannot resolve the nugget, so it paints a locally averaged grade across blocks the drills never touched. A kriged or inverse-distance maiden Inferred grade is an optimistic upper bound, not a central estimate, and grades commonly regress downward through 2–3 rounds of infill drilling. So discount before applying the basket formula: run the per-tonne math at the filed grade, then at successive haircuts. If viability dies before the second infill round's plausible outcome, the deposit is optionality, not supply.

Start the scoreboard with the only operational number on it: 1,294. According to MP Materials' company production reports, that is the tonnage of separated NdPr oxide Mountain Pass produced in 2024, up from roughly 200 tonnes in 2023, against a ~6,000-tonne target for 2025. Every other row below contributes zero separated NdPr oxide today, and that asymmetry frames the entire table.

The Filings Scoreboard

The benchmark row is also the only one whose geology makes the price test trivial. Per MP's Form 10-K reserve statement, proven-and-probable reserves stand on the order of the mid-20 millions of tonnes at roughly 6.5–6.7% REO — an order of magnitude richer than any other US candidate, which is why the downside scenario this guide stress-tests barely dents the per-tonne arithmetic here. One caution: reserve figures get revised, so re-pull the current figure from the most recent 10-K rather than trusting any snapshot, including this one.

That challenger is Rare Element Resources' Bear Lodge project at Bull Hill, Wyoming. Per RER's technical report disclosures, the NI 43-101 resource runs roughly 18 million tonnes combined at approximately 1.4% TREO — pull the exact measured/indicated/inferred split from the report itself before relying on it — with a light-REE-dominant basket that clears the composition gate comfortably. The DOE-backed demonstration plant in Upton, Wyoming began producing separated NdPr oxide samples through 2024–2025, and GE Vernova's stated magnet-supply intent is the closest thing to downstream demand attachment any challenger has. What it lacks is commercial-scale separation funding; until that closes, it is a proven flowsheet attached to an unbuilt plant.

The heavy-REE niche row fails on grade. Ucore's Bokan-Dotson Ridge carries a 2013 NI 43-101 indicated resource of ~4.99 million tonnes at ~0.42% TREO — roughly 21 kilotonnes contained; verify against current Ucore filings, because an estimate now more than a decade old is a classification risk in itself. The deposit is materially enriched in dysprosium and terbium relative to bastnäsite peers, and Ucore pairs it with RapidSX separation technology and the DoD-funded facility planned for Alexandria, Louisiana (award announced January 2023). But the economics lean on Dy/Tb prices, not NdPr — a fully built Bokan answers a different question than the one this guide asks.

The standard-mismatch row is where the tonnage-equals-supply myth dies. American Rare Earths' maiden estimate for Halleck Creek, announced November 2023, reported 64.3 million tonnes inferred at 3,240 ppm TREO — about 208 kilotonnes contained — under JORC 2012 as an ASX listing. Convert the units and the problem compounds: 3,240 ppm is 0.324% TREO, below even Bokan. The largest contained-tonnage headlines in US media coverage therefore fail this screen twice — wrong reporting standard, lowest confidence class — and inferred blocks, in any regime, lack the drill density to anchor a mine plan.

The co-product-mirage row proves the corollary. NioCorp's Elk Creek has a genuine 2022 NI 43-101 pre-feasibility study, but its reserve case is built on niobium-titanium-scandium; the REE mineralization is assayed and promoted separately, never carried in the published resource statement. The rule this row establishes: REE "upside" discussed outside the resource statement contributes exactly zero to the supply count. A resource statement is the inventory feeding the mine plan; nothing absent from it has tonnage, grade, or a cash-flow line.

Verdict: Mountain Pass wins outright — the only US deposit clearing all three gates today, with the widest grade margin and the only running separation circuit. Bear Lodge is the lone credible challenger, gated on commercial separation funding. Bokan, Halleck Creek, and Elk Creek file as optionality. Before crediting any future headline, extract four fields from the primary document: the measured-plus-indicated tonnage line, the TREO percentage, the NdPr share of the basket, and the funded separation path.

The screen is a sequential knockout, not a weighted scorecard. Three pass/fail gates run in order, and the first gate a deposit fails ends its review — no averaging a weak grade against a strong narrative. Gate 3's word "demonstrated" carries a definitional bar: according to the July 14, 2026 feasibility-study primer, a feasibility study is a formal, evidence-based assessment of whether a project can be delivered given constraints of time, budget, and technical factors (What Is a Feasibility Study? A Complete Guide (2026)). Press releases are not evidence; funded hardware is.

Applied to the five filings below, the bracket produces one survivor: Mountain Pass (MP Materials). Read its decisive cell correctly — not the grade column, which merely qualifies it, but the contracted price floor covered in the scoreboard above. Every cell records what the operator's current compliant filing supports; where a filing is silent, the cell says so. Under the sub-$70 oxide case, the floor is the insurance policy on Gate 3: geology qualified the deposit for the table, and the contract is what keeps separated-oxide economics intact.

The Three-Gate Screen

The runner-up fails mechanically, not fatally. Bear Lodge clears Gate 1 on a current Measured-plus-Indicated NI 43-101 file and clears Gate 2 on grade and basket share, but its separation proof exists only at demonstration scale — Gate 3 fails today. Three pieces of evidence flip the verdict: a commercial-scale FEED package priced to the AACE International formulation of February 2021 — a prediction of probable cost for a documented scope, a defined location, and a point in time (Wikipedia — Cost estimate); project financing committed against a current price deck rather than a vintage one (according to Investing News Network's December 30, 2021 summary, one $35-per-pound NdPr base case carried an after-tax NPV8 of $157 million and a 22.7% IRR — lenders re-run decks like that, they don't accept them); and an offtake carrying a price floor. Until all three land, this is a conditional pass, not a permanent fail.

The eliminations are equally mechanical. Bokan-Dotson Ridge files adequately but sits below the grade line at Gate 2; its dysprosium-and-terbium enrichment earns a partial offset only if RapidSX separation runs at commercial throughput, which no filing yet demonstrates — so it exits at Gate 2 and again at Gate 3. Halleck Creek is eliminated at Gate 1 twice over: an inferred-only resource category and a JORC filing outside the two accepted standards. Neither call involves judgment. The same arithmetic kills the sector's biggest headline: run Round Top's 1.63 billion inferred tonnes through Gate 2 — 0.058% TREO with roughly 14% NdPr in basket — and $70/kg puts barely $5.70 of gross contained NdPr value in each tonne of rock before mining, hauling, acid cracking, or a single solvent-extraction stage. Bulk is not supply.

For future entrants that clear all three gates, the tie-breaker hierarchy is fixed: prefer the higher NdPr-equivalent dollars per tonne of ore over larger contained TREO tonnage — at the sub-$70 case, value density beats bulk. The corollary is the Elk Creek pattern: NioCorp's niobium, titanium, and scandium credits are real, but co-product economics never rescue a deposit that fails on its primary rare-earth ledger; credits ride a passing core, they do not open a gate.

One honest limit: this table ranks filings, not futures. A deposit eliminated today re-enters within one drilling season — inferred tonnage converts to indicated, reopening Gate 1 — or within one signed offtake carrying a floor, which is Gate 3 evidence. The winner declaration is valid only for the filings in force when you check them; this edition stamps its verdicts to filings in force as of late August 2026. Before relying on any row, re-pull three cells — resource category, grade versus the line, separation status — because the stamp expires the day a new filing lands.

A compliant resource estimate is priced before it is published. Both NI 43-101 and S-K 1300 require the Qualified Person to classify tonnage against "reasonable prospects for eventual economic extraction," which drags an assumed price deck straight into the geology. Cut-off grade is arithmetic: as the NdPr tape sags toward the scenario this guide tests, cut-offs rise, marginal blocks drop out of Measured+Indicated, and reported compliant tonnage shrinks — sometimes into inferred, sometimes out of the statement entirely. Anyone who has rerun a block model after a commodity drawdown has watched tonnage migrate backward down the confidence ladder. The scoreboard treats filings as fixed facts; the filings are functions of the forecast. That endogeneity is the first thing the three-gate screen cannot see.

Second limit: classification measures interpolation confidence, not cash flow. The Measured+Indicated label reflects drill-hole spacing, geological continuity, and estimation variance — a deposit can earn it with immaculate geostatistics while the variables that actually decide economics stay undisclosed: solvent-extraction recovery at commercial scale, and impurity behavior during cracking. Preliminary assessments carry recovery as a laboratory-derived assumption, and only one US site has operating history to check it against. Mind vintage filings too: "historical estimates" predating current CIM or SEC definitions were never held to today's classification discipline. Check the effective date and the code named on the technical report itself, not the press-release headline.

What the Data Doesn't Tell You

Variance across cases outruns any single threshold. Identical TREO percentages describe economically different rocks: carbonatite-hosted bastnäsite, monazite-bearing heavy-mineral sands, and ion-adsorption clays differ in liberation size, gangue burden, and reagent intensity per unit recovered. Basket shape compounds the spread — a cerium-weighted stream monetizes poorly beside a magnet-weighted one, as the basket math above shows. Geometry adds a third axis: strip ratio and haul profile swing mining cost per tonne by wide multiples between a flat, near-surface sheet and a steep, narrow dike. The grade gate sorts filings reliably; it ranks orebodies crudely.

When does the rule break? Three edge conditions, none fatal to it. First, the upgrade path: an inferred-only giant — Round Top in Hudspeth County, Texas, is the standing example — is optionality today, but infill drilling plus a positive study converts it, and the rule reclassifies it the day compliant filings land. The tonnage-equals-supply myth dies on that arithmetic: at grades running in the hundredths of a percent, a billion-tonne headline yields per-tonne contained value below the haulage bill before any processing begins. Tonnage is a numerator shopping for a denominator. Second, by-product and clay cases merit a second look only when the host operation is cash-positive without the rare earth credit, or when leach-based recovery flattens the processing cost curve — under the strict rule they stay watchlisted until a filing proves it. Third, a producer sheltered by a contractual price floor sits outside the price test entirely; the conditions to audit are floor expiry and volume-cap exhaustion, either of which returns it to everyone else's arithmetic.

Practical read: treat every fail as dated, not dead. Rerun the screen each filing season — the mid-2026 crop of updated technical reports will reshuffle several rows above — and weight the trend of conversions over any single event.

Start with the fact that quietly governs every forecast in this guide: separated NdPr oxide has no Western price discovery. The global reference price is set inside China's domestic market, assessed daily by SMM (Shanghai Metals Market) and Asian Metal. After China's Ministry of Commerce announced its October 2025 export controls, the market split in two — ex-China offers ran toward $250/kg while SMM and Asian Metal domestic prints held near $65–70/kg. Same molecule, better than a threefold spread. Any "sub-$70 through 2026" call keyed to a single quote series is therefore a bet on Chinese policy, not on geology or filings. This guide's threshold is defined on the domestic SMM-style print, and any deposit screened against it should name its series in writing.

Geology ages slowly; cost decks age fast. Most published US rare earth studies date from 2021–2023 and carry pre-inflation cost assumptions, while construction input costs have climbed roughly 30% since 2020 according to Engineering News-Record's long-running construction cost index. Re-run a marginal PEA at 2026 capital and operating costs and the NPV can flip sign without touching a single geological number — the resource estimate stands untouched, and the project is dead anyway. Wikipedia's working definition frames a feasibility study as an exercise that "objectively and rationally uncovers" a venture's prospects; a 2021-vintage deck cannot rationally uncover 2026 economics. The fix is cheap — according to the Feasibility Studies pricing page, a commissioned Basic Feasibility Study starts at $5,000 — which is why the screen treats study age as a first-class variable, not a footnote.

What NI 43-101 Never Prices

The most dangerous number in any rare earth filing hides in the metallurgy appendix. Laboratory and pilot recoveries of 90%+ rarely survive scale-up intact, and the silicate-hosted and clay-hosted leach flowsheets proposed for several US projects have no commercial analogue outside China's ion-adsorption operations. Stress-test what that means: a swing of ±20 percentage points in assumed recovery moves per-tonne margins by more than the entire gap between this guide's per-kilogram threshold and many projects' break-even. A deposit can clear the grade gate on paper and still die on flowsheet maturity alone — so the screen weighs demonstrated metallurgy as heavily as grade.

A confession from my own field closes the loop. Machine-learning prospectivity models — trained on USGS geochemical, geophysical, and lithologic layers — genuinely can rank where rare earth occurrences are likely. But they predict occurrence, not orebodies. A cell in the top percentile of a prospectivity heatmap carries no information about grade continuity, metallurgy, or whether the ground could ever support a Measured-and-Indicated estimate filed under NI 43-101. The hottest pixels on any national map can sit on ground that fails the first screen gate instantly. Prospectivity maps are exploration triage; the filings scoreboard counts inventory. Never conflate the two.

Before crediting any US rare earth headline, de-price the filing along the five axes below. A deposit that survives all five earns a row on the scoreboard; nothing else does.

Round Top, in Hudspeth County, Texas, is the right specimen for watching a billion-tonne headline fail quietly. Held by USA Rare Earth, with Texas Mineral Resources Corp. retaining an interest, its sole published resource is the Behre Dolbear NI 43-101 technical report: 1.63 billion tonnes classified entirely INFERRED, grading 580 ppm TREO. Two flags rise before any economics are run. Inferred-only means no Measured-plus-Indicated filing exists anywhere in the record, so under this guide's Gate 1 the project files as optionality on status alone; and the estimate is more than a decade old, assembled under superseded classification practice that no Qualified Person would carry into a 2026 financing un-refreshed. The arithmetic below runs anyway, because the rejection turns out to be overdetermined — Round Top fails twice, independently.

Step one decomposes the basket, and the distribution table is where Round Top stops resembling the bastnäsite profiles earlier in this guide. Neodymium plus praseodymium together run only in the low-to-mid teens as a share of TREO, with an unusually pronounced yttrium-heavy tilt — dysprosium and terbium well above carbonatite norms. Multiply the grade by that teens-range share and each tonne of rock carries about eight-hundredths of a kilogram of contained NdPr — the first ledger line below. Nothing about 1.63 billion tonnes changes this division; spreading the same metal across more rock moves the denominator, never the per-tonne value.

Step three brings costs, where structure matters more than false precision. According to the U.S. Government Accountability Office, a cost estimate is "the summation of individual cost elements, using established methods and valid data, to estimate the future costs of a program, based on what is known today." USA Rare Earth's preliminary economic assessment — filed five years ago and, by CIM convention, too preliminary to ground an economic decision — summed its elements on then-current data: open-pit mining, crushing, agglomeration, acid heap leach, solvent extraction, plus initial capital. Restate each element to 2026 dollars using the inflation adjustment specified in this guide's limitations section, and the stack lands comfortably above the gross ceiling. Divide back through the basket and the implied break-even NdPr price sits far above the screen — closable only by outsized Dy/Tb prices or the undeveloped lithium and gallium co-product credits the sponsors promote, neither of which appears in any filing as separated-oxide revenue.

Worked Case

The lesson generalizes because the uncertainties stack independently: an inferred-only category, a decade-old estimate vintage, and a heap-leach-on-rhyolite flowsheet never demonstrated at commercial scale. Remove any one and the per-tonne math still dies at the screen. Upgrade the category tomorrow — the basket value is unchanged. Re-drill and re-date the estimate — the grade does not move. Prove the flowsheet — recovery haircuts push realized value further below the ceiling, not above it. The transferable rule: headline tonnage divided by grade tells you nothing until multiplied by basket share, price, and recovery, in that order. Run that four-factor chain on any billion-tonne headline; if the product cannot cover the project's own per-tonne cost line, file it as optionality, not supply.

According to Wikipedia's feasibility entry, judging any project reduces to two criteria — the cost required and the value to be attained — and most US rare earth pitches ask you to accept one of those numbers on faith. The five rules below turn any technical report into a pass/fail verdict in one sitting. They run in strict order: the first rule a deposit fails ends its review, and whenever two verdicts conflict, the stricter call wins.

Step one is where the tonnage-equals-supply myth dies. In estimation practice, Measured and Indicated are the only categories with drill density tight enough to defend a production decision; Inferred is a geological hypothesis with coordinates, and it cannot legally support one. A billion-tonne inferred headline is inventory, not supply, and it exits any honest supply count at step one — no exceptions. Round Top, dissected in the worked case above, is the canonical specimen; this rule lets you run the same autopsy on the next headline yourself.

Step two takes less effort than any geologic argument and kills more candidates: read the title page. Only NI 43-101 and S-K 1300 estimates count as verified US supply. A resource reported under JORC by a US-listed issuer stays unverified until a Qualified Person restates it under one of those two codes, however attractive the numbers look — classification definitions differ between codes, and every ranking in this guide assumes one common legal standard.

Step three: compute the basket yourself instead of trusting the study's own sensitivity table. Value per tonne of ore equals TREO grade times the combined neodymium-plus-praseodymium basket share times $70/kg, plus dysprosium and terbium credits at trailing twelve-month averages, and it must reach at least twice the modeled all-in operating cost per tonne. The 2× buffer is not decoration. According to the government evaluation authority IMED, as reported by The Business Standard, weak feasibility studies and weak economic analysis are consistent predictors of time and cost overruns — a thinner margin leaves a project one recovery shortfall from negative economics.

How to Choose Well

Step five checks the calendar before the cash flows. Reject any economic study older than 24 months or built on a pre-2022 cost deck, and accept no NPV or IRR claim until a refreshed study lands at current capex and opex. A compliant refresh is on the record: according to Newsfile Corp's August 21, 2026 release, the new PFS builds on previous technical work while incorporating an updated Mineral Resource Estimate, current engineering standards, updated cost estimates, revised economic assumptions and current regulatory requirements. For contrast, a December 30, 2021 Investing News Network item declared a project "economically compelling, even at a price of US$35/lb U3O8" — a resilience claim priced in a different commodity and a dead cost world, which fails step five on age alone no matter how compelling the wording.

Two housekeeping rules bind the set. When applying two rules yields conflicting verdicts, impose the stricter one — ambiguity never upgrades a deposit. And re-run all five steps against fresh filings every time a new technical report drops, because every verdict in this guide expires with the next filing. Calibrate what you will accept, too: according to Wikipedia's feasibility entry, a well-designed study carries historical background, product description, accounting statements, operations and management detail, marketing research, financial data, and legal and tax obligations — a document missing half that list is a brochure wearing a study's clothes.

Step two takes less effort than any geologic argument and kills more candidates: read the title page. Only NI 43-101 and S-K 1300 estimates count as verified US supply. A resource reported under JORC by a US-listed issuer stays unverified until a Qualified Person restates it under one of those two codes, however attractive the numbers look — classification definitions differ between codes, and every ranking in this guide assumes one common legal standard.

Step three: compute the basket yourself instead of trusting the study's own sensitivity table. Value per tonne of ore equals TREO grade times the combined neodymium-plus-praseodymium basket share times $70/kg, plus dysprosium and terbium credits at trailing twelve-month averages, and it must reach at least twice the modeled all-in operating cost per tonne. The 2× buffer is not decoration. According to the government evaluation authority IMED, as reported by The Business Standard, weak feasibility studies and weak economic analysis are consistent predictors of time and cost overruns — a thinner margin leaves a project one recovery shortfall from negative economics.

Step four reroutes rather than eliminates. Search the filings and press releases for a binding price floor — a take-or-pay offtake or a government equity deal with guaranteed pricing. At $100/kg or above, stop grading the asset against spot entirely: reclassify it as policy-backed and evaluate contract terms and duration instead. Mountain Pass, the floored California operation covered earlier, belongs in that bucket and nowhere else; a floored deposit and a spot-exposed deposit are different asset classes and must never share one ranking, because one's revenue is a contract line item while the other's is a price forecast.

Step five checks the calendar before the cash flows. Reject any economic study older than 24 months or built on a pre-2022 cost deck, and accept no NPV or IRR claim until a refreshed study lands at current capex and opex. A compliant refresh is on the record: according to Newsfile Corp's August 21, 2026 release, the new PFS builds on previous technical work while incorporating an updated Mineral Resource Estimate, current engineering standards, updated cost estimates, revised economic assumptions and current regulatory requirements. For contrast, a December 30, 2021 Investing News Network item declared a project "economically compelling, even at a price of US$35/lb U3O8" — a resilience claim priced in a different commodity and a dead cost world, which fails step five on age alone no matter how compelling the wording.

Two housekeeping rules bind the set. When applying two rules yields conflicting verdicts, impose the stricter one — ambiguity never upgrades a deposit. And re-run all five steps against fresh filings every time a new technical report drops, because every verdict in this guide expires with the next filing. Calibrate what you will accept, too: according to Wikipedia's feasibility entry, a well-designed study carries historical background, product description, accounting statements, operations and management detail, marketing research, financial data, and legal and tax obligations — a document missing half that list is a brochure wearing a study's clothes.

What to do next

Frequently Asked Questions

How far did neodymium-praseodymium oxide prices fall during the crash that gutted most US deposit economics?

NdPr oxide slid from about $220/kg in February 2022 to about $52/kg by February 2024 — a 76% collapse that pushed most American baskets below the $70/kg screen.

How much separated NdPr oxide has Mountain Pass actually produced, and what is it targeting?

Per MP Materials' company production reports, output rose from roughly 200 tonnes in 2023 to 1,294 tonnes in 2024, against a ~6,000-tonne target for 2025.

Why do the biggest US tonnage headlines — like Round Top and Halleck Creek — fail the three-gate screen?

Round Top's 1.63-billion-tonne endowment and Halleck Creek's 64.3 million tonnes inferred at 3,240 ppm (just 0.324%) TREO — filed under JORC 2012 as an ASX listing — both wash out once Inferred tonnes are barred, non-NI 43-101/S-K 1300 standards are excluded, and baskets price at $70/kg.

What exactly keeps Mountain Pass solvent when spot NdPr trades below $70/kg?

Its contractual $110/kg floor lifts the threshold case — 0.5% TREO ore with NdPr at 15% of basket — from roughly $53 to roughly $83 per tonne of ore, a 57% swing that truncates spot exposure entirely.

Can Inferred resources ever anchor a mine plan or feasibility study?

No — under the CIM Definition Standards (May 2014), only Measured and Indicated tonnes can support pre-feasibility or feasibility studies and convert to mineral reserves, while any PEA touching Inferred material must carry the 'preliminary in nature' label.

Which US challenger comes closest to being real supply, and has it actually made oxide?

Rare Element Resources' Bear Lodge — roughly 18 million tonnes combined at approximately 1.4% TREO under NI 43-101 — feeds a DOE-backed demonstration plant in Upton, Wyoming that began producing separated NdPr oxide samples through 2024–2025, with GE Vernova stating magnet-supply intent.

Quick answers

Also worth reading: How satellite imaging helps professionals scout for rare mineral deposits: How satellite imaging helps professionals · China 2026 Rare Earth Quotas: NdPr Oxide +12% in January: China 2026 Rare Earth Quotas: · Unlock hidden energy with vital trace minerals from the earth: Unlock hidden energy with vital