Approximate 2025 total
Rare earth mining race · 2026
No One Comes
Near China.
The rest of the world is building mines. China still dominates the factories that turn rare earths into magnets. That is the lead everyone else has to close.
See the gapUpdated 4 September 2026 · Production years shown on each chart · Project reporting through 30 August 2026
On this pageTakeaway
01 / Overall standings
The further down the chain,
the bigger China’s lead.
Share of world production, 2024.
China compared with everyone else combined.
ChinaEveryone else combinedEach track = 100% of world output
01
Mining
Taking magnet materials out of the ground
60%China’s share
Everyone else40%
02
Refining
Separating the materials factories need
91%China’s share
Everyone else9%
03
Making magnets
Producing sintered permanent magnets
94%China’s share
Everyone else6%
Mining and refining cover neodymium, praseodymium, dysprosium and terbium together. The final chart measures sintered magnets. “Everyone else” describes where output is made, not whether it is independent of Chinese inputs or ownership. Source: IEA, 2024 production shares ↗
02 / Country standings
China mines more than
the rest of the world combined.
All rare earths in the USGS mining total.
2025 estimates · A broader basket than the magnet-material charts above.
69.2% of world output
Approximately 30.8% of world output
5.3×
China’s mine output is 5.3 times that of the United States, the nearest country behind it.
Inside China’s lead ↗Rank / producerWorld share · 0–100% scale2025 output / share
- 01China270,000 t69.2%
- 02United States51,000 t13.1%
- 03Australia29,000 t7.4%
- 04Myanmar22,000 t5.6%
- 05Thailand4,800 t1.2%
See the remaining producers Ranks 6–12 + other countries
- 06India2,900 t0.7%
- 07Madagascar2,700 t0.7%
- 08Russia2,600 t0.7%
- 09Brazil2,000 t0.5%
- 10Nigeria1,500 t0.4%
- 11Vietnam150 t0.04%
- 12Malaysia110 t0.03%
Other countries: 550 tonnes (about 0.14%). This grouped remainder is not a ranked country.
Tonnes are measured as rare-earth-oxide equivalent: a common unit for comparing mine output. They are not tonnes of finished magnets. Shares use the USGS rounded world total of 390,000 tonnes, so the entries need not add to exactly 100%. The approximate rest-of-world total is 390,000 minus 270,000. Source: USGS Mineral Commodity Summaries 2026 [1]
03 / Material standings
Where can the challengers gain ground?
China leads the overall chain. The best alternative changes by material. Open a material to see what is made today, what is planned, and what remains unknown.
Plain-language guideMine output is material after mining and initial concentration, before individual rare earths are separated. A separated oxide is an individual rare-earth chemical ready for the next factory step. Capacity is what a plant is designed to make, not what it produces. Best alternative means the strongest documented non-China route at the same product stage. It describes operating output when available; a plant plan appears only as a separate future case.
How we countWe count the exact product made today: mined material, an individual rare-earth chemical, metal, or a product accepted by customers. A project plan is not current output. If the supply line is unclear, we do not count it as independent.
Compare all 17 materials side by side
| Material | Made now | Possible by 2030 | Our call |
|---|---|---|---|
| ScScandiumAluminium alloys and fuel cellsOxide is made today; alloy output is unclear | Canada makes scandium oxide today and is expanding to 9 tonnes a year; a Philippines-to-Japan route announced 7.5 tonnes a year in 2016 | An Australian project plans 60 tonnes a year from 2028 | Canada is the only operating source we could confirm. A Philippines-to-Japan line has announced capacity. Australia could become the largest new source after 2028. |
| YYttriumCeramics, lasers and phosphorsOxide can be made; later factory steps are unclear | France and India can separate yttrium; annual output is not published | Malaysia plans capacity of 1,100 tonnes a year, subject to approvals and start-up | France and India can separate yttrium, but their annual tonnes are not published. Malaysia has the largest plant plan, but its actual contribution cannot be measured until it produces. |
| LaLanthanumCatalysts, glass and batteriesOre and oxide are produced; later steps are unclear | Australian ore becomes oxide in Malaysia; no lanthanum-only figure is published | Eneabba plans a combined lanthanum-cerium product; no separate figure | Other countries know how to make it, but published totals are too incomplete to calculate their share of world lanthanum supply. |
| CeCeriumPolishing, catalysts and glassOre and oxide are produced; later steps are unclear | Australian ore becomes oxide in Malaysia; no cerium-only figure is published | Eneabba plans a combined lanthanum-cerium product; no separate figure | Cerium oxide is made outside China, but companies do not publish enough material-specific output to calculate the share made outside China. |
| PrPraseodymiumHigh-strength permanent magnetsOre and oxide are produced; US factories are growing | 9,859 tonnes is the sum of the latest disclosed combined neodymium-praseodymium outputs from different reporting periods; no praseodymium-only split or world percentage is available | Australia plans 7,740 tonnes a year of combined NdPr oxide; White Mesa plans 6,294 tonnes of overall NdPr capacity in 2029; France and Canada add different product forms | The Australia-to-Malaysia route leads current oxide supply outside China. Canada has the clearest new allied metal line, while the United States is closer to making the final magnets at home. |
| NdNeodymiumMotors, turbines and electronicsOre and oxide are produced; US factories are growing | 9,859 tonnes is the sum of the latest disclosed combined neodymium-praseodymium outputs from different reporting periods; no neodymium-only split or world percentage is available | Australia plans 7,740 tonnes a year of combined NdPr oxide; White Mesa plans 6,294 tonnes in 2029; France's Caremag adds didymium and neodymium; Canada adds NdPr metal | The strongest disclosed capabilities are spread across Australian ore, Malaysian separation, Canadian metal, Japanese buyers and growing US magnet factories. Public evidence does not yet show that they operate as one continuous commercial chain. |
| PmPromethiumNuclear batteries and researchMade in reactors, not mined | No commercial mined supply | Isotope production, not a mine project | Promethium is made in reactors, so it does not belong in a mine-output race with China. |
| SmSamariumSamarium-cobalt magnets and defenceOxide can be made; the magnet chain is not documented | Malaysia, India and France can make samarium oxide; annual output is not published | Malaysia plans up to 1,100 tonnes a year, subject to approvals, raw material and a successful start-up; the United States plans up to 140 tonnes a year | Lynas reported its first samarium oxide in Malaysia in March 2026. The United States has the clearest second plan for material used in samarium-cobalt magnets. |
| EuEuropiumDisplays, phosphors and security inksOxide can be made; later products are not traced | France and India can make europium oxide; annual output is not published | The United States plans up to 20 tonnes a year by the end of 2028 | White Mesa is the only new project with a public figure: up to 20 tonnes a year by the end of 2028. It is still being built. |
| GdGadoliniumMedical imaging and specialist magnetsOxide can be made; later products are not traced | France and India can make gadolinium oxide; annual output is not published | Malaysia plans up to 400 tonnes a year, subject to approvals, raw material and a successful start-up; the United States plans up to 140 tonnes a year | Malaysia publishes the larger plan. The US project would add another named future plant outside China, alongside operating capability in France and India. |
| TbTerbiumHigh-temperature magnetsCombined oxide is produced; the magnet route is incomplete | Malaysia reported 62 tonnes of combined dysprosium-terbium in the 2026 financial year; no terbium-only figure or world percentage is available | France, Malaysia, Canada and White Mesa plan 235 tonnes a year of separate terbium capacity; Malaysia's share is subject to approvals, raw material and a successful start-up; MP Materials plans 200 tonnes a year of combined dysprosium-terbium | Malaysia reports a combined dysprosium-terbium product, but not the terbium tonnes. France publishes the largest separate next capacity at 90 tonnes a year, followed by White Mesa's 80-tonne 2029 plan, Malaysia at 50 and Canada at 15. MP's 200-tonne line also combines both materials. |
| DyDysprosiumHeat-resistant magnetsCombined oxide is produced; the magnet route is incomplete | Malaysia reported 62 tonnes of combined dysprosium-terbium in the 2026 financial year; no dysprosium-only figure or world percentage is available | France, Malaysia, Canada and White Mesa plan 1,068 tonnes a year of separate dysprosium capacity; Malaysia's share is subject to approvals, raw material and a successful start-up; MP Materials plans 200 tonnes a year of combined dysprosium-terbium | Malaysia reports a combined dysprosium-terbium product, but not the dysprosium tonnes. France publishes the largest separate next capacity at 500 tonnes a year, followed by White Mesa's 288-tonne 2029 plan, Malaysia at 250 and Canada at 30. MP also reports both materials together. |
| HoHolmiumLasers, medical and specialist magnetsSeparation is possible; annual tonnes are not published | France can separate holmium; annual output is not published | Australia and the United States have plans, but neither gives a holmium-only figure | Australia has a funded refinery plan and the United States has a mine plan, but neither publishes annual holmium output. |
| ErErbiumFibre optics, lasers and alloysSeparation is possible; annual tonnes are not published | France can separate erbium; annual output is not published | Australia and the United States have plans, but neither gives an erbium-only figure | France can separate erbium. Australia and the United States have plans, but published data do not show how much erbium any route could supply at each stage. |
| TmThuliumLasers, imaging and radiation sourcesNo operating product route is documented | No reliable current output figure outside China | The United States and Australia have plans, but neither gives a thulium-only figure | This is one of the hardest materials to check. A mixed-product mine should not be counted as thulium output without a separate figure. |
| YbYtterbiumFibre lasers and specialist alloysSeparation is possible; annual tonnes are not published | France can separate ytterbium; annual output is not published | Australia and the United States have plans, but neither gives a ytterbium-only figure | France can separate ytterbium. Australia and the United States have plans, but none publishes usable annual output. |
| LuLutetiumMedical scanners, catalysts and researchOxide is planned, not operating | No reliable current output figure outside China | Malaysia plans capacity of 10 tonnes a year, subject to approvals and start-up | Malaysia's planned 10-tonne capacity is the only public project figure. It is not output yet. |
Sources for every row
Check the evidence for all 17 materials
Today's output stays separate from project plans. When a company reports several materials in one number, we keep that number combined.
NdNeodymiumAustralia and the US mine it; Malaysia and the US separate itView position +Close −
Raw materialMade todayIndividual oxideMade todayMetal or alloyProduction growingCustomer-ready magnetsProduction growing
The same disclosed floor applies as Pr: 7,260 tonnes of combined Lynas NdPr in the 2026 financial year and 2,599 tonnes from MP Materials in 2025. They come from different reporting periods, and no complete world total is available. An Nd-only split would be invented. 5 7
Eneabba and Nolans form the strongest funded oxide pipeline. Arafura describes Nolans' 4,440-tonne annual rate as about 4% of forecast global demand. White Mesa says its existing US line can make up to 1,000 tonnes a year and plans overall capacity of 6,294 tonnes in 2029; capacity is not output. France's Caremag separately plans 650 tonnes a year of didymium oxide and 150 tonnes of neodymium oxide. Canada's SRC has a separate 525-tonne-a-year NdPr metal design and targets commercial intake in the third quarter of 2027. If these separate projects establish compatible commercial links, US magnet plants could help connect them into a non-China chain. 9 10 8 11 25
The strongest disclosed capabilities are spread across Australian ore, Malaysian separation, Canadian metal, Japanese buyers and growing US magnet factories. Public evidence does not yet show that they operate as one continuous commercial chain.
PrPraseodymiumAustralian ore becomes oxide in MalaysiaView position +Close −
Raw materialMade todayIndividual oxideMade todayMetal or alloyProduction growingCustomer-ready magnetsProduction growing
Lynas reported 7,260 tonnes of combined NdPr in the 2026 financial year. MP Materials reported 2,599 tonnes of combined NdPr in 2025. The 9,859-tonne sum is therefore a disclosed floor drawn from different reporting periods, not a synchronized annual total or market share. Neither company publishes a separate Pr total. 5 7
Australia leads the funded oxide build: Eneabba has a 3,300-tonne NdPr financing case and Nolans is designed for 4,440 tonnes a year from around mid-2029. Arafura describes the Nolans rate as about 4% of forecast global demand. White Mesa says its existing US line can make up to 1,000 tonnes a year and plans overall capacity of 6,294 tonnes in 2029; capacity is not output. France's Caremag separately plans 650 tonnes a year of didymium oxide and 150 tonnes of neodymium oxide. Canada's SRC has a separate 525-tonne-a-year NdPr metal design and targets commercial intake in the third quarter of 2027. These different product forms are not added together. 9 10 8 11 25
The Australia-to-Malaysia route leads current oxide supply outside China. Canada has the clearest new allied metal line, while the United States is closer to making the final magnets at home.
DyDysprosiumAustralian ore becomes oxide in MalaysiaView position +Close −
Raw materialMade todayIndividual oxideMade today; reported with terbiumMetal or alloyNo usable output figureCustomer-ready magnetsFull chain not documented
Lynas's 62 tonnes of combined Dy and Tb is the only clearly reported commercial ex-China stream. 5
Caremag plans 500 tonnes a year of dysprosium, White Mesa plans overall capacity of 288 tonnes in 2029, Lynas a maximum 250 tonnes and Canada's SRC 30 tonnes. Lynas's Malaysian figure remains subject to approvals, enough raw material and a successful start-up. The separate-material designs total 1,068 tonnes a year. White Mesa's first heavy-circuit stage is 120 tonnes; the larger number belongs to a later company plan. MP is testing a 200-tonne-a-year combined dysprosium-terbium line. Serra Verde estimates 164 tonnes of dysprosium oxide-equivalent inside its 2027 mixed carbonate, not separated oxide. 6 8 11 24 16 25
Malaysia reports a combined dysprosium-terbium product, but not the dysprosium tonnes. France publishes the largest separate next capacity at 500 tonnes a year, followed by White Mesa's 288-tonne 2029 plan, Malaysia at 250 and Canada at 30. MP also reports both materials together.
TbTerbiumAustralian ore becomes oxide in MalaysiaView position +Close −
Raw materialMade todayIndividual oxideMade today; reported with dysprosiumMetal or alloyNo usable output figureCustomer-ready magnetsFull chain not documented
Lynas reported 62 tonnes of combined Dy and Tb in the 2026 financial year. It did not publish the individual Tb split. Estonia's Silmet has produced small-scale precursor solutions, not a comparable oxide total. 5 13
Caremag plans 90 tonnes a year of terbium, White Mesa plans overall capacity of 80 tonnes in 2029, Lynas a maximum 50 tonnes and Canada's SRC 15 tonnes. Lynas's Malaysian figure remains subject to approvals, enough raw material and a successful start-up. The separate-material designs total 235 tonnes a year. White Mesa's first heavy-circuit stage is 20 tonnes; the larger number belongs to a later company plan. MP is testing a 200-tonne-a-year combined dysprosium-terbium line; the split is not published. Serra Verde estimates 29 tonnes of terbium oxide-equivalent inside its 2027 mixed carbonate, not separated oxide. 6 8 11 24 16 25
Malaysia reports a combined dysprosium-terbium product, but not the terbium tonnes. France publishes the largest separate next capacity at 90 tonnes a year, followed by White Mesa's 80-tonne 2029 plan, Malaysia at 50 and Canada at 15. MP's 200-tonne line also combines both materials.
ScScandiumCanada makes oxide; Philippine material goes to JapanView position +Close −
Raw materialMade todayIndividual oxideMade todayMetal or alloyAmount unclearCustomer-ready productCapacity expanding; output undisclosed
Rio Tinto operates Canadian oxide capacity and is expanding to 9 tonnes a year. Sumitomo announced 7.5 tonnes a year of scandium-oxide-equivalent capacity for a Philippines-to-Japan route in 2016; current output is not public. 15 22
Australia's Syerston study plans 60 tonnes a year from 2028. It still needs proof that it can be built and reach steady output. 23
Canada is the only operating source we could confirm. A Philippines-to-Japan line has announced capacity. Australia could become the largest new source after 2028.
YYttriumFrance can separate it; Malaysia is buildingView position +Close −
Raw materialMade todayIndividual oxideMade today; tonnes unknownMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
Solvay in France has individual yttrium capability and IREL produces yttrium oxide in India, but neither publishes a dependable annual total. 12 21
Lynas plans a maximum 1,100 tonnes a year in Malaysia, subject to approvals, enough raw material and reaching steady output. Serra Verde estimates 1,534 tonnes of yttrium oxide-equivalent inside its 2027 mixed carbonate, not separated oxide. Round Top lists yttrium without a separate figure or a usable start date. 6 16 14
France and India can separate yttrium, but their annual tonnes are not published. Malaysia has the largest plant plan, but its actual contribution cannot be measured until it produces.
LaLanthanumAustralian ore becomes oxide in MalaysiaView position +Close −
Raw materialMade todayIndividual oxideMade today; tonnes unknownMetal or finished formAmount unclearCustomer-ready productLater steps unclear
Lynas uses Mt Weld material to make products ready to sell, but does not say how much of its 13,089-tonne 2026 financial-year total was lanthanum. IREL also makes lanthanum oxide in India without publishing an annual figure. 5 21
Eneabba's design contains lanthanum in a combined LaCe stream. It does not provide an individual La figure. 9
Other countries know how to make it, but published totals are too incomplete to calculate their share of world lanthanum supply.
CeCeriumAustralian ore becomes oxide in MalaysiaView position +Close −
Raw materialMade todayIndividual oxideMade today; tonnes unknownMetal or finished formAmount unclearCustomer-ready productLater steps unclear
Lynas makes cerium in Malaysia from Australian raw material, and IREL makes cerium oxide in India. Lynas does not publish a separate cerium figure inside its 5,829-tonne balance of products other than neodymium-praseodymium. 5 21
Eneabba's design contains cerium in a combined lanthanum-cerium stream, with no cerium-only annual figure. 9
Cerium oxide is made outside China, but companies do not publish enough material-specific output to calculate the share made outside China.
PmPromethiumMade in reactors, not minesView position +Close −
Raw materialNot minedIndividual oxideNot minedMetal or finished formNot minedEnd useMade in reactors
Promethium has no stable naturally occurring commercial ore supply. It should not be mixed into mine-output comparisons. 20
Future supply depends on isotope-production capability rather than a rare-earth mine or solvent-extraction plant. 20
Promethium is made in reactors, so it does not belong in a mine-output race with China.
SmSamariumMalaysia has output; the US is buildingView position +Close −
Raw materialMade todayIndividual oxideMade today; tonnes unknownMetal or alloyNo usable output figureCustomer-ready magnetsFull chain not documented
Lynas made its first samarium oxide in Malaysia in March 2026. IREL produces samarium oxide in India and Solvay has French capability. None publishes a dependable annual Sm total. 5 21 12
Lynas plans a maximum 1,100 tonnes a year, subject to approvals, enough raw material and reaching steady output. White Mesa plans up to 140 tonnes a year by the end of 2028. 6 8
Lynas reported its first samarium oxide in Malaysia in March 2026. The United States has the clearest second plan for material used in samarium-cobalt magnets.
EuEuropiumFrance can separate it; the US is buildingView position +Close −
Raw materialMade todayIndividual oxideMade today; tonnes unknownFinished formNo usable output figureCustomer-ready productLater steps unclear
IREL produces individual europium oxide in India and Solvay has French capability, but neither publishes an annual Eu total. 21 12
White Mesa publishes a maximum design quantity of 20 tonnes a year by the end of 2028. Lynas lists Eu only as a possible later process design addition, not committed capacity. 8 6
White Mesa is the only new project with a public figure: up to 20 tonnes a year by the end of 2028. It is still being built.
GdGadoliniumFrance can separate it; Malaysia and the US are buildingView position +Close −
Raw materialMade todayIndividual oxideMade today; tonnes unknownMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
Solvay in France has individual gadolinium capability and IREL produces it in India, but neither publishes a dependable annual total. 12 21
Lynas plans a maximum 400 tonnes a year in Malaysia, subject to approvals, enough raw material and reaching steady output. White Mesa plans up to 140 tonnes a year by the end of 2028. 6 8
Malaysia publishes the larger plan. The US project would add another named future plant outside China, alongside operating capability in France and India.
HoHolmiumFrance can separate itView position +Close −
Raw materialNot publishedIndividual oxideCan be made; tonnes unknownMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
Solvay can separate the material in France, but public records give no annual Ho output. 12
Round Top and the funded Eneabba refinery list holmium capability without an individual output split. Lynas says Ho may follow only if commercial agreements justify another process design step. 14 9 6
Australia has a funded refinery plan and the United States has a mine plan, but neither publishes annual holmium output.
ErErbiumFrance can separate itView position +Close −
Raw materialNot publishedIndividual oxideCan be made; tonnes unknownMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
France has separation capability, but an operating non-Chinese annual Er total is not disclosed. 12
Round Top and the funded Eneabba refinery list erbium capability without an individual output split. Lynas considers Er a later, contract-dependent addition. 14 9 6
France can separate erbium. Australia and the United States have plans, but published data do not show how much erbium any route could supply at each stage.
TmThuliumFrance can separate it; a US project is plannedView position +Close −
Raw materialAmount not publishedIndividual oxideSeparation capability; output unclearMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
No dependable current non-Chinese annual production figure is published. 20
Round Top and the funded Eneabba refinery both list thulium capability, but neither publishes a Tm production split or a financed date for making thulium oxide on its own. 14 9
This is one of the hardest materials to check. A mixed-product mine should not be counted as thulium output without a separate figure.
YbYtterbiumFrance can separate it; a US project is plannedView position +Close −
Raw materialAmount not publishedIndividual oxideCan be made; tonnes unknownMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
Solvay can perform the chemistry, but no current annual Yb total is disclosed. 12
Round Top and the funded Eneabba refinery list ytterbium without a separate figure. Lynas describes ytterbium as a possible later product, not committed output. 14 9 6
France can separate ytterbium. Australia and the United States have plans, but none publishes usable annual output.
LuLutetiumFrance can separate it; Malaysia is buildingView position +Close −
Raw materialAmount not publishedIndividual oxidePlanned; not operatingMetal or finished formNo usable output figureCustomer-ready productLater steps unclear
No dependable ex-China annual separated-oxide total is public. 12
Lynas plans a maximum 10 tonnes a year of lutetium in Malaysia, subject to approvals, enough raw material and reaching steady output. Eneabba and Round Top also list lutetium without separate annual figures. 6 9 14
Malaysia's planned 10-tonne capacity is the only public project figure. It is not output yet.
04 / Project standings
A mine outside China may still depend on China
We count supply as independent only when the product can reach the measured step without a required Chinese owner, buyer or factory. If that route is unclear, it does not count.
Mined in ChinaChinese material from the first step
Sent through ChinaA Chinese plant is still required
Controlled from ChinaA Chinese owner or committed buyer controls the output
Independent route documentedNo required Chinese step at the point measured
Route unclearDoes not count as independent
Check the supply route behind each project
| Project | Material and amount | Position now | What counts today |
|---|---|---|---|
| Mt Weld to Lynas MalaysiaAustralia to Malaysia | Neodymium-praseodymium; dysprosium-terbium7,260 tonnes of combined neodymium-praseodymium; 62 tonnes of combined dysprosium-terbium | Producing now | Counts as separated oxideAustralian ore becomes separated oxide in Malaysia without a required Chinese step. The same tonnes count once, not once for each country. 5 |
| Mountain PassUnited States | Neodymium-praseodymium; planned dysprosium-terbium line2,599 tonnes of combined neodymium-praseodymium; a 200-tonne-a-year combined dysprosium-terbium line in start-up tests | Neodymium-praseodymium is producing; the heavy line is being tested | Current neodymium-praseodymium counts; heavy output does not yetMP stopped selling to China in 2025, and its Shenghe purchase agreement expired in January 2026. The heavy line stays out of current supply until it makes material ready to sell. 7 24 |
| EneabbaAustralia | Neodymium-praseodymium; dysprosium-terbium; other rare earths3,300 tonnes of neodymium-praseodymium in the published project case; 725 to 750 tonnes of dysprosium-terbium at maximum | Being built | No current outputThe refinery is funded and does not require a Chinese step, but planned capacity is not production. The higher heavy-rare-earth figure also needs ore from other suppliers. 9 |
| White Mesa heavy circuitUnited States | Neodymium-praseodymium, dysprosium, terbium, samarium, europium and gadoliniumExisting design: up to 1,000 tonnes a year of NdPr. First heavy stage: 120 Dy, 20 Tb, 140 Sm, 20 Eu and 140 Gd. Planned 2029 overall capacity: 6,294 NdPr, 288 Dy and 80 Tb | Light-separation capacity exists; heavy circuit is being built; a second expansion is planned | Capacity is not output; no current commercial heavy-rare-earth outputTest samples show that the heavy process works. The first commercial circuits target the end of 2027 and 2028. The much larger 2029 figures remain company plans subject to raw material, finance, buyers and permits. 8 |
| CaremagFrance | Neodymium-praseodymium, neodymium, dysprosium and terbiumPlanned yearly capacity: 650, 150, 500 and 90 tonnes, respectively | Being built; first output targeted for late 2026 | No current outputThe plant has finance and a Japanese buyer, but planned capacity is not current supply before start-up. 11 |
| SRC minerals-to-metals facilityCanada | Neodymium-praseodymium metal; dysprosium oxide; terbium oxideDesign capacity of 525, 30 and 15 tonnes a year, respectively | Early start-up testing; commercial intake targeted for the third quarter of 2027 | No regular commercial output yetCanada has funded metal and individual heavy-oxide capacity, but capacity is not output. The page keeps it separate from Australian and Malaysian oxide production. 25 |
| Pela Ema / Serra VerdeBrazil | Neodymium, praseodymium, dysprosium, terbium and yttrium in one mixed productThe 2027 company estimate includes 164 tonnes of dysprosium, 29 tonnes of terbium and 1,534 tonnes of yttrium inside the mixed product | Selling mixed carbonate today | The mixed product counts; separate oxides do notThe material figures are company estimates inside mixed carbonate. Public evidence does not yet show that these individual oxides are separated outside China. 16 |
| NguallaTanzania | Concentrate rich in neodymium and praseodymiumThe study plans 16,200 tonnes a year of total rare-earth oxide inside concentrate | Permitted; not producing | No output counts as independentShenghe controls the project and is due to buy all planned concentrate. A mine outside China is not independent when control and the buyer remain Chinese. 18 19 |
| Kachin and Shan ionic claysMyanmar to China | Mixed material rich in dysprosium and terbium22,000 tonnes of rare-earth oxide in the national mine estimate | Operating, but the data and route are unclear | No separate output counts as independentThe estimate gives no reliable amount for each material, and the main route uses Chinese technicians, buyers and separation plants. 1 17 |
| Round TopUnited States | Yttrium, gadolinium, terbium, dysprosium, holmium, erbium, thulium, ytterbium and lutetiumNo annual amount for any individual oxide | Funded project; no usable start date | No current outputFunding makes the project more credible, but a list of materials in the deposit is not a yearly production figure. 14 |
What counts today
The 9,859 tonnes of combined neodymium-praseodymium is a sum of the latest annual disclosures from different reporting periods. The 62 tonnes is Lynas's combined dysprosium-terbium output in the 2026 financial year. Both sit on supply lines documented outside China, but neither is a complete world total or an individual-material figure, so no world percentage is calculated. Ngualla's project plan, Myanmar's broad mine estimate and every plant still being built add zero to today's output of individual rare-earth chemicals.
05 / The race to 2030
Can anyone close the gap by 2030?
These are EarthRarest editorial judgements. The revision triggers are part of the forecast, not small print.
By 31 December 2030
China loses mine share but keeps refining control.
China's share of magnet rare earth mining falls below 55 per cent, while its refining share remains at least 80 per cent. The ex-China mine pipeline is advancing faster than separation, and China continues to expand and consolidate its own processing system. 3
ConfidenceMedium
What would change this view
Revise if several large non-Chinese separation plants reach sustained output before 2029, or if Chinese refining is disrupted for more than six months.
By 31 December 2030
Mine plans outrun chemical-factory plans two to one.
By 2030, planned mining capacity outside China will be at least twice the capacity to make individual rare-earth chemicals. Australia, Brazil, the United States and several newer projects have much larger mining plans than factory plans for individual rare-earth chemicals. 3
ConfidenceHigh
What would change this view
Revise if published operating data show ex-China individual oxide capacity above half the corresponding mine and mixed-product capacity.
Through 31 December 2030
At least three projects still need government-backed buyers.
At least three projects outside China will still depend on governments guaranteeing a buyer or minimum price. New projects outside China often cost more. Government-backed buyers and minimum prices are becoming part of how they get financed, not a temporary rescue. 2
ConfidenceHigh
What would change this view
Revise if support contracts expire without replacement and the projects sustain market-priced output for four consecutive quarters.
Beyond the 17 rare earths
China’s lead extends beyond rare earths.
Hafnium and rhenium are the exceptions in the production data shown here. Every card states the year and whether it measures mining, refining or another production step.
Why separate charts? Antimony shows both its mine share and China's later refining share. Gallium uses new low-purity output. Bismuth and indium use refinery output. Tellurium follows the listed USGS table, where Sweden's entry is concentrate. Hafnium uses a five-year processed-output average. Rhenium uses primary recovery. Combining these tonnes would create a false comparison.
China mines 36% of antimony, then refines about 74%.
Antimony Flame retardants, lead-acid batteries, ammunition and infrared systems.
0%Share of world output100%
Russia and Tajikistan provide substantial mine supply, but China's share is slightly more than twice as high at the refining step. 26
Read the antimony guideChina refines seven of every eight tonnes of bismuth.
Bismuth Medicines, lead-free solders, pigments and fusible alloys.
0%Share of world output100%
South Korea is the next-largest producer at about 1,000 tonnes, one fourteenth of China's output. 27
Read the bismuth guideFor every 100 kilograms of new low-purity gallium, roughly 99 come from China.
Gallium Power chips, radio-frequency chips, LEDs and high-efficiency solar cells.
0%Share of world output100%
Australia's Wagerup project is being built for 100 tonnes a year, while Kazakhstan plans 15 tonnes. Neither project is current supply. 28
Read the gallium guideChina refines roughly three-quarters of germanium; no complete 2025 country tonnage table is public.
Germanium Fibre optics, infrared optics, chips and specialist solar cells.
0%Share of world output100%
The latest full USGS country table is from 2022. Without a current published country total, the market is harder to check during a disruption. 29
Read the germanium guideHafnium breaks the China pattern: France and the United States accounted for about 95% in the 2020 to 2024 benchmark.
Hafnium Jet-engine alloys, nuclear control rods and advanced chip dielectrics.
0%Share of world output100%
This historical five-year benchmark is used because a complete current country total is not published. 30
Read the hafnium guideChina refines seven in ten tonnes of indium.
Indium Touchscreens, displays, solders, solar coatings and semiconductors.
0%Share of world output100%
South Korea is the only alternative producer above 100 tonnes in the 2025 table. 31
Read the indium guideChile, not China, leads rhenium supply.
Rhenium Jet-engine superalloys and petroleum-reforming catalysts.
0%Share of world output100%
China ranks second with about one-quarter. Poland and the United States each add roughly one-eighth. 32
Read the rhenium guideChina refines about four of every five tonnes of tellurium.
Tellurium Thin-film solar panels, thermoelectric devices and specialist alloys.
0%Share of world output100%
Russia and Japan are the largest visible alternatives, but neither reaches 7% of the listed total. Sweden's separate entry is concentrate, not refinery output. 33
Read the tellurium guideCountry index
Go inside each country’s race.
Twenty country briefings: operating mines, factory plans, the materials they make and what could change by 2030.
Countries that refine or manufacture
Countries with operating processing or magnet factories, or strong plans to build them.
ChinaDominates processing and magnets
The conversion superpower
- Mine output
- 270,000 tonnes of rare-earth oxide · 69.2% of 2025 world output
- Constraint
- Depends on Myanmar for heavy rare earths
United StatesLarge mine, much smaller processing industry
The state-backed rebuild
- Mine output
- 51,000 tonnes of rare-earth oxide · 13.1% of 2025 world output
- Constraint
- Raw material and customer tests
AustraliaNo. 3 miner
The refinery race behind the mine numbers
- Mine output
- 29,000 tonnes of rare-earth oxide · 7.4% of 2025 world output
- Constraint
- Chemicals ready to sell
MalaysiaSeparation power
A separation power with a small mine
- Mine output
- 110 tonnes of rare-earth oxide · under 0.1% of 2025 world output
- Constraint
- Not enough Malaysian raw material
JapanState-backed buyer
The buyer that helps foreign mines secure finance
- Mine output
- Not reported
- Constraint
- Enough dysprosium and terbium raw material
South KoreaKorean companies are building plants abroad
The industrial network built across four countries
- Mine output
- Not reported
- Constraint
- Plants must pass customer tests
Countries trying to open new mines
Current miners and project countries trying to move from ore or mixed material to a lasting supply business.
BrazilOperating challenger
The clay challenger with one mine already running
- Mine output
- 2,000 tonnes of rare-earth oxide · 0.5% of 2025 world output
- Constraint
- No large factory to split mixed material
CanadaProcessor first
Processing first, mine later
- Mine output
- 0 tonnes of rare-earth oxide · 0% of 2025 world output
- Constraint
- Roads and finance
IndiaOxide producer
Oxides today, magnets still to build
- Mine output
- 2,900 tonnes of rare-earth oxide · 0.7% of 2025 world output
- Constraint
- No mass production of high-strength magnets
VietnamFactory before mine
The reserve reset and hidden magnet base
- Mine output
- 150 tonnes of rare-earth oxide · 0.04% of 2025 world output
- Constraint
- Deposits have not become working mines
SwedenIndustrial challenger
Waste first, new mine later
- Mine output
- Not listed
- Constraint
- Rights and permits
RussiaSmall operating chain
A real mine, an unfinished magnet chain
- Mine output
- 2,600 tonnes of rare-earth oxide · 0.7% of 2025 world output
- Constraint
- Separation and sanctions
Frontier and opaque supply
Material flows and projects where politics, logistics, disclosure or finance matter as much as the deposit.
MyanmarKey source of dysprosium and terbium ore
The heavy rare earth pressure point
- Mine output
- 22,000 tonnes of rare-earth oxide · 5.6% of 2025 world output
- Constraint
- Armed control
ThailandMine output cannot be independently checked
The 4,800-tonne production mystery
- Mine output
- 4,800 tonnes of rare-earth oxide · 1.2% of 2025 world output
- Constraint
- No named mines behind the national total
MadagascarSeventh-largest estimated miner
The monazite exporter facing its value-add test
- Mine output
- 2,700 tonnes of rare-earth oxide · 0.7% of 2025 world output
- Constraint
- Permits and finance
NigeriaTenth-largest estimated miner
The producer whose missing data is the story
- Mine output
- 1,500 tonnes of rare-earth oxide · 0.4% of 2025 world output
- Constraint
- Disclosure
GreenlandMining licence, no funded construction
Licensed geology, outsourced chemistry
- Mine output
- Not reported
- Constraint
- Not enough money; processing would happen abroad
TanzaniaMine controlled by a Chinese company
The permitted NdPr project now routed through Shenghe
- Mine output
- Not reported
- Constraint
- No funded construction approval
AngolaConstruction under pressure
The Western-backed build beside the Lobito Corridor
- Mine output
- No separate USGS figure
- Constraint
- US$150 million outstanding
South AfricaTwo routes near the gate
The waste-to-oxide route beside a high-grade restart
- Mine output
- Not reported
- Constraint
- Scale and product proof
Australia, the United States and Brazil can all add mine or mixed product before the end of the decade. Canada, Greenland and several African projects could follow if finance and infrastructure arrive. Yet the announced non-Chinese pipeline becomes progressively narrower after mining. IEA expects concentration to ease, while warning that planned refining and magnet capacity still trails new mining plans. 3
A tonne changes meaning as it moves through this chain because total rare earth oxide equivalent includes abundant elements with very different markets. Magnet security turns first on neodymium and praseodymium. Much smaller flows of dysprosium and terbium carry a different risk. Purity and passing customer tests can matter more than gross volume.
Malaysia matters despite having little domestic ore. Japan and South Korea matter despite not leading reserve tables. Processing knowledge, long-term finance and magnet customers can redirect a supply chain even when the rock is elsewhere.
Production is only part of the lead. Ownership, factory know-how and customer relationships determine who can sell the material and who can stop it crossing a border. Water, waste, radiation and community impacts also need to be measured, even when a project is promoted as essential for national security.
Myanmar shows why these questions cannot be separated: its heavy rare-earth raw material has supplied Chinese processing plants, while armed conflict makes exact production hard to establish. Environmental oversight is weak as well. Buyers may reduce direct exposure and still consume material that crossed the border through an opaque chain.
Western price floors and government-backed purchase agreements can keep new projects alive against lower Chinese domestic prices. The contracts buy resilience rather than instant competitiveness, making a higher-cost second market plausible for heavy rare earths and magnets accepted by defence customers. 2
USGS supplies the common mine output and reserve baseline, while IEA provides magnet-specific supply-chain shares. National geological agencies publish resources that often use different cut-offs and confidence levels. Each briefing keeps those categories apart rather than forcing them into a single number. 1 2
Project entries carry a dated status such as operating, construction, feasibility, pilot or early exploration. A feasibility study is useful evidence, though it is not a funded mine. A plant that has started up matters, but it is not the same as steady output accepted by customers.
Each country briefing ends with three dated predictions and the evidence that would force a revision. The deadlines turn broad claims into a record that can be checked.
Method and evidence
Every forecast has a deadline and a revision test
Project data are checked through 30 August 2026. Later national updates should be checked against the dated status in each briefing.
Editorial and commercial disclosure: Product listings and commercial relationships do not determine this report's findings, comparisons or forecasts. Sources, dates and revision rules are published so readers can audit the analysis. This is general information, not personal investment advice.
Production and reserves
USGS mine output and reserves are used for cross-country comparison. National resources are retained only with their original category and unit.
China share figures
IEA shares for mining, refining and magnets refer to the four principal magnet rare earths. They are not interchangeable with total USGS rare earth output.
Forecast protocol
Each prediction has a deadline, confidence score and an observable condition that would trigger a revision.
Sources used for this overview
- Mineral Commodity Summaries 2026: Rare Earths, version 1.3, U.S. Geological Survey, May 2026. Common mine output and reserve baseline.
- Rare Earth Elements: Pathways to Secure and Diversified Supply Chains, International Energy Agency, 8 April 2026. Magnet rare earth concentration, demand and diversification analysis.
- Global Critical Minerals Outlook 2026, International Energy Agency, 16 July 2026. Announced capacity and concentration outlook.
- Announcement No. 18 on export control of selected rare earth items, Ministry of Commerce of China, 4 April 2025. Original licensing scope for seven rare earth families and related products.
- Lynas Rare Earths FY2026 results presentation. Saleable NdPr, combined Dy/Tb, total products and first samarium output.
- Lynas expanded heavy rare-earth separation facility. 29 October 2025. Individual Sm, Gd, Tb, Dy, Y and Lu design capacity and product timing.
- MP Materials 2025 Form 10-K. Separated NdPr output, end of the Shenghe route and US downstream status.
- Energy Fuels White Mesa heavy circuit. 29 July 2026. Existing NdPr capacity, the heavy-circuit build and the larger 2029 expansion plan.
- Iluka rare-earth products and Eneabba refinery. Construction status, product suite and feed-dependent NdPr and Dy/Tb maximum capacities.
- Arafura Rare Earths Nolans project update. Funded mine-to-oxide design, NdPr output and first-product target.
- Caremag rare-earth recycling and refining plant. 650 tonnes a year of didymium oxide, 150 tonnes a year of neodymium oxide, 500 tonnes a year of dysprosium oxide and 90 tonnes a year of terbium oxide.
- Solvay La Rochelle rare-earth production expansion. Commercial magnet-material production and capability to separate the full rare-earth suite.
- Neo Silmet heavy separation line. 10 April 2026. Small-scale Dy and Tb precursor production in Estonia.
- US Commerce support for USA Rare Earth. Round Top recoverable-material list, funding and processing design. The release gives no usable start date.
- Rio Tinto and Canada Growth Fund scandium expansion. Operating Canadian route and funded capacity of nine tonnes a year.
- Serra Verde transaction presentation. Material-specific 2027 company estimates inside mixed carbonate, based on assumed recoveries. They are not separated-oxide output.
- Myanmar rebels disrupt China's rare-earth trade. Reuters, 28 March 2025. Heavy-feed routing, border control and disruption.
- Completion of Shenghe acquisition of Ngualla. Project control after the September 2025 transaction.
- Ngualla annual report and offtake route. Planned concentrate capacity and binding Shenghe purchase route.
- China and Rare Earth Supply Chains. Royal United Services Institute, June 2026. Individual elements, uses and supply-chain structure.
- IREL Rare Earths Division, Aluva. Indian individual oxide capability including La, Ce, Sm, Eu, Gd and Y; individual annual volumes are not disclosed.
- Sumitomo Metal Mining scandium recovery route. Philippines feed, Japanese refining and 7.5-tonne scandium-oxide-equivalent capacity.
- Australia's Sunrise project targets US critical-minerals supply. Reuters, 4 February 2026. Syerston scandium design and 2028 target.
- MP Materials second-quarter 2026 results. Combined Dy/Tb capacity of 200 tonnes a year and 2026 start-up testing status.
- REalloys second-quarter 2026 results. Canada's funded designs for 525 tonnes a year of NdPr metal, 30 tonnes of dysprosium oxide and 15 tonnes of terbium oxide; commercial intake target in the third quarter of 2027.
- USGS Mineral Commodity Summaries 2026, Antimony; IEA 2025 refining concentration (China's 74% refining share). The chart keeps each year and production step visible.
- USGS Mineral Commodity Summaries 2026, Bismuth. The chart keeps each year and production step visible.
- USGS Mineral Commodity Summaries 2026, Gallium; ERG Kazakhstan gallium project (Planned capacity of 15 tonnes a year and the planned start date); Australian Government, Wagerup gallium build decision (Final investment decision and planned capacity of 100 tonnes a year); Australian Government, Wagerup ground-breaking (Construction milestone on 24 August 2026; no production is credited). The chart keeps each year and production step visible.
- IEA 2025 refining concentration; USGS Mineral Commodity Summaries 2026, Germanium (Current market context and the absence of a complete country total); USGS OFR 2024-1057, table 1.14 (The latest complete country production table, for 2022). The chart keeps each year and production step visible.
- SCRREEN3 Hafnium factsheet; USGS Mineral Commodity Summaries 2026, Zirconium and Hafnium (USGS says current world primary hafnium production data are unavailable). The chart keeps each year and production step visible.
- USGS Mineral Commodity Summaries 2026, Indium. The chart keeps each year and production step visible.
- USGS Mineral Commodity Summaries 2026, Rhenium. The chart keeps each year and production step visible.
- USGS Mineral Commodity Summaries 2026, Tellurium. The chart keeps each year and production step visible.
Continue when you’re ready
Bring us your questions.
Talk through the metals, the buying process, storage or resale with EarthRarest.