Complete rare earth reference

Rare Earths: From Rock to Product

What the seventeen elements are actually used for, every step that turns a rock into a magnet, a catalyst, a laser or a phosphor, and who controls what along the way. Written for a reader with no chemistry or mining background.

A World Finance Edu research report · August 2026
How to read this document. Part A answers "what is this stuff for". Part B answers "how does a rock become that". They are written to be read in order, but each stage in Part B is self-contained, so you can jump to the step you care about. A glossary at the end of the report explains the recurring jargon and, just as importantly, where each mine and plant physically is. Part C answers "who controls it and why does it matter", including the Malaysia and Japan policy picture.
Part A

What rare earths are actually used for

The magnet story dominates the headlines, but it is roughly a third of the market. The other two thirds are refineries, glass, screens, lasers and ceramics, and they matter because they are what keeps the light elements economic.

1The seventeen elements, one by one

Rare earths are fifteen lanthanides plus scandium and yttrium. They are not scarce in the ground. Cerium is more common in the earth's crust than copper. What is hard is separating them from each other, because chemically they behave almost identically. Every element below occurs in the same ore body, so a producer cannot choose to make only the profitable ones. That single fact, called the balance problem, shapes the entire industry.

ElementClassWhat it is actually used forWhy that element, and supply note
Scandium
Sc, 21
SpecialAluminium-scandium alloys for aerospace frames, high-end bicycles and additive manufacturing. Electrolyte in solid oxide fuel cells. Metal halide stadium lamps.Adding under 0.5% scandium to aluminium gives large strength and weldability gains. Almost no primary mine exists; it comes as a by-product of titanium, uranium and nickel laterite processing, which is why it stays expensive.
Yttrium
Y, 39
HeavyYttria-stabilised zirconia: thermal barrier coatings on jet engine turbine blades, oxygen sensors in cars, dental crowns, cutting tools. Red phosphors. Host crystal for YAG lasers. YBCO high-temperature superconductors. Microwave filters in radar.Yttria stabilises zirconia in its tough tetragonal form and survives above 1,200°C, which is why jet engines can run hotter than their metal should allow. Grouped with the heavies commercially and comes largely from ionic clay.
Lanthanum
La, 57
LightFluid catalytic cracking catalyst in oil refineries, the single largest use by volume. Nickel metal hydride battery electrodes in hybrids. High refractive index optical glass for camera and phone lenses. Hydrogen storage alloys.In a refinery, lanthanum stabilises the zeolite so it keeps cracking heavy oil into petrol at 700°C. Abundant and cheap, and the world has more of it than it needs.
Cerium
Ce, 58
LightCerium oxide polishing powder for display glass, optical lenses and semiconductor wafers. Oxygen storage in petrol autocatalysts. Ultraviolet blocking glass. Diesel fuel-borne catalyst. Mischmetal and lighter flints. Cerium-doped YAG, the yellow phosphor that makes white LEDs white. Scintillator crystals.Cerium flips easily between two oxidation states, so it can absorb and release oxygen on demand. That is what an autocatalyst needs. The most abundant rare earth, and structurally in surplus.
Praseodymium
Pr, 59
LightBlended with neodymium as NdPr in permanent magnets. Yellow ceramic and glass pigment. Didymium glass in welding and glassblowing goggles. High-strength alloys in aircraft engines.Chemically so close to neodymium that separating the two adds cost for little benefit, so the industry ships and prices them together as NdPr. Didymium glass absorbs the sharp yellow sodium flare, which is why a glassblower can see the work.
Neodymium
Nd, 60
LightNdFeB permanent magnets, by far the dominant use. Nd:YAG industrial and medical lasers. Violet glass and ceramic colouring. Capacitor dielectrics.The compound Nd2Fe14B has an exceptionally strong preference for magnetising along one crystal axis, which is what makes the strongest commercial magnet possible. This is the element that carries the entire investment case.
Promethium
Pm, 61
SpecialNuclear batteries for spacecraft and pacemaker research, luminous paint, thickness gauges in manufacturing.Has no stable isotope and does not occur in usable quantity in nature; it is made in reactors. Commercially negligible, listed here only for completeness.
Samarium
Sm, 62
MidSamarium-cobalt permanent magnets for defence, aerospace, motorsport and downhole tools. Neutron-absorbing reactor control rods. Samarium-153 for bone cancer pain therapy.SmCo is weaker than NdFeB but keeps working past 300°C and does not corrode, so it goes where heat or reliability beats raw strength. Japan's 2026 supply agreements specifically added samarium, which tells you where defence demand is heading.
Europium
Eu, 63
HeavyRed and blue phosphors in displays, fluorescent lamps and LED lighting. Anti-counterfeiting markers in euro banknotes. Reactor control rods.Europium gives an extremely pure narrow-band red, which no cheap substitute matches. Demand fell sharply when compact fluorescent lamps died out, so it is one of the few rare earths in genuine oversupply.
Gadolinium
Gd, 64
HeavyContrast agents injected for MRI scans. Neutron shielding and control rods. X-ray intensifying screens. Magnetic refrigeration research.It has the highest neutron capture cross-section of any stable element and seven unpaired electrons, which is exactly what makes it visible on an MRI. A rare case where a rare earth ends up inside a patient.
Terbium
Tb, 65
HeavyDiffused into the surface of NdFeB magnets so they survive engine bay heat. Green phosphor in displays and lamps. Terfenol-D magnetostrictive alloy for sonar transducers and precision actuators.The most expensive of the magnet elements and the tightest supply of all. Terfenol-D physically changes length in a magnetic field, which is how a naval sonar makes sound.
Dysprosium
Dy, 66
HeavyThe heat-resistance additive in NdFeB magnets for electric vehicles and wind turbines. Nuclear control rods. Terfenol-D. Radiation dosimeters.Without dysprosium or terbium an NdFeB magnet starts losing its magnetism around 80°C, which is below the temperature inside a car motor. This is the element that export controls are really aimed at.
Holmium
Ho, 67
HeavyHolmium lasers used in urology to break kidney stones and treat the prostate. Magnetic flux concentrators in research magnets. Yellow and red glass colouring.Holmium has the highest magnetic moment of any element. Its laser wavelength is strongly absorbed by water, so it cuts tissue precisely without burning deep, which is why surgeons like it.
Erbium
Er, 68
HeavyErbium-doped fibre amplifiers, which regenerate the light signal every 80 to 100 km along every long-haul and submarine fibre optic cable in the world. Er:YAG dental and dermatology lasers. Pink glass and glaze colouring. Burnable neutron poison in reactor fuel.Erbium amplifies light at exactly 1,550 nm, the wavelength where glass fibre loses the least signal. The modern internet runs on this coincidence. Volumes are tiny and the strategic value is enormous.
Thulium
Tm, 69
HeavyPortable X-ray sources for field radiography using thulium-170. Tm:YAG surgical lasers. Blue security phosphor in euro banknotes.The rarest of the stable lanthanides. A thulium source lets a technician X-ray a weld on a pipeline with no electrical power supply.
Ytterbium
Yb, 70
HeavyYtterbium-doped fibre lasers for industrial cutting, welding and marking, a large and fast-growing use. Optical atomic clocks. Stainless steel grain refinement. Stress and pressure gauges.Fibre lasers have displaced CO2 lasers across sheet metal cutting, and almost all of them are ytterbium-doped. This is the quiet growth story that gets no press.
Lutetium
Lu, 71
HeavyLSO and LYSO scintillator crystals inside PET scanners. Petroleum cracking and alkylation catalysts. Lutetium-177 radiopharmaceuticals for prostate and neuroendocrine cancer.The densest and heaviest lanthanide, and the most expensive. Its scintillator stops gamma rays fast enough to give a PET scan usable resolution.
The vocabulary that trips people up. Light rare earths means lanthanum through samarium: abundant, cheap, mostly used in catalysts, glass and NdFeB magnets. Heavy rare earths means gadolinium through lutetium, plus yttrium: much scarcer, far more valuable, and concentrated in Chinese and Southeast Asian ionic clay. When a headline says China restricted rare earth exports, it almost always means the heavies, and usually dysprosium and terbium specifically. NdPr is the neodymium-praseodymium mix sold as one product. REO means rare earth oxide, the standard unit for quoting grade and production. Mischmetal is unseparated rare earth metal, mostly cerium and lanthanum, sold cheap.
The balance problem, in one paragraph. A rare earth deposit gives you every element at once, roughly in the proportions nature put there. A typical light-dominant ore is around half cerium and a quarter lanthanum by content, but almost all the revenue comes from the few percent that is neodymium and praseodymium. So a producer chasing magnet demand is forced to also produce large volumes of cerium and lanthanum it may not be able to sell. That is why the industry cares so much about finding new uses for cerium, and why a project's economics can be wrecked by a mineralogy that looks fine on a headline grade. When you assess any rare earth project, ask for the distribution, not the grade.

2Where the demand actually sits

Two different pictures are both true, and confusing them causes most of the bad analysis in this sector. By tonnage, catalysts, polishing and glass are enormous. By value and by strategic leverage, magnets dominate and everything else is a rounding error.

ApplicationShare of demandElements involvedCharacter of the demand
Permanent magnetsRoughly 31% to 41%, the largest single blockNd, Pr, Dy, Tb, SmGrowing fast with electric vehicles, wind, robotics and defence. Carries almost all the strategic risk.
CatalystsRoughly 16% to 20%La, CeRefinery cracking catalyst plus petrol autocatalysts. Large, stable, and slowly declining as internal combustion vehicles fade.
Polishing powdersRoughly 12%CeDisplay glass, optical lenses and semiconductor wafer polishing. Tracks electronics production.
Glass additivesRoughly 10%Ce, La, Nd, ErUltraviolet blocking, high refractive index lenses, decolourising, tinting. Mature.
Phosphors and pigmentsRoughly 8%Y, Eu, Tb, CeFell hard when fluorescent lamps died, now stabilised by LED and display use.
Nickel metal hydride batteriesRoughly 7%La, Ce, Nd, PrHybrid vehicles and industrial cells. Shrinking as lithium ion takes over, though hybrids have lasted longer than expected.
Metallurgy and alloysRemainderCe, La, Y, ScSteel and aluminium treatment, superalloys, thermal barrier coatings. Low value per kilogram, high volume.
Ceramics, lasers, medicalSmall but criticalY, Nd, Yb, Er, Gd, LuTiny tonnage, extremely high value per kilogram, and no substitute in most cases.

Shares are indicative and vary between sources depending on whether they measure tonnes or dollars, and whether they count oxide or magnet value. Treat the ranking as reliable and the decimals as noise.

The number that puts this sector in proportion. Rare earth oxides are roughly a US$8 billion market and rare earth magnets roughly US$16 billion, heading toward about US$30 billion by 2033. The entire global oxide market is smaller than one hyperscaler's quarterly data centre spend. This will never be a high-volume industry. It is a strategic one: few decisions, large plots, heavy utilities, and political attention out of all proportion to the revenue.

3Permanent magnets, the main event

If you understand one application, make it this one. Everything about the geopolitics, the export controls, the Japanese contracts and the Malaysian policy follows from how a neodymium magnet is built and why it fails.

How the magnet actually works

The working material is a compound called Nd2Fe14B. Its crystal has an extremely strong preference for pointing its magnetism along one particular axis. Manufacturing exploits this: the alloy is ground into powder so fine that each particle is essentially a single crystal, the powder is aligned in a magnetic field so all those axes point the same way, and then it is pressed and sintered solid. The result stores far more magnetic energy per unit volume than anything else available commercially, which is why an electric motor using it can be dramatically smaller and lighter than one that is not.

Two properties define the product. Remanence is how strong the magnet is. Coercivity is how hard it is to demagnetise. A magnet grade like N45SH encodes both: the number is the energy product, and the letters are the temperature class.

The problem is heat. Plain NdFeB starts irreversibly losing magnetism at around 80°C. An electric vehicle traction motor runs hotter than that. The fix is to add dysprosium or terbium, which raise the coercivity and push the usable temperature up toward 150°C to 200°C. That single engineering requirement is what turned two obscure heavy rare earths into instruments of trade policy.

ClassMax operating temperatureTypical Dy or Tb contentWhere it is used
NAbout 80°CNoneSpeakers, headphones, sensors, magnetic separators, consumer electronics
MAbout 100°CLowHard drives, small motors, appliances
HAbout 120°CModerateAir conditioning compressors, pumps, e-bike motors
SHAbout 150°CHigherIndustrial servo motors, robotics, generators
UHAbout 180°CHighElectric vehicle traction motors
EH / AH200°C and aboveHighestAerospace, defence, downhole tools, demanding traction applications

Temperature limits also depend on the shape of the magnet and how it sits in the magnetic circuit, so grade tables are a guide, not a guarantee.

The other magnet: samarium cobalt

SmCo magnets are roughly 20% to 30% weaker than the best NdFeB, and cobalt makes them expensive. They win on three things: they keep working past 300°C, they barely change strength with temperature, and they resist corrosion so well that they usually need no coating. That combination is why they end up in missile guidance, aircraft actuators, satellite mechanisms, motorsport, and oil and gas measurement tools that sit in a hot borehole. When a defence supply chain worries about rare earths, samarium is usually in the conversation alongside dysprosium.

How much magnet is actually in things
ProductNdFeB content, indicativeNotes
Electric vehicle traction motorAbout 1.5 kg, range 1 to 2 kgRoughly 7 g per kW of drive power. Of that, around a third is NdPr metal. Around 80% to 95% of electric vehicles use a rare earth magnet motor.
Hybrid vehicle1 to 2 kg magnetOlder hybrids also carried 10 to 15 kg of lanthanum-rich mischmetal in the nickel metal hydride battery, a large but declining use.
Conventional car, no electrification0.2 to 0.5 kgSpread across dozens of small motors: windows, seats, wipers, fuel pump, steering assist, speakers, sensors.
Direct drive offshore wind turbine160 to 650 kg per MWLow speed direct drive designs sit at the top of that range, mid speed around 160 kg per MW, geared high speed designs around 80 kg per MW. A 15 MW direct drive turbine can carry several tonnes.
Industrial robot1 to 5 kgSpread across six or more servo motors. Robotics and humanoid development is the demand story most often underestimated.
Inverter air conditioner compressor100 to 250 gVery high unit volumes in Asia. This is a genuinely large aggregate use.
E-bike or scooter hub motor300 to 500 gEnormous unit numbers in China and Southeast Asia.
Hard disk drive10 to 20 gVoice coil motor and spindle. Declining per unit but data centre drive shipments keep the total meaningful, and drives are the best feedstock for recycling.
SmartphoneUnder 1 gSpeaker, haptic vibration motor, camera autofocus and stabilisation, magnetic accessories. Tiny per unit, but billions of units.
Consumer drone50 to 300 gFour to eight brushless motors, all rare earth magnet based.
F-35 fighter aircraftAbout 417 kg of rare earth materialWidely cited United States Congressional estimate, covering magnets, alloys and coatings across the whole airframe.
Virginia class submarineAbout 4,200 kgSame source basis. An Arleigh Burke destroyer is usually put around 2,360 kg. These figures explain the defence procurement urgency better than any market forecast.

Defence figures are widely quoted United States Congressional Research Service and Government Accountability Office estimates, are dated, and cover total rare earth content rather than magnet mass alone. Use them directionally.

4Catalysts: refineries and exhaust pipes

This is the largest use nobody talks about, and it is the reason lanthanum and cerium have any market at all.

Fluid catalytic cracking

Every oil refinery has a fluid catalytic cracking unit that breaks heavy oil molecules into petrol and diesel. The catalyst is a zeolite, a crystalline aluminosilicate, and it works at roughly 700°C in a stream of steam and abrasive particles. Left alone the zeolite would collapse in that environment. Lanthanum ions sit inside the zeolite cage and hold the structure open, which keeps the catalyst active and improves how much petrol you get per barrel. A large refinery consumes tonnes of catalyst a day and the rare earth is not recovered, so this is genuinely consumed demand rather than a stock. Cerium is used alongside lanthanum in some formulations.

Automotive catalytic converters

A three-way catalyst has to oxidise carbon monoxide and hydrocarbons while simultaneously reducing nitrogen oxides, and those two jobs want opposite amounts of oxygen. Cerium oxide solves this by absorbing oxygen when the engine runs lean and releasing it when the engine runs rich, smoothing the swings so the platinum group metals can do their work. Without that oxygen buffer, modern emissions limits are not achievable. Demand here declines slowly as internal combustion vehicles are replaced, which is one of the few structurally negative lines in the rare earth demand picture.

There is a neat symmetry worth noticing. The catalyst business consumes the cheap, abundant light elements that the magnet business produces as unavoidable by-product. If catalyst demand falls faster than magnet demand rises, the balance problem gets worse, not better, and cerium and lanthanum stockpiles grow. Anyone building a separation plant needs a plan for that surplus.

5Polishing powders, glass and optics

Polishing. Cerium oxide polishes glass better than anything else known, and it does it through a mild chemical reaction with the glass surface rather than pure abrasion, which is why it leaves an optically perfect finish. It is used on smartphone cover glass, display panels, precision optics, and in chemical mechanical planarisation of semiconductor wafers. Every screen you have ever looked at was polished with cerium.

Optical glass. Lanthanum raises the refractive index of glass without raising dispersion, which is exactly what a lens designer wants. Compact camera modules in phones, high-end camera lenses and microscope optics depend on lanthanum glass. Cerium blocks ultraviolet light and prevents glass browning under radiation, so it goes into aerospace windows, medical glass and television glass. Neodymium and erbium are used as colourants, giving violet and pink respectively.

Specialist glass. Didymium, the neodymium and praseodymium mix, absorbs the intense yellow flare that sodium gives off when glass is worked at high temperature. That is why glassblowers' and welders' goggles have that characteristic colour: it is not tinting, it is a notch filter.

6Phosphors, lasers and light

A phosphor absorbs energy at one wavelength and re-emits it at another. Rare earths are unusually good at this because their inner electron shells are shielded from the surrounding crystal, so they emit very pure, very narrow colours that do not shift with the host material.

UseElementsWhat is going on
White LED lightingCe in a YAG hostA blue LED chip is coated with cerium-doped yttrium aluminium garnet, which converts part of the blue into yellow. Blue plus yellow reads as white to the eye. Almost every white LED in the world works this way.
Display and lamp phosphorsEu red and blue, Tb green, Y hostEuropium gives a narrow-band red that colour standards were effectively written around. This was a huge market in the fluorescent lamp era and shrank dramatically afterwards.
Medical imagingLu, Gd, Y, CeLSO and LYSO scintillator crystals in PET scanners convert gamma rays into light fast enough to locate the event. Gadolinium screens do the same job in X-ray. Gadolinium is also the injected MRI contrast agent.
Industrial lasersNd, Yb, Er, Ho, TmNd:YAG has been the workhorse solid state laser for decades. Ytterbium-doped fibre lasers have since taken over metal cutting and welding. Holmium and thulium lasers are surgical tools.
Fibre optic telecommunicationsErErbium-doped fibre amplifiers boost the optical signal directly, without converting it to electricity and back. Every submarine cable and long-haul route on earth depends on them. Tiny tonnage, total dependence.
Security and authenticationEu, Tm, YInvisible phosphors printed into banknotes, including euro notes, that fluoresce in a specific pattern under ultraviolet light.

7Alloys, batteries and ceramics

Metallurgy
  • Steel treatment. Small additions of cerium and lanthanum tie up sulphur and oxygen and control the shape of inclusions, which improves toughness. Low value per kilogram but very large volume.
  • Aluminium and magnesium alloys. Rare earth additions refine grain structure and improve creep resistance at temperature. Scandium is the premium version of this effect.
  • Superalloys and coatings. Yttrium improves how well a protective oxide layer sticks to a turbine blade, which directly extends engine life.
  • Mischmetal. Unseparated rare earth metal, used in lighter flints, as a steel additive, and historically in battery alloys.
Batteries and ceramics
  • Nickel metal hydride cells. The negative electrode is a lanthanum-rich alloy that stores hydrogen. This was the battery in every early hybrid and still ships in large numbers, though lithium ion has taken the growth.
  • Yttria-stabilised zirconia. Yttrium locks zirconia into a tough crystal form. It becomes the thermal barrier coating on jet engine blades, the sensing element in a car's oxygen sensor, dental crowns, and cutting tools.
  • Solid oxide fuel cells. Both yttrium and scandium stabilised zirconia are used as the oxygen-conducting electrolyte.
  • Hydrogen storage. Lanthanum-nickel alloys absorb and release hydrogen reversibly, a niche that grows if hydrogen infrastructure does.

8What can and cannot be substituted

Substitution is real but partial, and the pattern is consistent: you can usually design the rare earth out if you accept a worse product, more weight, more cost, or more complexity somewhere else.

ApplicationSubstitutable?What the alternative costs you
Electric vehicle traction motorPartlyInduction motors and externally excited synchronous motors use no magnets at all and are already in production at Tesla, Renault, BMW and others. They are typically heavier, less efficient at part load, or need brushes and more power electronics. Ferrite-assisted designs cut magnet content rather than eliminating it.
Dysprosium and terbium in magnetsReducible, not removableGrain boundary diffusion has already cut heavy rare earth content by well over half for the same temperature rating. Going to zero means accepting a lower operating temperature or a bigger magnet.
Wind turbine generatorYesGeared designs use far less magnet, but add a gearbox, which is historically the least reliable part of a turbine and the most expensive thing to replace offshore.
Refinery cracking catalystPartlyLow-rare-earth zeolite formulations exist and were adopted quickly during the 2011 price spike, at some cost in yield and catalyst life.
Cerium polishing powderPoorlyZirconia and alumina alternatives exist but give a worse finish or work slower. Cerium survived the 2011 price spike as the standard.
Erbium fibre amplifiersNoNo practical alternative for long-haul optical amplification at 1,550 nm.
PhosphorsPartlyQuantum dots substitute in some display applications. Colour purity and lifetime trade-offs remain.
MRI contrast, PET scintillatorsNoThe physics is specific to the element.
The lesson from 2011. When rare earth prices spiked roughly tenfold, the market responded within about two years: thrifting cut the amount used per unit, substitution removed rare earths from some applications entirely, and demand destruction was permanent in phosphors. Prices then collapsed and stayed low for a decade, wiping out most of the Western projects that the spike had funded. Any investment case built purely on a price spike should be read with that history in mind. The current case is different because it rests on policy and security of supply rather than price, but the thrifting reflex is still there.
Part B

From mine to product: every step

Ten stages. The first three are ordinary mining. The middle four are chemistry, and that is where China's control is close to absolute. The last three are precision manufacturing. Value, difficulty and political leverage all rise as you move down the list.

9The full chain at a glance

STAGE 0GeologyWhich deposit type you have decides everything downstream
STAGE 1MiningOre at 0.05% to 8% REO
STAGE 2BeneficiationConcentrate at 30% to 60% REO
STAGE 3CrackingMineral lattice broken open
STAGE 4Leach and purifyMixed rare earth carbonate
STAGE 5SeparationThe chokepoint. Single element streams
STAGE 6CalcinationIndividual oxides at 99.9%+
STAGE 7ReductionOxide becomes metal
STAGE 8AlloyingStrip cast NdFeB flake
STAGE 9Magnet makingPowder, align, sinter, coat
STAGE 10The productMotor, generator, actuator
StageWho can do itChina's shareWhy the difficulty rises
1 to 2
Mining and concentration
Many countriesAround 70% of mined oreStandard mining and mineral processing. Capital intensive but not technically exotic. Genuinely contestable.
3 to 4
Cracking and leaching
A handful of operatorsHighHot concentrated acid or caustic at scale, plus radioactive residue handling. The permitting problem starts here.
5 to 6
Separation to oxide
Very fewAround 85% to 90% of oxides, and up to 99% of heavy rare earthsHundreds to thousands of extraction stages, decades of operating know-how, and enormous reagent inventories. This is the real moat.
7 to 8
Metal and alloy
Very fewAround 90%Molten salt electrolysis at 1,000°C and above, with fluoride chemistry and severe corrosion problems. Power hungry.
9
Magnet manufacture
China and Japan, with new Western entrantsAround 90% to 94% of sintered NdFeBPrecision powder metallurgy with oxygen control, magnetic alignment and tight dimensional tolerance. Highest value step.

10Stage 0: the four kinds of deposit

Before any processing question can be answered you have to know what kind of rock you are dealing with, because the deposit type decides the flowsheet, the radioactivity problem, the element mix and the capital cost. There are four that matter commercially.

Deposit typeTypical gradeElement biasWhat it means in practice
Bastnasite in carbonatite
Mountain Pass, Bayan Obo, Mount Weld
3% to 8% REO, sometimes higherStrongly light: Ce, La, Nd, PrThe classic high-grade hard rock deposit. Relatively low thorium. Excellent for NdPr, gives you almost no dysprosium or terbium. Mount Weld feeds the Lynas plant in Malaysia.
Monazite and xenotime
Mineral sands by-product, India, Brazil, Australia
Variable, often recovered as by-productMonazite light, xenotime heavierMonazite carries thorium, typically several percent. That single fact creates the licensing, storage and public opposition problem that has defined the industry outside China. Often available as a by-product stream from titanium mineral sands operations, so the mining cost is already paid.
Ionic adsorption clay
Southern China, Myanmar, Malaysia, Laos
0.05% to 0.3% REO, very lowHeavy rich: Dy, Tb, YWeathered granite where rare earth ions are loosely stuck to clay particles. The grade looks terrible but the ions wash off with a simple salt solution at room temperature, so there is no crushing, no grinding and no acid bake. Low thorium. This is where the world's dysprosium and terbium actually come from, and it is the resource Malaysia holds.
Secondary and recycled
Magnet scrap, drives, phosphate gypsum, coal ash, tailings
Effectively very high for magnet scrapFollows the sourceEnd-of-life magnets are roughly 30% rare earth by weight, which beats any ore on earth. The constraint is collection and sorting, not chemistry.
Why Malaysia's ionic clay matters more than its grade suggests. A 0.1% REO clay deposit sounds worthless next to an 8% hard rock deposit. It is not, for three reasons. First, the enrichment is in the heavies, which is where the value and the scarcity are. Second, the extraction is a room temperature ion exchange rather than a high temperature acid bake, so the plant is far simpler and cheaper. Third, Malaysian clay is low in thorium, which sidesteps the radiological permitting problem that has dogged Lynas for fifteen years. The trade-offs are real: in-situ leaching consumes large volumes of ammonium salt and water and carries a serious groundwater and ammonia contamination risk if done badly, and southern China's history with this exact method is not a good advertisement.

11Stage 1: mining

1

Getting the material out of the ground

Three completely different methods depending on deposit type
InputRock, sand or weathered clay in the ground
OutputRun of mine ore at 0.05% to 8% rare earth oxide, or a pregnant leach solution
  • Open pit hard rock. Drill, blast, load, haul. Mountain Pass and Mount Weld work this way. Conventional, well understood, and the least of anyone's problems.
  • Mineral sands dredging. Rare earth minerals come out attached to a heavy mineral concentrate that is mined mainly for titanium and zirconium. Monazite is separated out later as a by-product, which is why the economics can work at grades that would never support a standalone mine.
  • In-situ leaching of ionic clay. No pit and no crushing. Injection wells are drilled into the weathered clay and an ammonium sulphate solution is pumped in. The ammonium ions swap places with the rare earth ions stuck on the clay surfaces, and a pregnant solution carrying the rare earths is collected downslope. Stage 2 and part of Stage 3 effectively disappear, which is why this route is so much cheaper.
Timeline to build3 to 10 years including permitting
Land footprintLarge. Hundreds of hectares
Key riskPermitting, water, community consent
SitingMine site only. Not a general industrial estate activity
The in-situ route is the one with the bad environmental history. Done properly it needs an impermeable base layer under the ore body, full recovery of the leach solution, and ammonia treatment. Done badly, and it was done badly for years in southern China, it acidifies soil, kills vegetation and puts ammonia nitrogen into groundwater. Any Malaysian project will be judged against that record whether or not it deserves to be.

12Stage 2: beneficiation

2

Throwing away the waste rock

Physical separation only. No chemistry yet
InputRun of mine ore, a few percent rare earth oxide at best
OutputMineral concentrate at roughly 30% to 60% rare earth oxide

The ore is crushed and then ground to a powder fine enough that individual mineral grains come free of each other, usually tens of micrometres. Then the rare earth minerals are separated from the gangue using differences in physical properties:

  • Froth flotation. The dominant method. Reagents are added that make the rare earth mineral surfaces water repellent, air is bubbled through the slurry, and the target minerals stick to the bubbles and float off as a froth. Getting the reagent chemistry right for a specific ore is a genuine technical exercise and is often where a new project stumbles.
  • Magnetic separation. Monazite and xenotime are weakly magnetic and can be pulled out with high intensity magnetic separators.
  • Gravity separation. Rare earth minerals are dense, so spirals, shaking tables and centrifugal concentrators work well, especially on sands.
  • Electrostatic separation. Used on mineral sands, exploiting differences in electrical conductivity between the mineral species.
EnergyHigh. Grinding is the single largest power draw in most mines
RecoveryTypically 60% to 85% of contained rare earth
WasteVery large tailings volume
Ionic claySkipped entirely. This is the clay route's big advantage
Everything up to this point is a commodity and is worth very little. A rare earth concentrate is the mineral industry equivalent of dirt with potential. This is precisely what Malaysia's export ban was written to stop leaving the country.

13Stage 3: cracking

3

Breaking the mineral open

The first chemical step, and where the radioactivity problem is created
InputMineral concentrate, 30% to 60% rare earth oxide
OutputA soluble rare earth compound plus a solid residue containing thorium and uranium

Rare earths are locked inside a crystal lattice, chemically bound as phosphates or fluorocarbonates. They will not dissolve in water. Cracking is the brute force step that destroys that lattice so the rare earths can be washed out. Three routes are used:

RouteConditionsHow it works and what it leaves behind
Sulphuric acid bakeConcentrated H2SO4, roughly 200°C to 500°C in a rotary kilnThe standard route for monazite and for bastnasite. Converts rare earths into water-soluble sulphates. Generates large volumes of acidic off-gas including sulphur dioxide and hydrogen fluoride, which needs a serious scrubbing plant. Thorium mostly reports to the residue.
Caustic crackConcentrated sodium hydroxide, roughly 140°C to 200°CConverts phosphate minerals to rare earth hydroxides and produces trisodium phosphate as a saleable by-product. Cleaner off-gas than the acid route, higher reagent cost, and it works best on high-grade concentrate.
Oxidative roastAir, roughly 600°CUsed on bastnasite to drive off carbon dioxide and oxidise cerium so it can be selectively rejected early. Often a precursor step rather than a complete crack.
This is where the thorium goes, and this is the whole Lynas controversy. Monazite naturally contains thorium, a mildly radioactive element, and often some uranium. Cracking concentrates it into a solid residue. A commonly cited figure is roughly 2.6 kg of thorium generated per tonne of rare earths produced, with the great majority reporting to solid waste that has to be stored under a radiation licence effectively forever. There is nothing exotic or dangerous about the chemistry; the problem is that no community wants the storage facility, and no regulator wants to sign the permanent disposal permit. That, far more than any technical barrier, is why Western separation capacity is so thin. It is also exactly why Malaysia's low-thorium ionic clay is politically attractive: the clay route largely avoids this step altogether.

Where the radioactivity actually comes from

The single most misread part of this industry. Worth getting exactly right.

The rare earths themselves are not radioactive. All seventeen are stable in the forms that occur in ore. The radioactivity is a passenger, and it is thorium.

  1. Thorium is a chemical lookalike for the light rare earths. Its ion is close enough in size and charge that when the mineral monazite crystallises, thorium simply occupies slots meant for cerium, lanthanum and neodymium. Monazite is effectively (Ce,La,Nd,Th)PO4. This is not contamination that got mixed in. It is built into the crystal.
  2. So the ore decides everything. See the table below. This is why a project's mineralogy matters more than its grade.
  3. Mining is not the trigger. Processing is. In the ground the thorium is dilute and locked in a lattice, and nobody cares. Stage 3 dissolves the rare earths out and the thorium reports to the solid residue. You have taken a harmless dilute natural material and concentrated it past the legal threshold for regulated radioactive material. Roughly 2.6 kg of thorium per tonne of rare earths produced, with 96% to 98% going to that solid waste. At Lynas the material has a name, WLP residue, water leach purification residue, and it is the specific physical object that fifteen years of Malaysian protest have been about.
  4. The hazard is the decay chain, not the thorium. Thorium-232 has a half-life of about 14 billion years, so it barely decays, so it is barely radioactive on its own. What matters is what it decays into: radium-228, thorium-228, and radon-220 gas. Radon gas and radium carried in dust are the real exposure pathways. That is why the argument is always about storage, ventilation and dust control rather than about the thorium sitting still.
Ore typeThorium contentWhat it means
MonaziteCommonly about 5% to 7% thorium oxide, up to around 10%The problem child. Usually recovered as a mineral sands by-product, so the ore is cheap but the residue liability is real.
XenotimeLower thorium, but carries uraniumDifferent element, same regulatory category.
BastnasiteWell under 1%Why Mountain Pass and Mount Weld are the comparatively easy ores.
Ion adsorption clayEssentially noneThorium is insoluble, so it never becomes a loosely held exchangeable ion in the first place. Weathering leaves it behind in the residual minerals. The same physics that puts the rare earths loosely on the clay is the physics that excludes the thorium. This, not luck, is why Malaysia's clay is politically attractive.

Why Malaysia reacts the way it does. This is the part usually missing from the coverage. Malaysia has done this before. Asian Rare Earth, a joint venture involving Mitsubishi Kasei, now Mitsubishi Chemical, processed monazite from tin tailings at Bukit Merah in Perak from 1982. After sustained litigation and protest it ceased operating in 1992 and became one of Asia's largest radioactive cleanup operations, reported at around USD 100 million. Malaysian opposition to Lynas is not abstract anxiety. It is a specific memory, and any project in this sector will be measured against it.

UtilitiesVery high heat load, large acid or caustic consumption
EffluentAcidic gas scrubbing, wastewater neutralisation
Permit classRadiological, if monazite is in the feed
Ionic clayNot needed. The ions are already free

14Stage 4: leaching and purification

4

Getting to a clean mixed solution

The product Malaysia banned exporting
InputCracked material, or the pregnant leach solution straight from an ionic clay operation
OutputMixed rare earth carbonate, all seventeen elements still jumbled together

The cracked solids are washed with water or dilute acid, dissolving the rare earths into solution. Then everything that is not a rare earth has to come out, because the separation circuit downstream is intolerant of impurities. Iron, aluminium, calcium, phosphate, thorium, uranium and radium are removed by controlled pH adjustment and selective precipitation: raise the pH in steps and different impurities drop out at different points.

The clean solution is then treated with sodium carbonate or oxalic acid and the rare earths drop out together as a mixed rare earth carbonate or oxalate, typically 92% to 96% rare earth oxide equivalent after drying. This is the standard traded intermediate product.

This is the exact product at the centre of Malaysian policy. Mixed carbonate is a commodity worth a fraction of what the separated oxides are worth, and shipping it means exporting almost all of the value. The export ban that took effect on 1 January 2024 was designed to make sure that anything mined in Malaysia moves at least past this point, and ideally past Stage 5, before it leaves the country. Understanding the difference between mixed carbonate and separated oxide is the single most useful thing a non-specialist can learn about this industry, because it is where almost all the confusion in press coverage comes from.
ChemistrypH controlled precipitation, well understood
DifficultyModerate. Not the bottleneck
WasteNeutralisation salts, radioactive residue if monazite fed
Value addedModest. Still a commodity

15Stage 5: separation, the chokepoint

5

Splitting seventeen nearly identical elements apart

The single step that explains China's position
InputMixed rare earth solution or redissolved carbonate
OutputIndividual element streams at 99.9% to 99.999% purity

Every rare earth carries the same +3 charge and differs from its neighbour only by a tiny change in ionic radius, a fraction of a percent. Nothing separates them cleanly in one pass. What the industry does instead is repeat a barely-effective separation an enormous number of times.

Countercurrent solvent extraction is the method. Two liquids that do not mix, an acidic water phase carrying the rare earths and an organic phase carrying a specialised extractant, are contacted in a mixer-settler: a stirred chamber that mixes them, followed by a quiet chamber where they separate by density. In each contact, one element moves into the organic phase very slightly more readily than its neighbour. The separation factor between adjacent elements is often only 1.5 to 3, and for the hardest pairs it approaches 1.

Hundreds of these mixer-settlers are plumbed in series, with the aqueous phase flowing one way and the organic phase the other, so the tiny per-stage advantage compounds. A full plant separating all the elements to high purity can run hundreds to well over a thousand stages, arranged as a tree of circuits that split the feed into two groups, then split each group again, and so on. Heavy rare earths need more stages, more reagent, and tighter control than the lights.

Element pairRelative difficultyWhy
Cerium from the restEasyCerium can be oxidised to the +4 state, which behaves completely differently. It is usually rejected first and cheaply.
Light group from heavy groupModerateEnough radius difference across the gap to give a workable separation factor.
Neodymium from praseodymiumHardAdjacent elements, very small separation factor. The industry mostly does not bother and sells NdPr as one product.
Dysprosium from terbium, and the heavies generallyVery hardSeparation factors close to 1, so stage counts and reagent inventories balloon. This is why heavy rare earth separation is around 99% Chinese.
Why this is a moat and not a secret. The chemistry is published. The extractants, P507, D2EHPA, Cyanex 272 and similar organophosphorus compounds, are commercially available. What is not transferable is thirty years of accumulated operating practice: how to hold an emulsion from forming, how to manage reagent degradation, how to keep trace impurities from poisoning the circuit, how to restart a plant without losing months of inventory. A separation circuit holds a very large working inventory of dissolved rare earth in its pipes, and a plant can take six months to a year to reach steady state and design purity after commissioning. That is a brutal cash flow profile for a new entrant and an enormous barrier to a fast Western build.

The Chinese head start. The chemist Xu Guangxian developed a countercurrent extraction theory in the 1970s that let engineers design a separation circuit mathematically rather than by trial and error. It let Chinese plants hit 99.99% purity at a fraction of the previous cost, and it was diffused across the industry as national policy rather than held as a company secret. Combined with cheap power, a tolerance for the environmental cost, and thirty years of continuous operation, that is how a single country ended up with roughly 90% of world separation capacity.

Can anything replace solvent extraction?
TechnologyMaturityStatus
Countercurrent solvent extractionCommercialThe global standard. Nothing displaces it today.
Ion exchange and chromatographyCommercial, nicheUsed for very high purity small volumes, historically for the heaviest elements.
Ligand-assisted displacement chromatographyPilotThe most credible challenger. Demonstrated high purity neodymium at far higher throughput than older chromatography, still pre-commercial.
Membrane processes and electrodialysisLab to early pilotCut solvent use dramatically. Not yet at scale.
Bioleaching and biosorptionEarly researchLower footprint biological route. Long way from a plant.
FootprintLong low buildings, very large. Mixer-settler halls run hundreds of metres
UtilitiesLarge water and chemical demand, moderate power, extensive bunding
EffluentAcidic and saline wastewater, organic losses, requires full treatment plant
Commissioning6 to 12 months to design purity

16Stage 6: precipitation and calcination

6

Turning a purified solution into a saleable oxide

Straightforward chemistry, high value product
InputSingle element solution at 99.9%+ purity
OutputA white or coloured powder: neodymium oxide, dysprosium oxide and so on

Oxalic acid or a carbonate is added and the single element drops out as an oxalate or carbonate. That solid is filtered, washed and then calcined, meaning heated in air to roughly 800°C to 1,000°C, which drives off the organics and leaves a pure metal oxide. Particle size and morphology are controlled here, and they matter to the downstream customer, because a magnet maker and a polishing powder maker want very different powders from the same chemical formula.

This is the point at which the material becomes a recognisable traded commodity with a published price. When you read a headline about neodymium oxide at US$113 per kilogram inside China and about US$184 per kilogram for export in March 2026, this is the product being quoted, and that roughly 63% spread is the entire commercial argument for building capacity outside China.

17Stage 7: reduction to metal

7

Oxide to metal

Power hungry, corrosive, and almost entirely Chinese
InputSeparated rare earth oxide
OutputRare earth metal, usually NdPr metal ingot for the magnet chain

An oxide is not a metal, and a magnet needs metal. Two routes:

  • Molten salt electrolysis for the light elements. Neodymium oxide is dissolved in a molten bath of neodymium fluoride and lithium fluoride at roughly 1,000°C to 1,100°C and a current is passed through it. Metal collects at the cathode. It is essentially the same idea as aluminium smelting, and it has the same characteristics: very high electricity consumption, on the order of ten kilowatt hours or more per kilogram of metal, severe corrosion of everything the bath touches, and fluoride emissions that must be captured. Cheap reliable power is a genuine siting requirement.
  • Metallothermic reduction for elements whose melting points are too high for electrolysis, notably dysprosium, terbium, samarium and yttrium. The oxide or fluoride is reduced with calcium or lithium metal under vacuum or inert gas. Batch, expensive, and lower volume.
PowerVery high. Around 10 kWh or more per kg of metal
EmissionsFluoride and particulate, requires capture
ConcentrationAround 90% Chinese
Siting driverReliable low cost baseload electricity
This stage is the least discussed and one of the most binding. Several Western projects can now produce separated oxide but still send it to China to be turned into metal, which means the supply chain is not actually de-risked. Any credible claim to a non-Chinese magnet chain has to include this step.

18Stage 8: alloying

8

Making the magnet alloy

Strip casting, a small step with a large effect on quality
InputNdPr metal, iron, boron, and dysprosium or terbium if used in the melt
OutputThin alloy flake, roughly 200 to 400 micrometres thick

The metals are melted together under vacuum or argon in the right proportions and then strip cast: the molten alloy is poured onto a water-cooled rotating copper wheel, where it solidifies almost instantly into thin flakes. The speed of cooling is the point. Cool it slowly in a mould and you get coarse grains and patches of free iron, both of which ruin magnetic performance. Strip casting gives a fine, uniform grain structure with the neodymium-rich grain boundary phase in exactly the right place, and that grain boundary phase is what allows the magnet to be sintered properly later.

Strip casting replaced simple book mould casting across the industry in the 1990s and is one of the reasons magnet performance improved so much. It is a good example of a step that looks trivial on a flow diagram and is actually decisive for product quality.

19Stage 9: making the magnet

9

Precision powder metallurgy

Nine sub-steps, and the highest value point in the chain
InputStrip cast NdFeB alloy flake
OutputFinished, coated, magnetised, dimensionally precise magnet blocks or arcs
#Sub-stepWhat happens, and why it matters
1Hydrogen decrepitationThe flake is exposed to hydrogen gas at roughly 0.1 to 0.25 MPa for one to three hours. Hydrogen enters the crystal lattice, expands it, and shatters the alloy into coarse friable powder without any mechanical force. The hydrogen is then driven off by heating to about 300°C to 400°C. Elegant, and much gentler on the material than crushing.
2Jet millingThe coarse powder is fired against itself in a high velocity nitrogen stream until the particles are ground down to a mean size of about 2 to 5 micrometres, which is roughly one crystal grain each. Everything from here on is done under nitrogen or vacuum, because powder this fine will oxidise, and oxygen is the enemy of magnetic performance. It is also pyrophoric, so this is a real fire and explosion hazard step.
3Alignment and pressingThe powder is placed in a die and a strong magnetic field is applied, which rotates every particle so its easy magnetisation axis points the same way. While the field holds them aligned, the powder is pressed into a compact. This is the step that turns a pile of powder into an anisotropic magnet, and alignment quality directly sets the final strength. Isostatic pressing is often used afterwards to raise density evenly.
4SinteringThe compact is heated in a vacuum furnace to roughly 1,000°C to 1,100°C. The neodymium-rich grain boundary phase melts and pulls the grains together, densifying the part to nearly full theoretical density. The part shrinks substantially and not evenly, which is why the next step is unavoidable.
5Two stage annealingHeat treatment at roughly 850°C to 950°C and again at 400°C to 600°C. This optimises the structure of the grain boundary phase, which is what actually sets coercivity. Get this wrong and an otherwise perfect magnet demagnetises in service.
6MachiningSintered NdFeB is hard and brittle and cannot be turned or milled conventionally. It is cut with diamond wheels and wire saws and ground to final tolerance, often within a few hundredths of a millimetre. Material loss at this step is significant, commonly a fifth to a third of the block, and that swarf is a valuable recycling stream.
7Grain boundary diffusionThe clever modern step. Instead of mixing expensive dysprosium or terbium through the whole magnet, a heavy rare earth compound is painted or sputtered onto the machined surface and the part is heated so the heavy element diffuses along the grain boundaries into the interior. Coercivity is set at the grain boundaries, so this achieves the same temperature rating using a fraction of the heavy rare earth. This single technique has cut dysprosium and terbium intensity dramatically over the last decade and is a large part of why heavy rare earth demand has not grown as fast as electric vehicle sales.
8CoatingNdFeB corrodes readily, so magnets are nickel-copper-nickel plated, zinc plated, epoxy coated or given a combination. Coating failure is one of the most common field failure modes.
9Magnetising and testThe part is finally exposed to a very strong pulsed field to saturate it, then measured for flux, dimension and coating integrity. Many magnets are shipped unmagnetised and magnetised by the customer after assembly, simply because a magnetised block is dangerous and difficult to handle.
EnvironmentInert atmosphere throughout, tight oxygen control
HazardsFine pyrophoric powder, strong fields, diamond machining dust
UtilitiesHigh power, vacuum furnaces, nitrogen plant, clean water
Building typeClean, climate controlled, high specification manufacturing. Suited to a general industrial estate
This is the stage that can actually sit in a general industrial estate. Stages 1 to 4 are mine site activities. Stages 5 to 7 are heavy chemical plants with radiological and effluent baggage that most parks cannot and should not host. Stage 8 and Stage 9 are ordinary advanced manufacturing: a clean, well-serviced building with reliable power, nitrogen supply, good logistics and a trained workforce. That is precisely what a general industrial estate is built to accommodate. The magnet plant announced for Kuantan is exactly this profile.

20Stage 10: into the product

10

Assembly into the thing that gets sold

Where the magnet stops being a material and becomes a component

Finished magnets are bonded or clipped into a rotor, encapsulated, balanced and assembled into a motor, generator, actuator, speaker or sensor. For an electric vehicle traction motor this means rotor lamination stacking, magnet insertion, adhesive cure, balancing and end-of-line test. The magnet is typically 10% to 20% of the motor's cost but effectively 100% of its supply risk, which is why automakers have spent the last five years trying to either secure magnet supply or design around it.

Worth noting on siting: motor and rotor assembly is a much larger employer and occupies far more floor space than magnet making itself, and it has none of the chemical baggage. If a magnet plant lands somewhere, the assembly work often follows, and that is where the employment sits.

21The value ladder

The same kilogram of rare earth, tracked from concentrate to finished magnet. Prices in this sector are volatile and differ enormously between Chinese domestic and export markets, so treat the ranges as indicative and the shape of the ladder as the real point.

ProductIndicative valueWhat has been added
Run of mine oreEffectively no marketNothing. Not a traded product.
Mineral concentrate, 30% to 60% REOLow, single digit US$ per kg of contained oxideCrushing, grinding, flotation. A commodity.
Mixed rare earth carbonateLow, roughly US$10 per kg order of magnitudeCracking and leaching. Still a commodity, and the product Malaysia banned exporting.
Separated NdPr oxideNeodymium oxide around US$113 per kg in China and about US$184 per kg for export, March 2026Separation. This is where roughly an order of magnitude of value appears, and it is the reason the entire industry structure looks the way it does.
Separated dysprosium oxideSeveral hundred US$ per kgMuch harder separation, far scarcer feed.
Separated terbium oxideRoughly an order of magnitude above dysprosiumThe tightest and most expensive of the magnet elements.
NdPr metalOxide price plus a conversion marginElectrolysis. Modest value added, high power cost, and a genuine bottleneck outside China.
Strip cast NdFeB alloyMetal price plus alloying marginMelting and rapid solidification.
Finished sintered magnetHighly variable, driven by grade and heavy rare earth contentPowder metallurgy, alignment, sintering, machining, diffusion, coating. Highest value and highest barrier step.
Assembled motor or generatorMagnet is 10% to 20% of costComponent assembly. Large employer, low technical barrier, follows the magnet plant.
Read the ladder as a policy document. The gap between mixed carbonate and separated oxide is where the money is, and every producing government now knows it. Malaysia's export ban, Indonesia's downstreaming policy, the United States Department of Defense price floor under NdPr, and the European Union's Critical Raw Materials Act are all versions of the same instruction: do not let the material leave before Stage 5.

22Utilities, waste and footprint by stage

The practical table for anyone assessing whether a given stage can physically be hosted on a given site.

StagePowerWaterEffluent and wasteRadiologicalSite implications
1 MiningModerateHighTailings, in-situ leach solutionDepends on oreMine site only. Not a park activity.
2 BeneficiationHigh, grindingHighVery large tailings volume, flotation reagentsDepends on oreAt the mine. Needs a tailings facility.
3 CrackingHigh heat loadModerateAcid gas, sulphur dioxide, hydrogen fluoride, solid residueYes if monaziteDedicated heavy chemical site, full scrubbing, licensed residue storage.
4 LeachingModerateHighAcidic and saline wastewater, neutralisation sludgePossibleBunded, corrosion resistant, on-site effluent treatment.
5 SeparationModerateHighSaline and acidic wastewater, organic solvent lossesLow by this pointVery large flat footprint. Long mixer-settler halls, extensive bunding, solvent fire risk, dedicated treatment plant.
6 CalcinationHigh heat loadModerateCombustion off-gas, particulateLowStandard chemical plant.
7 Metal makingVery highModerateFluoride emissions, spent salt, slagLowNeeds firm cheap baseload power and fluoride capture. Effectively a smelter.
8 AlloyingHighModerate, coolingMinimalNoneVacuum melting shop. Conventional heavy manufacturing.
9 Magnet makingHighModerate, plating and coolingPlating effluent, grinding swarf, both recoverableNoneClean advanced manufacturing. Nitrogen supply, vacuum furnaces, climate control, fire protection for pyrophoric powder. Fits a normal high specification industrial site.
10 AssemblyModerateLowMinimalNoneOrdinary light to medium manufacturing. The largest employer of any stage.
A commonly cited Chinese industry estimate for the traditional light rare earth route is that producing one tonne of rare earth oxide generates on the order of 2,000 tonnes of tailings, tens of thousands of cubic metres of waste gas and a couple of hundred cubic metres of acidic wastewater. Modern plants perform far better than that, and the figure describes historic Chinese practice rather than a benchmark, but it is the number that shapes public perception and it will be quoted at any project that faces a community consultation.

23Recycling: the parallel chain

Recycling is a genuine second source, not a green talking point, but it is limited by collection rather than by chemistry. Global end-of-life recycling of rare earths is still in the low single digit percentages.

Direct reuse

Pull the magnet out of a hard drive or a motor, clean it, re-magnetise it, and put it back to work. Cheapest by far, and limited to cases where the original geometry suits a new application. Mostly a niche.

Hydrogen processing

The same hydrogen decrepitation trick used in Stage 9, applied to scrap. Hydrogen shatters the magnet into powder inside its own casing, which separates it from steel and glue. The powder can be re-pressed into a new magnet, skipping Stages 1 through 8 entirely. This is the most promising route and is being commercialised in the United Kingdom and elsewhere.

Chemical recycling

Dissolve the scrap and feed the solution back into a separation circuit at Stage 4 or 5. Handles mixed and contaminated feed, tolerates any magnet grade, and produces virgin quality oxide. Most flexible, most expensive, and it needs a separation plant to feed into.

Where the feedstock actually is. Manufacturing scrap comes first: machining swarf from Stage 9 alone is a fifth to a third of every block cut, it is clean, it is concentrated, and it never leaves the factory. That is why magnet plants and recyclers want to sit next to each other. End-of-life feedstock is harder: hard drives are the best source because data centres decommission them in bulk, while electric vehicle motors will become the dominant stream in the 2030s but are not there yet because the first large fleets have not been scrapped. Japan has run a national urban mining programme for years and treats it as part of supply security rather than as waste management. A recycler is a small, clean, light-industrial operation and a natural neighbour to a magnet plant.

24Where each stage can be sited

Reading the chain backwards from the siting question rather than forwards from the geology.

StageCan a general industrial estate host it?Assessment
1 to 2 Mining, beneficiationNoMine site activity, tied to the ore body.
3 to 4 Cracking, leachingOnly a dedicated siteRadiological licensing if monazite is in the feed, heavy acid handling, licensed residue storage. This needs a purpose-built chemical estate with the right permits, not a general industrial park, and it carries reputational risk that attaches to the host site as much as to the operator.
5 to 6 Separation, calcinationConditionallyVery large flat land, heavy water and effluent infrastructure, solvent fire risk. Possible only on a dedicated chemical-zoned estate. Low employment per hectare. High political visibility.
7 Metal makingConditionallyEffectively a smelter. Needs firm cheap power above all else. Power price and reliability decide the location more than any other factor.
8 to 9 Alloying and magnet makingYes, and this is the entry pointThe realistic entry point. Clean advanced manufacturing, high specification building, reliable power, nitrogen, good logistics, skilled workforce. No radiological issue. Strong Japanese and Korean corporate interest, and a live precedent already in Kuantan.
10 Assembly, and recyclingYesOrdinary manufacturing, largest employment per hectare, follows wherever the magnet plant lands. Recyclers are small and clean, and want to sit next to the magnet plant for its swarf.
Three things that shape any siting decision. First, reputational and permitting risk is the defining feature of this sector in Malaysia and it attaches to the host site as much as to the operator. Second, effluent liability, residue handling, decommissioning and site restoration decide whether a project is viable at a given location, and where those obligations sit is settled at the contracting stage rather than later. Third, this is a policy-driven market: the January 2024 export ban created the onshore processing opportunity, and the August 2026 signal that it may be eased shows how fast the underlying logic can move.
Part C

The industry, the market and the politics

Merged in from the standalone global rare earth report of July 2026. Who controls what, how China built the position, what Japan did about it, and where the money is going. Findings carry the confidence rating and vote count from that report's adversarial verification pass.

25China's position, stage by stage

China's grip is best understood as a mining share that is large but contestable, sitting on top of a processing share that is close to absolute. Confusing the two produces most of the bad commentary in this sector.

StageChina's shareSource basis
Mine production, oreAbout 62% in 2019, rising to about 70% by 2024European Commission Joint Research Centre and United States Geological Survey
Rare earth oxidesAbout 85%Joint Research Centre, citing Adamas Intelligence
Rare earth metalsAbout 90%Joint Research Centre
All rare earth processingAbout 90%Center for Strategic and International Studies
Heavy rare earth processingAbout 99%Center for Strategic and International Studies
Sintered NdFeB magnetsAbout 94%International Energy Agency
Refining leadership across strategic mineralsLeads 19 of 20 minerals, about 70% on averageInternational Energy Agency, Global Critical Minerals Outlook 2025
High confidenceverified 3-0 Global ore production is not monopolised. The United States, Myanmar and Australia each hold roughly 10% to 12% of mining. The dominance that matters is downstream.
How it was built, three decades of policy
  • Export quotas, 2010. China imposed strict rare earth export quotas, spiking global prices and alarming importers. The quotas were lifted in 2014 after a World Trade Organization ruling (case DS431) found them non-compliant.
  • Consolidation, from 2016. Mining and separation companies were merged into state-owned groups, first six, then rationalised further. China Rare Earth Group was formed in December 2021. The result is essentially two quota-eligible groups, China Northern Rare Earth and China Rare Earth Group, controlling substantially all concentrate and refining quota.
  • Downstream pull. Industrial policy drew alloy and magnet manufacturing into China as well, so the highest value step concentrated there too.

Outside China only a handful of processors operate at scale, chiefly Lynas in Australia and its Advanced Materials Plant in Malaysia.

High confidenceverified 3-0 The policy sequence above, and the fact that Lynas is the only at-scale non-Chinese processor.

26Producing countries and anchor companies

CountryRoleNotes
ChinaAbout 70% of mining, about 90% of processingBayan Obo in Inner Mongolia is the world's largest deposit. Ionic clay mines in the south supply the heavy rare earths.
United StatesAbout 12% of miningMountain Pass, operated by MP Materials, is the only at-scale mine-and-process site in North America.
MyanmarAbout 10% of miningA critical source of heavy rare earths feeding Chinese separation plants. Supply is politically unstable.
AustraliaAbout 10% of miningHome to Lynas, the largest non-Chinese processor.
China, state-owned

China Northern Rare Earth Group

  • Operates the Bayan Obo deposit
  • Produces on the order of 270,000 tonnes of rare earth oxide a year, a large share of global output
  • One of the two quota-eligible state groups
China, state-owned

China Rare Earth Group

  • Formed December 2021 by merging several state enterprises
  • Consolidates southern and heavy rare earth capacity
  • The second of the two dominant quota holders
United States, listed (NYSE: MP)

MP Materials

  • Largest rare earth producer in the Western Hemisphere
  • Owns and operates Mountain Pass, California
  • Supplies more than 10% of global rare earth content
  • Building United States metal and magnet capacity, see section 30
Australia and Malaysia, listed

Lynas Rare Earths

  • Largest processor outside China
  • Mines in Australia, separates at the Lynas Advanced Materials Plant in Malaysia
  • Expanding into processing in Texas
High confidenceverified 3-0 MP Materials is the largest producer in the Western Hemisphere, owns the only at-scale North American mine-and-process site, and supplies more than 10% of global rare earth content.

27Asia-Pacific: Japan, Malaysia, Indonesia, Singapore

The response to China's dominance has an Asian centre of gravity. Japan wrote the playbook, Malaysia hosts the hardware, Indonesia is a late entrant with more potential than delivery, and Singapore's role is money rather than material.

Japan: first mover on supply security, and still the magnet technology leader

Japan is the country that first treated rare earths as a national security question, and its response over fifteen years is the template the West is now copying.

September 2010
The supply shock. After a trawler collision near the disputed Senkaku Islands, China halted rare earth shipments to Japan for roughly two months, though Beijing denied a formal ban. Japan was then about 90% dependent on China, and prices spiked roughly tenfold over the following year.
October 2010
Japan passed an unprecedented supplementary budget of about JPY 100 billion, roughly USD 1.2 billion, to fund diversification, stockpiling, recycling and substitution.
March 2011
The Lynas lifeline. JOGMEC, the government resources agency, and trading house Sojitz committed USD 250 million in loan and equity to Lynas, in return for about 8,500 tonnes a year of rare earths for ten years, roughly 30% of the Japanese market. This financing is a key reason a large non-Chinese processor exists at all.
2022 to 2023
Further JOGMEC and Sojitz top-ups, including a 2023 investment aimed specifically at securing heavy rare earths, dysprosium and terbium.
Early 2020s
Result. Japan cut its reliance on Chinese rare earths from about 90% in 2010 to roughly 58%, through overseas investment, substitution and recycling.
2026
The revised Lynas agreement reported in 2026 raises Japan's access to 75% of medium and heavy rare earth output, adds samarium and other elements, and covers at least 5,000 tonnes a year of NdPr through 2038. That material is separated in Kuantan, so Malaysia is a physical node in Japan's supply security.
Magnet technology

Japan invented the modern magnet

  • The sintered NdFeB magnet was invented by Masato Sagawa at Sumitomo Special Metals, with the breakthrough around 1982 to 1984, in parallel with General Motors. Sumitomo Special Metals became Hitachi Metals in 2007 and is now Proterial.
  • Shin-Etsu Chemical, TDK and Proterial together make roughly 15% to 20% of global sintered NdFeB magnets, the leading share outside China.
  • Substitution leadership: Daido Steel has supplied heavy-rare-earth-free magnets used in Honda hybrids since 2016, and Proterial has cut its heavy rare earth use to about a tenth of its early 2000s level.
New frontiers

Deep sea mud and urban mining

  • Minamitorishima. Rare-earth-rich seabed mud inside Japan's exclusive economic zone, estimated at about 1.2 million tonnes of oxide in the most promising zone, with an unusually high heavy-to-light ratio of roughly 50:50 against about 25:75 for typical Chinese ore, and low radioactivity.
  • Status: a pilot lift test from about 6,000 metres depth was planned to begin January 2026 at roughly 350 tonnes a day, with a demonstration phase targeted for 2027. Unproven at scale.
  • Recycling. Shin-Etsu has recycled magnets industrially since 2008. A Daikin, Shin-Etsu and Hitachi consortium will recover magnets from used air conditioner compressors, around 10,000 a year, from 2027.
Malaysia: the largest processing plant outside China
  • What LAMP is. Operating since 2012 on a 100 hectare site at Gebeng near Kuantan, Pahang, the Lynas Advanced Materials Plant does not mine. It processes concentrate shipped from Mount Weld in Western Australia through cracking and leaching, solvent extraction and finishing, mainly into NdPr oxide. In 2025 it also began producing separated heavy rare earths, dysprosium and terbium, at Kuantan. That was previously an almost exclusively Chinese capability.
  • The radioactive residue fight. Cracking and leaching produces a low-level radioactive residue containing thorium, which drove years of protest. A 2019 and 2020 licence condition ordered Lynas to move that step out of Malaysia, with a deadline of 1 July 2023 and then 1 January 2024. In October 2023 Malaysia reversed course after Lynas proposed thorium extraction technology, allowing continued import and cracking through March 2026. Lynas is separately building cracking and leaching capacity at Kalgoorlie in Australia.
  • Domestic ambitions. Malaysia holds its own non-radioactive ionic clay deposits, a government estimate of about 16.1 million tonnes, and in September 2023 announced a ban on exporting raw or unprocessed rare earths to force domestic value capture. In 2026 Lynas signed a magnet plant joint venture with South Korea's JS Link near the Gebeng refinery, a step toward finished magnets in Malaysia.
Indonesia: ambitious late entrant, still mostly potential
  • Resource. Indonesia's rare earths are mainly monazite recovered as a by-product of tin mining in Bangka-Belitung, through state miner PT Timah, with further potential in Kalimantan, Sulawesi and Papua. Widely quoted figures in the billions of tonnes are government potential estimates, not proven reserves. The United States Geological Survey still lists Indonesia's rare earth reserves as not available.
  • Policy. Indonesia is applying its nickel downstreaming playbook, raw export bans plus mandated domestic processing, to rare earths, aiming to build separation capacity rather than export concentrate. PT Timah is running monazite pilot projects and negotiating with foreign technology suppliers.
  • Reality check. Indonesia is essentially early stage. As of 2026 it has no commercial separation or refining capacity, which leaves any concentrate at risk of flowing to Chinese processors. Potential is large, delivery is unproven.
Singapore: a finance and trade hub, not a producer

Singapore has no mines, no refining and no operating magnet production. Its relevance is real but almost entirely as a finance, trade and policy node. An honest read by angle:

AngleVerdictDetail
Trading and finance hubReal but genericA genuine commodity trading and mining finance centre. Traders have held rare earth oxide in Singapore bonded warehouses as a hedge against Chinese supply cuts. But the large books here are battery metals, lithium, cobalt and copper, not rare earths, and no rare-earth-dedicated desk is identifiable.
Recycling and urban mining researchAdjacentNTU's SCARCE programme, with France's CEA, does advanced separation from electronic waste and batteries, but there is no evidence of a dedicated NdFeB magnet or rare earth separation programme.
Manufacturing that uses rare earthsLegacySingapore was the world's hard disk drive hub, using Nd, Pr and Dy magnets, into the 1990s, but Seagate closed its last local drive plant in 2009. Today's semiconductor and precision base is a rare earth consumer, not a maker.
Policy and neutral hubStrongest angleSingapore positions itself as a neutral logistics, finance and trade hub amid United States and China rivalry, pursuing supply chain diversification and upstream investment in resource-rich neighbours such as Indonesia and Australia.
Rare earth corporate headquarters or joint ventureNot foundNo rare earth company headquarters, joint venture or deal is anchored in Singapore. The nearest regional activity, Lynas processing and the JS Link magnet joint venture, sits in Malaysia.
Singapore takeaway. For a Singapore-based investor or adviser the opportunity is not production but the money and logistics layer: financing, trading, warehousing, certification, and structuring deals into resource-rich neighbours. Any claim that Singapore is a rare earth producer, refiner or magnet maker is unsupported.

28Demand outlook and price dynamics

Evidence note. Quantified demand and price forecasts did not survive the source report's adversarial verification pass. The framing below reflects broad industry consensus and is medium confidence. Treat specific numbers as directional context, not verified data points.

Demand drivers. Magnet-grade demand, meaning neodymium, praseodymium, dysprosium and terbium, is anchored to three long-run trends: electric vehicles, wind power and defence. Industry analysts including the International Energy Agency, IDTechEx and Adamas Intelligence broadly describe magnet material demand as having roughly doubled since 2015 and continuing to grow through 2030 and beyond, with electric vehicles the dominant single driver. Because each vehicle and each turbine embeds a fixed magnet content, demand scales directly with electrification and installed wind capacity.

Price dynamics. NdPr is the benchmark price. Prices are notoriously volatile because supply is concentrated and policy sensitive: quota changes, export controls and stockpiling decisions in China move the market more than incremental demand does. Heavy rare earths carry a scarcity premium over the lights because their supply is even more concentrated. The 2025 export control cycle drove renewed price strength in medium and heavy rare earths.

Why a United States price floor exists. Volatility is severe enough that the Department of Defense agreed a ten year price floor of USD 110 per kg on NdPr for MP Materials. That is a signal that the West views price risk, not only supply risk, as the barrier to building domestic capacity. High confidenceverified 3-0

29Geopolitics and export controls

Since mid-2023 China has used its processing dominance as strategic leverage, through a widening sequence of export controls on critical minerals.

July 2023
Gallium and germanium export controls.
October 2023
Graphite export controls.
August 2024
Antimony and superhard materials controls.
February 2025
Tungsten and tellurium controls.
April 2025
Medium to heavy rare earths, gadolinium, dysprosium and terbium, brought under licensing. This hit magnet supply directly.
October 2025
Expanded rare earth and critical minerals measures announced, subsequently reported to be delayed by about a year.
The scale of the leverage. The International Energy Agency estimates that if China's expanded October 2025 controls are fully enacted, around USD 6.5 trillion in annual downstream production outside China could be jeopardised, with automotive alone accounting for over USD 3 trillion of that exposure. The figure spans China's broader critical minerals controls, rare earths plus graphite and others, but rare earths are the dominant driver. High confidenceverified 3-0

This is the essence of the chokepoint dynamic. A country does not need to control the ore if it controls the one processing step through which nearly all supply must pass. Western governments have responded with stockpiling, friend-shoring, price floors and direct equity stakes, a sharp break from prior reliance on open markets.

30Investment landscape and outlook

The thesis is straightforward. Demand is anchored to electric vehicles, wind and defence, while supply security is the dominant risk. Value therefore accrues to whoever can build non-Chinese processing and magnet capacity, which is the scarce and defensible step. The flagship case is MP Materials.

The MP Materials and United States Department of Defense deal, July 2025. A rare government price-floor-plus-offtake structure de-risking domestic magnet production:
  • USD 400 million equity investment in newly issued Series A convertible preferred stock
  • USD 150 million loan for heavy rare earth separation
  • Ten year price floor of USD 110 per kg for NdPr products, from the fourth quarter of 2025
  • Ten year Department of Defense offtake of 100% of magnets from the planned "10X" facility
  • About 15% Department of Defense ownership on an as-converted basis
High confidenceevery term verified 3-0 against SEC filings
MP's operational ramp and strategic pivot
  • Record third quarter 2025 NdPr production of 721 tonnes, up 51% year on year, and second-best-ever quarterly rare earth oxide production of 13,254 tonnes.
  • A new Mountain Pass heavy rare earth separation facility began commissioning in mid-2026, roughly 3,000 tonnes a year of feedstock and a 200 tonne a year dysprosium and terbium circuit, supporting planned NdFeB magnet output of about 10,000 tonnes a year.
  • Decoupling from China. MP ceased concentrate shipments to China in April 2025 and agreed to cease all product sales to China under the Department of Defense arrangements, recognising zero concentrate revenue in the third quarter of 2025 against USD 43.1 million a year earlier. That is a deliberate strategic revenue hit, not a demand failure.
A common claim that is now out of date. The framing that United States producers "still depend on China" because MP sells concentrate to Shenghe, a Chinese affiliate, for refining was refuted 0-3 in verification. MP cut those China sales in 2025 and the Shenghe offtake agreement expired in January 2026. If you hear this argument in a meeting, it is stale.

The broader field. Beyond MP and Lynas, capacity is emerging in Texas, Canada, Brazil, Africa and Europe, plus recycling ventures. The open strategic question is whether these can collectively break China's roughly 90% processing share within a decade. That is a question of capital, permitting, technical scale-up and sustained political will, not geology.

31Verified findings, caveats and open questions

Each finding below survived a three-vote adversarial verification in the source report. A claim was killed only if two of three independent reviewers refuted it. Votes are shown so the reasoning is auditable.

High3-0

Rare earths are indispensable to high-performance NdFeB magnets, the dominant magnet for electric vehicle motors and wind generators, using neodymium and praseodymium for strength and dysprosium and terbium for heat resistance. This makes them structurally critical to clean energy technology.

High3-0

Global rare earth ore production is not monopolised by China, about 62% in 2019 rising to about 70% by 2024, with the United States, Myanmar and Australia each around 10% to 12%. China's true dominance is downstream: about 90% of processing, about 99% of heavy rare earth processing, and refining leadership in 19 of 20 strategic minerals at about 70% on average.

High3-0 / 2-1

China produces about 85% of rare earth oxides and about 90% of metals, with separation carried out almost exclusively in China. Its sintered NdFeB magnet share has risen to about 94%.

High3-0

China built this dominance through policy: 2010 export quotas, lifted in 2014 after a World Trade Organization ruling, post-2016 consolidation into two state groups controlling nearly all quota, and industrial policy pulling magnet making into China. Only Lynas processes at scale outside China.

High3-0

China rolled out escalating export controls: gallium and germanium in July 2023, graphite in October 2023, antimony in August 2024, tungsten and tellurium in February 2025, medium and heavy rare earths in April 2025, expanded in October 2025.

High3-0

The International Energy Agency estimates around USD 6.5 trillion in annual downstream production outside China could be jeopardised if the October 2025 controls are fully enacted, with automotive accounting for over USD 3 trillion of that exposure.

High3-0

MP Materials is the largest rare earth producer in the Western Hemisphere and owns Mountain Pass, the only at-scale mine-and-process site in North America. It supplies more than 10% of global rare earth content.

High3-0

MP ceased concentrate shipments to China in April 2025 and agreed to cease all sales to China in July 2025, recognising zero concentrate revenue in the third quarter of 2025 against USD 43.1 million in the same quarter of 2024.

High3-0

The July 2025 Department of Defense partnership includes a USD 400 million preferred equity investment, a USD 150 million heavy rare earth loan, a ten year USD 110 per kg NdPr floor, a ten year magnet offtake, and about 15% Department of Defense ownership.

High3-0

MP posted record third quarter 2025 NdPr production of 721 tonnes, up 51% year on year, and 13,254 tonnes of rare earth oxide. A Mountain Pass heavy rare earth separation facility began commissioning in mid-2026, supporting planned NdFeB magnet output of about 10,000 tonnes a year.

Caveats on the merged material.
  • Data vintage. Production share figures blend 2019 Joint Research Centre data with 2023 to 2025 figures from the International Energy Agency, the Center for Strategic and International Studies and MP. China's mine share has risen toward 70% and its processing and magnet dominance has intensified rather than eased, so directionally the numbers are conservative.
  • Single-analyst figures. The 85% oxide and 90% metal shares trace to one specialist analyst, Adamas Intelligence via the Joint Research Centre, and drew the only non-unanimous vote at 2-1, though the direction is confirmed by the International Energy Agency.
  • The USD 6.5 trillion figure spans China's broader critical minerals controls, not rare earths alone, and is contingent on full enactment of measures later reported to be delayed by about a year.
  • Company self-interest. MP Materials filings are primary SEC disclosures, but the company has an obvious interest in its Western Hemisphere and Department of Defense narrative. Those points were cross-checked against independent sources. Quarterly production metrics are self-reported and unaudited, which is standard for production releases.
Open questions, thin evidence, worth a follow-up
  • Demand and price forecasts. Specific quantified NdPr and dysprosium and terbium tonnage and price trajectories to 2030 and 2035 did not survive verification. Section 28 rests on qualitative framing.
  • Recycling and substitution maturity. How cost competitive are magnet-to-magnet recycling and rare-earth-free motors, and what share could they realistically displace by 2035?
  • Non-China build-out. What is the true cost and scale of Lynas and the emerging projects in Canada, Brazil, Africa and Europe, and can they collectively break the roughly 90% processing share within a decade?
  • Enforcement reality. Following the reported one year delay of the October 2025 controls, what is actual licensing behaviour, and how much of the USD 6.5 trillion exposure has materialised?
Appendix

Terms, mines and places

Kept at the back on purpose. Nothing here is needed to follow the report, but every recurring word and every site named above is defined here.

TermsThe words and places that keep coming up

A reference list, not a prerequisite. Every term and every mine, plant and port named in the report is here, so come back the moment something stops making sense.

The one thing to get straight first

The most misunderstood fact in this industry is that the place the material is dug up and the place it is processed are different countries.

DigMount WeldWestern AustraliaMine, crush, concentrate. Nothing more.
ShipAbout 4,000 km by seaFremantle to KuantanIt crosses the ocean as concentrate
SeparateGebeng, KuantanPahang, MalaysiaCracking, solvent extraction, finished oxide

Both ends are the same company, Lynas. The mine is Australian and the plant is Malaysian. Malaysia is not processing its own ore. It supplies the site, the workforce, the power and the water, and it keeps the radioactive residue. Almost every argument in Malaysia traces back to that single arrangement.

Terms
TermIn one lineThe fuller version
REO
rare earth oxide
The industry's common yardstickGrade, reserves and production are all quoted as the oxide, not the metal. Two reasons. First, the oxide is the first stable, weighable, saleable form the material takes. Second, each of the seventeen elements has a different metal-to-oxide weight ratio, so you cannot add them up unless you convert everything to oxide. "A deposit grading 5% REO" means 100 kg of rock yields 5 kg of oxide, not 5 kg of metal. For the light elements the metal weighs roughly 80% to 85% of the oxide.
ConcentrateWhat is left after you throw the waste rock awayOre that has been physically upgraded by removing worthless gangue. A few percent in the ground becomes 30% to 60% in the concentrate. No chemistry has happened yet.
TailingsThe rock you threw awayThe very large volume of waste rock and slurry rejected during beneficiation. The volume is the problem: a tailings storage facility needs land, engineering and permanent management, and it is one of the biggest liabilities on a mine site.
SlurryGround rock mixed with water into a thick mudOre crushed to powder and suspended in water. Why bother making it a liquid? Because a liquid can be pumped, stirred, and dosed with reagents evenly. A dry powder can do none of those three things. Almost all wet mineral processing runs on slurry. Solids are typically 30% to 60% by weight, so the consistency is close to what is inside a concrete mixer before it is poured.
FlotationStick it to a bubble and float it offReagents are added to the slurry that make the surface of the target mineral water repellent, then air is blown through. The target mineral sticks to the bubbles and rises as a froth, which is skimmed off. It looks simple, but finding the reagent combination that works on a specific ore is genuinely hard, and it is a classic place for a new project to stumble.
Ion adsorption clayRare earths merely stuck to the surface of clay, not locked inside a crystalGranite weathered by rain over millions of years. The rare earths released from the original minerals did not form new crystals. They are held on the surface of clay particles as loose ions, by electrical attraction alone. Because they are stuck on rather than chemically bonded, a room-temperature salt solution swaps them off. That single property produces three consequences: grade is extremely low at 0.05% to 0.3%; processing is nonetheless cheap because there is no crushing, grinding or acid bake; and the weathering process concentrates the heavy rare earths while leaving thorium behind. This is where the world's dysprosium and terbium actually come from, and it is the resource Malaysia holds.
Three-way catalystThe device that cleans a petrol car's exhaustThe canister in the exhaust line that cleans the gases. "Three-way" means it deals with three pollutants at once: carbon monoxide, unburned hydrocarbons, and nitrogen oxides. The hard part is that the first two must be oxidised, meaning oxygen added, while nitrogen oxides must be reduced, meaning oxygen removed. Two opposite reactions in the same box. Platinum, palladium and rhodium do the chemistry, but cerium oxide absorbs and releases oxygen to smooth the swings, which is what allows both reactions to work. Diesel aftertreatment is a different architecture and is not called a three-way catalyst.
Fluid catalytic cracking
FCC
The refinery unit that breaks heavy oil into petrolDistilling crude alone leaves far too much heavy oil, so a catalyst is used to break those large molecules into petrol and diesel. The catalyst is a fine powder that circulates through the unit behaving like a fluid, which is where the name comes from. Lanthanum is what keeps the zeolite catalyst from collapsing at operating temperature.
Roasting and crackingBurning the mineral to break it openRare earths are locked inside a crystal lattice and will not dissolve in water. Baking the concentrate with concentrated sulphuric acid at 200°C to 500°C, or boiling it in concentrated caustic, destroys that lattice and converts the rare earths into a soluble form. This is the step that concentrates thorium into the residue.
Solvent extraction
Mixer-settler
Using two liquids that will not mix to separate elements a sliver at a timeContact two liquids that do not mix, like water and oil, and one element will move into the oil side very slightly more readily than its neighbour. One pass is far too small a difference to be useful. So a unit that combines a stirred mixing chamber with a quiet settling chamber is repeated hundreds of times in series, compounding the tiny advantage. China's dominance lives in this step.
Strip castingPouring molten alloy onto a spinning cooled wheel so it freezes instantly into thin flakesMolten magnet alloy is poured onto a water-cooled copper wheel spinning at speed, and freezes within a fraction of a second into flakes 0.2 to 0.4 mm thick. Why the hurry? Metal cooled slowly grows large crystals and throws out free iron, and both wreck magnetic performance. Cooled fast, the crystals are fine and uniform and the neodymium-rich grain boundary phase is spread thinly in exactly the right places. That grain boundary phase is what makes sintering and grain boundary diffusion work later. Ice is a fair comparison: frozen slowly it grows big cloudy crystals, frozen fast the crystals are tiny. This is the same effect in an alloy. The technique spread in the 1990s and magnet performance rose sharply as a result.
SinteringFusing powder into a solid with heat, without fully melting itPressed powder is heated to just below the melting point, 1,000°C to 1,100°C for NdFeB, so the particles bond and the gaps close to near full density. Because it is not fully melted, the crystal alignment set during pressing survives. That is how an anisotropic magnet is made.
Grain boundary diffusionPainting the expensive element onto the surface instead of mixing it throughRather than blending dysprosium or terbium through the whole magnet, it is applied to the surface of the finished part and heated so it travels inward along the boundaries between crystals. Heat resistance is set at those boundaries, so the element only needs to be where it does the work. The same performance for a fraction of the heavy rare earth, which is why intensity has fallen so much over the last decade.
Coercivity
Remanence
Resistance to demagnetising / raw strengthRemanence is how strong the magnet is. Coercivity is how hard it is to demagnetise. In an electric vehicle motor the second one is the problem: coercivity falls as temperature rises, and past a certain point the loss is permanent. Dysprosium and terbium are added specifically to raise it.
Thorium
NORM
The reason any of this is a radiation questionThorium rides along inside monazite because it is a chemical lookalike for the light rare earths. The rare earths themselves are not radioactive. NORM means naturally occurring radioactive material, the regulatory category that also covers phosphate fertiliser plants, oil and gas pipe scale and tin slag. Full explanation in Stage 3.
NdPrNeodymium and praseodymium sold as one productThe two are so chemically similar that separating them costs more than it is worth. The industry ships and prices them together. When you see an NdPr price quoted, this is what is being quoted.
MischmetalUnseparated rare earth metalMostly cerium and lanthanum. Cheap because no separation was done. Used as a steel additive and in lighter flints.
Places and mines
NameWhere it isWhat it is, and why it matters
Bayan OboInner Mongolia, China
About 150 km north of Baotou
The world's largest rare earth deposit. It was developed as an iron ore mine and still produces iron ore. The rare earths come out alongside the iron, which is part of why Chinese costs are so hard to match. This single deposit accounts for a large share of world light rare earth supply, operated by China Northern Rare Earth. The name means "rich mountain" in Mongolian.
Mountain PassCalifornia, United States
Mojave Desert near the Nevada border
About 85 km southwest of Las Vegas
The only site in North America that both mines and processes at commercial scale. Owned and operated by MP Materials. It is a bastnasite deposit that supplied most of the world's rare earths from the 1960s to the 1980s, closed in 2002 under Chinese price pressure and environmental problems, and was later restarted. This is the asset the United States Department of Defense took a stake in.
Mount WeldWestern Australia
Near Laverton
About 800 km northeast of Perth, inland desert
One of the highest grade deposits in the world, owned by Lynas. What happens here is mining and concentration only. No separation. The concentrate is shipped from Fremantle to Kuantan in Malaysia. Mount Weld is the mine, Kuantan is the plant. Confusing those two makes the entire Malaysian debate impossible to follow.
KuantanCapital of Pahang state, Malaysia
On the peninsula's east coast, facing the South China Sea
About 250 km east of Kuala Lumpur, roughly three and a half hours by road
An industrial city with a deepwater port. The Lynas separation plant sits inside the Gebeng industrial estate just north of the city. Press coverage calls it "the Kuantan plant" and "the Gebeng plant" interchangeably; they are the same facility. The magnet plant joint venture with South Korea's JS Link announced in 2026 is planned near the same refinery. Malaysia's east coast is less industrialised than the west, which is part of why the government is keen to attract industry there.
GebengImmediately north of KuantanA heavy chemical and petrochemical industrial estate. The Lynas Advanced Materials Plant occupies about 100 hectares inside it. Because the estate is zoned for heavy chemicals, it can host processes an ordinary industrial park cannot.
Pahang, Perak
and Kerian
Pahang is on the east of the peninsula and contains Kuantan
Perak is in the northwest
Kerian is an industrial park in Perak
Malaysia's ionic clay resource sits in five states: Kelantan, Terengganu, Pahang, Perak and Kedah. Kerian in Perak is a frequently cited candidate location for magnet manufacturing, because it sits next to the northern semiconductor cluster and because Perak is one of those five states.
MinamitorishimaInside Japan's exclusive economic zone
About 1,900 km southeast of Tokyo
Seabed mud containing rare earths lies at roughly 6,000 metres depth. The mix is attractive, high in heavies and low in thorium, but nobody anywhere has ever lifted mud commercially from that depth. A lift test was targeted for January 2026 and a demonstration phase for 2027. Manage expectations on this one.
KalgoorlieInland mining city, Western AustraliaLynas is building cracking and leaching capacity here. It is the move that would take the radioactive residue step out of Malaysia and back to Australia. The implication is that the Malaysian plant's role could narrow over time to separation and downstream only.

Sources and basis. Process descriptions draw on EuRARE and EAG Laboratories process documentation, a 2025 Journal of Sustainable Metallurgy review of rare earth separation, published sintered NdFeB manufacturing process literature including strip casting, hydrogen decrepitation and jet milling parameters, IRENA's 2022 critical materials paper, and peer reviewed work on ion adsorption clay leaching in Malaysia. Application intensity figures draw on published magnet intensity studies for electric vehicles and direct drive wind, and on widely quoted United States Congressional Research Service and Government Accountability Office estimates for defence platforms. Market share and price figures draw on S&P Global and industry price reporting on neodymium oxide as at March 2026.

Part C sources. Part C is merged from the standalone global rare earth industry report of July 2026, which drew on the European Commission Joint Research Centre, the United States Geological Survey, the Center for Strategic and International Studies, the International Energy Agency Global Critical Minerals Outlook 2025, World Economic Forum and JOGMEC material on Japan's post-2010 response, Lynas Rare Earths disclosures, South China Morning Post reporting on the Malaysian licence conditions, National Bureau of Asian Research on Indonesian downstreaming, Global Trade Alert on the export control sequence, and MP Materials SEC filings and quarterly releases. Findings carrying a vote count survived a three-vote adversarial verification in that report; a claim was killed only where two of three independent reviewers refuted it.

Caveat. Figures are indicative, are drawn from sources of differing dates and methodologies, and should be re-verified before any external use or investment decision. Process parameters are typical industry ranges, not the specification of any particular plant. Nothing in this document is investment advice.

World Finance Edu · August 2026