Rare Earths

The rare earths are 17 chemically similar elements: the 15 lanthanides plus scandium and yttrium. They are not geologically rare, but they occur mixed together at low concentrations and separating them is a filthy, capital-intensive chemical business that the West largely abandoned in the 1990s and China perfected. The money is in magnets: neodymium and praseodymium (traded together as NdPr oxide), boosted with dysprosium and terbium for heat resistance, make the NdFeB permanent magnets inside EV motors, wind turbine generators, hard drives, fighter jets, and guided munitions. China mines roughly 70 percent of world output, performs roughly 91 percent of separation and refining, and manufactures roughly 90 percent of sintered magnets. The only significant western producers are MP Materials at Mountain Pass, California, and Australia's Lynas, which separates in Malaysia; its planned Texas plant stalled in 2025.

The name is a double misnomer worth unpacking. They are called "rare" not because they are scarce, cerium is more common in the earth's crust than copper and even the scarcest of them beats gold, but because they almost never occur in concentrated, easily separated deposits, so for a century chemists could isolate only minute amounts. "Earths" is archaic chemistry for a metal oxide that early furnaces could not reduce any further (the same sense as "alkaline earths"), and the rare earths were first isolated as oxides. And to head off a common confusion: these are chemical elements, the sense in which a nutrition label lists "minerals" like iron and zinc (inorganic elements the body needs), rather than the crystalline compounds a geologist calls minerals. The common thread is simply that all three, dietary minerals, rock minerals, and rare earths, are inorganic and come out of the ground.

The market's defining precedent came in September 2010, when China halted rare earth shipments to Japan during the Senkaku Islands trawler dispute. Prices rose as much as tenfold within a year, dysprosium more, governments panicked, Molycorp and Lynas were financed into existence, and Japan built stockpiles through JOGMEC. The spike collapsed almost as fast once China relaxed quotas, and a 2014 WTO ruling forced their formal removal, but the lesson held: rare earth supply is a policy variable. In April 2025 China proved it again, imposing export licensing on seven medium and heavy rare earths and on finished magnets in response to US tariffs. Magnet shipments stopped for weeks, western auto plants idled, and the controls were expanded in October 2025. After a trade truce with Washington, China suspended the October expansion for one year in November 2025, while the April 2025 licensing regime stayed in force with general licenses issued to selected exporters, the position that still held as of June 2026.

Where the ore comes from. About 70 percent of mined rare earths come out of China itself, from three kinds of deposit. The largest is Bayan Obo in Inner Mongolia, an iron ore mine where the rare earths are a by-product, which holds more than 80 percent of China's reserves and feeds the light rare earth plants at nearby Baotou. Sichuan adds more bastnaesite, the other main light rare earth ore. The heavies come from ionic clays in Jiangxi, Guangdong and Fujian, weathered granite in which the rare earths cling loosely to the clay and can be washed out with a salt solution, and increasingly from the same clays across the border in Myanmar's Kachin State, which sent China about 27,000 t in 2024 and supplies much of the world's dysprosium and terbium; a rebel army took the main mining district late that year. Outside China, ore comes from Mountain Pass in California (sold to China for separation until 2025), Mt Weld in Western Australia (separated by Lynas in Malaysia), and the monazite beach sands of India and Brazil, a mineral that carries radioactive thorium.

What a refinery does. A mine produces a concentrate holding all the rare earths together. It is first "cracked" in hot concentrated sulphuric acid or caustic soda to dissolve them, and then separated by solvent extraction: the solution is stirred with an organic solvent in a tank called a mixer-settler, and one element moves into the solvent slightly more readily than its neighbour. Adjacent rare earths differ so little that a single stage barely shifts the balance, so the tanks are run in cascades of dozens for each split and hundreds for a full separation. A plant splits whatever its ore contains, one pair after another, so a single plant can in principle produce every element in its feed; in practice each plant is designed around one ore, and one built for light ore cannot separate heavies it was never sent. Lynas in Malaysia produced only light rare earths until May 2025, when it became the only commercial producer of separated dysprosium outside China.

What one costs. Outside China, from several hundred million dollars to about two billion. Lynas built its Malaysian plant for about RM 1.2 billion over two phases from 2011, roughly $400 million at the time. Iluka's refinery at Eneabba in Western Australia, designed to separate both light and heavy rare earths, was re-estimated in 2024 at A$1.7 to 1.8 billion, up from A$1 to 1.2 billion at approval, and is funded mostly by Australian government loans. Lynas's planned heavy rare earth plant in Texas stalled in 2025 when its costs outgrew a US government contribution of $258 million, wastewater handling being one of the overruns. A large new oil refinery costs ten to twenty billion dollars, so the plant itself is smaller; the problem is that it sells into a market whose price China sets.

Pollution and labour. The hazard differs from an oil refinery's. A refinery's main danger is fire and explosion in a plant handling hundreds of thousands of barrels a day. A rare earth plant handles far smaller volumes, but every tonne of ore brings acid, fluorine and the radioactive thorium and uranium found in most rare earth minerals, and the residue has to be stored permanently. Chinese estimates put each tonne of separated oxide at about 75 cubic metres of acidic wastewater and around a tonne of radioactive residue. The southern clay mines did their damage through the ground: ammonium sulphate pumped into hillsides to leach the clay seeped into rivers and groundwater, and China's industry ministry put the cost of repairing the land around Ganzhou in Jiangxi at 38 billion yuan, about $6 billion. Mountain Pass had about 60 pipeline spills between 1984 and 1998, putting some 600,000 gallons of radioactive and hazardous wastewater onto the desert, and separation there stopped in 1998. Malaysia made Lynas move its acid-cracking step, which produces the radioactive residue, to Kalgoorlie in Western Australia as a condition of renewing its licence. The plants themselves are not especially labour intensive (Lynas Malaysia runs with about a thousand employees and contractors, like a mid-sized chemical works); the hand labour is at the clay pits in Myanmar. What a plant needs is chemists and engineers who know how to run the cascades.

Why China. Its lead was built over forty years from several advantages at once. Bayan Obo's rare earths come out of an iron mine that already pays its way, so the ore is cheap. In the 1970s the Peking University chemist Xu Guangxian worked out the theory of cascade extraction, which let a plant calculate a separation circuit instead of tuning it by trial, and his students staffed the industry. Deng Xiaoping said in 1992 that "the Middle East has oil, China has rare earths", and the state treated the industry as strategic. Through the 1990s and 2000s environmental rules were weakly enforced, so the cost of the waste fell on the land rather than the price, and Chinese output undercut Mountain Pass until it closed. Today the advantage is a complete chain: China separates, makes the metals and alloys, and makes about 90 percent of the magnets, so a separator has its customers next door. In December 2023 Beijing banned the export of rare earth extraction and separation technology, and it can move prices with quotas and export licences faster than a western plant can be built, which is the risk every outside project has to price. That is why the western plants that exist lean on government money: the Pentagon's floor price for MP Materials, Australian loans for Iluka, US funds for Lynas.

There is no exchange contract of any consequence. Prices are assessments published by Asian Metal, Shanghai Metals Market (SMM), and Argus, quoting NdPr oxide, dysprosium oxide, and terbium oxide mostly in yuan per tonne with USD equivalents, and most physical volume moves under term contracts and offtakes tied to those assessments. The western policy response escalated in July 2025 when the US Department of Defense took a large equity stake in MP Materials, guaranteed a $110 per kg floor price for NdPr oxide, roughly double the prevailing Chinese price, and committed to offtake from a new magnet plant. A two-tier market, Chinese price and western strategic price, is now openly forming.

Private investors can buy the metals themselves, almost entirely through a handful of German firms. The best known, Tradium in Frankfurt, sells neodymium, praseodymium, dysprosium and terbium, the four with real scarcity value (nobody stockpiles cerium), in sealed drums carrying a batch number and a laboratory analysis certificate. They sit in a converted Second World War bunker east of Frankfurt run by its sister company Metlock, behind a door weighing more than four tonnes. Because the bunker is a bonded customs warehouse, no import duty or VAT is charged while the drums stay inside, which is much of the pitch; take delivery and German VAT of 19 percent falls due. The difficulty is the exit. There is no exchange and no public bid, so the natural buyer is the firm that sold the metal, at a price it sets against the assessed market, and an industrial buyer will usually want its own analysis rather than rely on the seller's certificate. The price itself follows Chinese quota and export policy rather than any store-of-value logic: NdPr oxide halved within two years of its 2022 peak. Germany's consumer-testing foundation Stiftung Warentest told readers in March 2025 not to buy industrial metals in physical form at all, and in Britain the Financial Services Authority warned in 2012 about telephone salesmen selling rare earths to the public at several times the market price; four of them were later jailed.

Osmium, which turns up in the same sales literature, belongs to the platinum group rather than the rare earths. The densest element there is, it is recovered at a rate of a tonne or two a year as a by-product of platinum and nickel refining, mostly in South Africa. Its industrial uses are small: osmium tetroxide stains cell membranes so they show up under an electron microscope, and serves as a catalyst in organic chemistry (part of the work behind Barry Sharpless's 2001 Nobel Prize), while hard osmium-iridium alloy tips fountain-pen nibs, instrument pivots and electrical contacts. It also lit the first metal-filament lamps, which is why Osram is named after osmium and Wolfram, the German word for the tungsten that replaced it within a few years. Raw osmium powder slowly forms osmium tetroxide in air, a volatile and highly toxic compound, so the investment product is crystalline osmium: melted and crystallised in Switzerland, scanned so that each piece carries a unique identification code in a central database, and sold through the Osmium-Institut in Germany, which Bloomberg described as "a sort of for-profit advocacy group" founded by the entrepreneur Ingo Wolf. The same network crystallises, certifies and resells the metal, so a holder's exit price is whatever that network will pay. There is no exchange, no daily fix, and no industrial buyer who needs a polished crystal.