Rare Earth Magnet Supply Chain Explained: Mining, Separation, NdFeB Magnets and China's Hold on Each Step
A neodymium-iron-boron (NdFeB) magnet is the strongest permanent magnet you can buy at scale. It sits inside most electric vehicle motors, many wind turbines, missile guidance fins, drones, hard drives and phone speakers. Getting one made takes four separate industrial steps: mining the ore, separating it into individual rare earth oxides, turning those oxides into metal and alloy, then pressing and sintering the alloy into a finished magnet. China leads every one of those steps, and its lead gets bigger the further down the chain you go. This page walks the chain stage by stage, explains why two obscure metals (dysprosium and terbium) matter so much, and sets out where things stand as of October 2026.
The chain at a glance
The International Energy Agency’s April 2026 rare earths report gives the cleanest set of shares, based on 2024 data. China mined about 60% of the world’s magnet rare earths, produced 91% of refined output and made 94% of permanent magnets. In 2005 its magnet share was roughly 50%. US Geological Survey figures for 2025 put Chinese mine output at about 270,000 tonnes of rare earth oxide out of a world total near 390,000 tonnes, which is closer to 70%.
| Stage | What happens | China’s share | Main producers outside China |
|---|---|---|---|
| Mining | Ore dug and concentrated | About 60% to 70% | MP Materials (US), Lynas (Australia), Myanmar, Thailand, Madagascar |
| Separation and refining | Concentrate split into individual oxides | About 91% overall; heavy rare earths over 99% | Lynas (Malaysia), MP Materials (US), Neo (Estonia) |
| Metal and alloy | Oxides reduced to metal, alloyed with iron and boron | Dominant, with very few producers elsewhere | Less Common Metals (UK), Japanese and Vietnamese plants |
| Magnet making | Alloy milled, pressed, sintered, machined, coated | About 94% | Japan (Shin-Etsu, Proterial), Neo (Estonia), MP Materials and Noveon (US) |
Mining is the stage where the rest of the world is least dependent. Refining and magnet making are where the chokehold actually sits.
Stage by stage
Mining. Rare earths are not rare in the crust; they are rarely concentrated enough to mine cheaply. The main ore types are bastnaesite (Mountain Pass in California, Bayan Obo in Inner Mongolia), monazite from mineral sands, and ionic adsorption clays found in southern China and northern Myanmar. Clays matter out of proportion to their tonnage because they carry most of the world’s heavy rare earths. Myanmar’s mines in Kachin State produced an estimated 27,000 tonnes in 2024, nearly all of it shipped to China, and they have supplied at least 60% of China’s rare earth imports since 2017. That makes China the processor of other countries’ heavy rare earth ore as well as its own.
Separation and refining. This is the hard part. Rare earth elements are chemically almost identical, so the concentrate is dissolved and run through solvent extraction circuits: long banks of mixer-settler tanks, often hundreds of stages, each nudging one element slightly apart from its neighbours. The output is a set of high-purity oxides, with neodymium-praseodymium (NdPr) oxide the most valuable by volume. Building these circuits takes years, generates acidic and sometimes mildly radioactive waste, and depends on process know-how that China added to its list of banned technology exports in December 2023.
Metal and alloy. Oxide powder can’t be pressed into a magnet. It has to be reduced to metal, usually by electrolysis in molten salt for NdPr, then melted with iron and boron and strip-cast into thin flakes of alloy. This step gets little attention and has almost no capacity outside China. Less Common Metals in Cheshire, England, now owned by USA Rare Earth, describes itself as the only company outside China offering a full range of light and heavy rare earth metals and alloys.
Magnet making. Sintered NdFeB magnets are made by breaking the alloy apart with hydrogen, jet-milling it to a fine powder, pressing it inside a strong magnetic field so the grains line up, then sintering, machining, coating and magnetising the blocks. Bonded magnets (powder mixed with a polymer) are weaker and used for smaller parts. NdFeB accounted for about 96% of rare earth magnet demand in 2025. Japan is the main non-Chinese producer, with Shin-Etsu at roughly 3,400 tonnes a year across Japan and Vietnam and Proterial at 2,500 to 3,000 tonnes, but Japanese magnet makers still buy most of their oxides and metals from China.
Light vs heavy rare earths: why dysprosium and terbium matter
Rare earths split into two groups by atomic weight. The light rare earths include lanthanum, cerium, praseodymium and neodymium. They are relatively plentiful, and neodymium and praseodymium make up most of the rare earth content of an NdFeB magnet. The heavy rare earths include terbium, dysprosium, holmium, erbium, ytterbium and lutetium. They are scarcer, pricier and concentrated in clay deposits.
A plain NdFeB magnet loses its resistance to demagnetisation as it heats up. Traction motors, wind generators and military actuators run hot, so magnet makers swap a small amount of neodymium for dysprosium or terbium. Both raise coercivity (the field strength needed to wipe a magnet out) and keep the magnet working at temperature. A newer technique called grain boundary diffusion places the heavy rare earth only at the edges of each grain, which cuts the amount needed but does not remove it.
That is why the heavy end of the chain is the real pressure point. China and Myanmar supply most heavy rare earth feedstock, and more than 99% of heavy rare earth separation happens in China. The US had 100% import reliance for heavy rare earths from 2021 through 2025, and China supplied all of its terbium imports.
Where NdFeB magnets go
Electric vehicles are the biggest growth market. Most EV traction motors are permanent magnet synchronous motors, and a single large motor can hold close to 200 individual magnets. Wind turbines are the second big user, especially direct-drive designs common offshore, where the generator needs no gearbox. Industrial motors, robots, drones, e-bikes, air conditioners and consumer electronics make up much of the rest.
Defence is a smaller market by tonnage but the one governments worry about most. Rare earth magnets go into F-35 fighters, Virginia-class and Columbia-class submarines, Tomahawk missiles, radar systems, Predator drones and JDAM guided bombs. (A widely repeated figure says each F-35 holds about 920 pounds of rare earth materials. It traces to an unpublished 2012 Pentagon study and has never been backed by a public bill of materials.) Under an amended US defence procurement rule, NdFeB and samarium-cobalt magnets in US defence contracts may not be mined, refined, separated, melted or produced in China, Russia, Iran or North Korea from 1 January 2027. Few supply chains outside China can meet that standard today.
The IEA expects magnet rare earth demand, which roughly doubled between 2015 and 2024, to grow by about a third by 2030.
China’s 2025 controls and what followed
Beijing’s licensing system is covered in detail in our timeline of China’s critical mineral export controls. The magnet story runs like this.
On 4 April 2025, China put seven medium and heavy rare earths under export licensing: samarium, gadolinium, terbium, dysprosium, lutetium, scandium and yttrium, along with magnets containing them. Shipments stalled while exporters waited for licences. Chinese customs data showed magnet exports of 1,238 tonnes in May 2025, down 74% from a year earlier and the lowest monthly total in more than five years. Ford paused Explorer production in Chicago that month, and suppliers in Europe, Japan and India reported shortages.
On 9 October 2025, China announced a far wider package. It added five more rare earths and, more importantly, claimed jurisdiction over foreign-made products containing Chinese rare earths or made with Chinese rare earth technology, a mirror of the US foreign direct product rule (see our guide to EAR, FDPR and other export control acronyms). After the Trump-Xi meeting in Busan on 30 October, Beijing suspended that package for a year. The April 2025 licensing stayed in force.
In early December 2025 China began issuing “general licences” that let major magnet makers ship to approved customers without a licence for each shipment. JL Mag was cleared to supply nearly all its customers; Ningbo Yunsheng and Zhong Ke San Huan got cover for part of theirs.
Flows have since recovered, unevenly. China exported 36,880 tonnes of magnets in January to July 2026, and July’s 5,375 tonnes was slightly above July 2024. Shipments to the EU, South Korea, India and Thailand ran well ahead of 2024 levels. The United States is the exception: it took 3,521 tonnes in January to July, about 16% below the same period of 2024, and August shipments to the US fell 21% from July to 512 tonnes.
Three 2026 events show the controls are still in active use. On 6 January, China restricted dual-use exports to Japan, including the seven rare earths used in magnets. On 22 June, it added 10 US firms to its export control list, among them MP Materials and USA Rare Earth. In late September, the two governments extended the Busan truce from 10 November 2026 to 10 January 2027, which keeps the October 2025 package suspended for now.
Building a chain outside China
Most non-Chinese capacity is new, small or still under construction. The main players as of October 2026:
- MP Materials (US). Owns Mountain Pass, the only operating US rare earth mine, and produced 840 tonnes of NdPr oxide in the second quarter of 2026. In July 2025 the Pentagon bought $400 million of preferred stock (about 15% on conversion), set a 10-year price floor of $110 per kilogram for NdPr, and agreed to buy all output from a planned 10,000-tonne magnet plant for a decade. That “10X” plant, at Northlake near Fort Worth, Texas, is due to start testing in 2028. MP’s existing Independence plant in Fort Worth began initial magnet production in December 2025, with General Motors as the anchor customer, and Apple has a separate $500 million supply deal. MP is also adding heavy rare earth separation.
- Lynas Rare Earths (Australia/Malaysia). The largest separated rare earth producer outside China, mining at Mt Weld in Western Australia and refining at Gebeng, Malaysia. It produced 7,260 tonnes of NdPr in the year to June 2026, separated its first dysprosium in May 2025 and terbium in June 2025, began samarium in March 2026, and aims to produce a full suite of heavy rare earths in Malaysia by late 2026. Malaysia renewed its operating licence for 10 years in March 2026. Plans for a heavy rare earth plant in Texas have been uncertain since 2025.
- USA Rare Earth (US/UK). Owns Less Common Metals and is building a magnet plant in Stillwater, Oklahoma, targeting a 600-tonne annual run rate in the fourth quarter of 2026. It holds a Commerce Department award of $277 million in grants plus up to $1.3 billion in loans, one of the newer awards in our CHIPS Act awards tracker.
- Vulcan Elements (US). Plans a 10,000-tonne magnet plant backed by a $620 million Pentagon loan and $50 million from Commerce.
- Neo Performance Materials (Estonia). Opened Europe’s first large sintered magnet plant at Narva in September 2025, with 2,000 tonnes a year of capacity.
- Others in development. Serra Verde in Brazil (ionic clay, with US development finance support), Iluka’s Eneabba refinery and Arafura’s Nolans project in Australia, and Norra Kärr in Sweden, a potential dysprosium source.
Governments are pushing in the same direction. G7 leaders agreed in June 2026 to aim for no more than 60% dependence on any single outside supplier for rare earths and magnets by 2030. Even so, the IEA calculates that announced capacity outside China would meet only about half of projected demand at the mining stage by 2035, a quarter at refining, and well under a fifth at magnet making.
What to watch
The near-term date is 10 January 2027, when the extended truce expires and the October 2025 extraterritorial package could return. The US Affiliates Rule on export controls sits on the same negotiating table (see the US chip export controls timeline). Nine days earlier, on 1 January, the Pentagon’s ban on Chinese-origin magnet material takes effect, and it will show quickly whether MP, Lynas, Neo and the Japanese producers can supply defence programmes at volume.
The longer test is heavy rare earth separation. Mines can be permitted and magnet lines bought off the shelf. A dysprosium and terbium circuit that runs reliably at scale is the piece the rest of the world has never had, and it is the step China has worked hardest to keep. For the wider picture of how minerals fit into the tech split between the two countries, see the complete timeline of US-China technology decoupling.