Rare Earth Magnets: Who Makes Them, and What China’s Export Pause Covers

The Mountain Pass rare earth mine and processing facility in California, photographed in February 2022. Photo: Tmy350 / CC BY-SA 4.0 (cropped and resized).

As of 4 November 2026, China's suspension of its widest export controls on rare earth magnets and their materials runs until 10 November 2026. Beijing has not said, in a legal notice, whether it will extend it. The suspended package covers five more rare earth elements, the equipment and know-how to process them, superhard materials, battery materials, and a rule that would reach foreign-made goods containing Chinese rare earths. ^{[6] [7] [8] [10] [37] [38]}

Rare earth magnets sit inside electric car motors, many wind turbine generators, drones and military systems. ^{[2] [18]} China made about 94% of the world's sintered rare earth magnets in 2024, the International Energy Agency (IEA) estimates. It also did 91% of the refining and 60% of the mining of the four elements that matter most for rare earth magnets. ^[2] On the IEA's later measure, China's refining share was 85% in 2025. ^[4]

The numbers differ because the steps differ. The headline "China controls the supply" hides that.

Mining is the least concentrated step. Refining is slowly diversifying. Magnet making is the hardest step to copy, and it is where the gap outside China is widest. ^{[2] [4]}

The legal position is narrow. China's official notice suspends the October 2025 package until 10 November 2026. ^[10] Separate April 2025 licences on seven heavy and medium rare earths were never paused. ^{[1] [5]}

After talks on 20 to 23 September, Beijing's commerce ministry said the two sides agreed to extend the wider US-China arrangement to 10 January 2027. That text does not name rare earths. ^[12] The ministry's announcement list, read on 4 November, shows no notice that changes the 10 November date. ^[14]

Where things stand:

• Verified: China's lead is deepest in refining and magnet making, and its April 2025 licences were still in force on 4 November 2026. ^{[2] [5]}
• Contested: whether the January 2027 extension covers the rare earth pause, which officials did not say in writing. ^{[12] [14]}
• Unresolved: what Beijing will do when the suspension runs out on 10 November, and whether non-Chinese magnet plants reach scale in time. ^{[2] [4]}

Rare earths are a group of 17 elements: the 15 lanthanides, plus scandium and yttrium. ^[2] An older USGS fact sheet counts 16, leaving out scandium. ^[33] They are not rare in the ground. Workable deposits are uncommon, and the elements are chemically so alike that separating them is hard. ^[2]

Lithium, cobalt, nickel and graphite are critical minerals, but they are not rare earths. ^[3] They go into batteries.

Rare earths go into magnets, catalysts, glass polish and other uses. USGS says the leading global use is rare earth magnets. ^[1] The IEA puts permanent magnets at about 95% of rare earth consumption by value. ^[2]

Two families matter for magnets:

• Neodymium-iron-boron (NdFeB). The strongest common rare earth magnet. It is made mainly from neodymium and praseodymium, often with dysprosium and terbium added. ^[2]
• Samarium-cobalt. Used where heat and corrosion matter. It relies on samarium. ^[5]

The four "magnet rare earths" are neodymium, praseodymium, dysprosium and terbium. ^[2] Neodymium and praseodymium are light rare earths. Dysprosium and terbium are heavy rare earths. ^[33]

Standard NdFeB grades can be built from neodymium and praseodymium alone. Dysprosium and terbium raise a magnet's resistance to demagnetisation. Peer-reviewed work says this helps rare earth magnets keep their strength at the high temperatures of traction motors, which one review puts above 150 °C. ^{[34] [35]} So high-performance grades often carry the heavy elements, which are scarcer and pricier.

Making a rare earth magnet takes five steps. Rock is mined and concentrated. Mixed material is separated into individual oxides. Oxides become metals, then alloys, and alloy powders are pressed, sintered and finished as rare earth magnets. ^[2]

China's share differs at each step, so no single number describes the chain.

Two details explain the mining rows. The IEA counts only the four magnet elements. USGS counts all rare earths, including cerium and lanthanum, which are mined in bulk. ^{[1] [2]}

The USGS estimate for 2025 puts the United States second, at 51,000 tons. Australia follows with 29,000 tons and Burma with 22,000 tons. ^[1]

Reserves tell a similar story. USGS lists China with 44 million tons of the world's more than 75 million. Brazil has 11 million. ^[1] That suggests mining is where others can compete most easily, though that is our reading, not the agencies'.

The 2025 refining figure comes with a trend. In July 2026 the IEA said new US projects and more output in Malaysia cut the top refiner's share from over 90% in 2023 to 85% in 2025. It projects 70% by 2035 if planned projects open on time. ^[4] That figure is on the IEA's own basis, so it is not a fall from the 91% above.

Outside China, Japanese and German firms are the established makers of sintered magnets. This research found no reliable tonnage or share for any of them, so it names none. The IEA's 94% leaves about 6% of 2024 sintered output made elsewhere. ^[2]

The sources point to three reasons.

First, separation is hard. Chemically near-identical elements must be pulled apart in long chains of solvent-extraction tanks. Know-how and skilled staff matter as much as the ore. ^{[2] [3]}

Second, scale. China's rare earth magnet industry buys huge volumes of oxide. The IEA says that creates economies of scale and a stable base of demand for upstream material. ^[2]

Third, policy. China's October 2025 notices would have controlled magnet-making equipment, such as hydrogen decrepitation furnaces, jet mills, sintering furnaces and grain-boundary diffusion equipment. ^[6]

They would also have controlled the technology for mining, separation, metal smelting, magnet making and recycling. ^[9] Both notices are paused. They show which stages the controls targeted.

Outside China, projects face higher costs. The IEA lists smaller scale, higher input prices, complex permits and strict environmental rules. ^[2] Customers also want proof a plant works before they sign long contracts, so developers carry risk first. ^[2]

A caution on headlines. Figures of "90%" and "99%" circulate for rare earth magnets and heavy rare earths. This article uses only numbers tied to a named agency and year. For heavy rare earths specifically, no share is cited here, because the IEA reports the four magnet elements together. ^[2]

Demand for the four elements behind rare earth magnets has doubled since 2015, the IEA says. Under today's policies it will grow by about a third by 2030. ^[2] Outside China it is projected to rise by 50% by 2035, with electric vehicles making the largest contribution. ^[2]

The uses are everyday ones:

• Electric cars: traction motors use NdFeB rare earth magnets. ^[2]
• Wind turbines: many use permanent-magnet generators. ^[2]
• Industry: motors and robots matter more after 2030. ^[2]
• Data centres, aerospace, defence and medicine: all are on the IEA's list. ^[2]

Small electronics use them too. USGS notes that small, strong rare earth magnets let makers shrink audio gear, computers and phones. ^[33] MP Materials launched a drone-industry demand project, Project Swarm, in 2026. ^[18] The IEA publishes no drone figure, and this article gives none.

Defence is not the main source of demand, but it shapes policy. A US Defense Department rule is due to reach the whole magnet supply chain on 1 January 2027, for Defense Department contracts. For NdFeB magnets it bars delivery of items made with material mined, refined or produced in China, Russia, Iran or North Korea. ^[16]

Law-firm coverage reports a class deviation that eases the rule for older contracts, and industry calls for a delay. This research did not confirm whether the date has moved. ^[39]

The IEA makes a point about price. Rare earths are about 40% of a permanent magnet's cost but less than 1% of a car's value. ^[3]

A tripling of rare earth prices would add about 0.1% to a car's cost. ^[3] The risk is not that rare earth magnets cost too much. It is that they may not be available at any price.

China has acted in two waves. The words "ban" and "embargo" are wrong for both. They are licensing regimes. An exporter needs a licence from the commerce ministry, and the ministry decides.

April 2025: Announcement 18

Issued on 4 April, it took effect the same day. It requires licences for seven elements: samarium, gadolinium, terbium, dysprosium, lutetium, scandium and yttrium. ^[5]

It covers their metals, alloys, oxides and compounds. It also covers samarium-cobalt magnets and NdFeB magnets that contain terbium or dysprosium. ^[5]

NdFeB magnets made from neodymium and praseodymium alone are not on that list. This regime was still in force on 4 November 2026. ^[1]

October 2025: six notices

All were issued on 9 October.

• Announcement 55: superhard materials. ^[37]
• Announcement 56: equipment and raw materials for rare earth processing and magnet making. ^[6]
• Announcement 57: holmium, erbium, thulium, europium and ytterbium. ^[7]
• Announcement 58: lithium batteries and artificial graphite anodes. ^[38]

Announcement 56 named the machines that make rare earth magnets. They include strip-casting and hydrogen furnaces, jet mills, sintering furnaces and grain-boundary diffusion kit. ^[6] It would have taken effect on 8 November 2025. ^[6]

Two more notices went further. Announcement 61 is the "foreign direct product" rule. It would require a Chinese licence to export certain foreign-made goods that contain at least 0.1% Chinese-origin controlled rare earths by value. It would also cover goods made abroad with Chinese rare earth technology. ^[8]

Announcement 61 adds stricter conditions. Applications from foreign military users are in principle refused. Applications for chips at 14 nanometres or below, and for some military-use artificial intelligence, are decided case by case. ^[8] That ties rare earth magnets to the chip fight described in Semiconductor Leverage.

Announcement 62 controls rare earth technology and the people who share it. It also bars Chinese citizens and firms from giving substantial help to foreign rare earth mining, separation, smelting, magnet making and recycling without a licence. ^[9]

Announcement 62 took effect on issue. The foreign-made-goods rule in Announcement 61 was due from 1 December 2025. ^{[8] [9]}

In 2026 China also barred dual-use exports to Japanese military users and uses that raise Japan's military strength. The IEA read this as a sign that the underlying risk remains. ^{[2] [15]}

After April 2025, exports of the controlled elements and magnets fell sharply in April and May. The IEA says many automakers in the United States, Europe and beyond struggled to source rare earth magnets. Some cut output or stopped production for a time. ^[2]

Ford is one named case. The company idled its Chicago Explorer plant for a week in May 2025, citing magnet supply. ^[32]

Licences were later granted and volumes recovered. Outside China a premium remained. The IEA says rare earth magnets made outside China still sell at a significant premium. ^[2] In Europe, dysprosium and terbium cost about five times the Chinese domestic price when the IEA wrote in July 2026. ^[3]

The premium shows a lasting effect. China did not cut off supply; licensing slowed it. Prices elsewhere stayed higher afterwards, but how much of that is risk and how much is higher non-Chinese cost is not answered in the public sources.

The text of Announcement 70 is short. From 7 November 2025 until 10 November 2026, Announcements 55 to 58, 61 and 62 are suspended. ^[10] Announcement 18 is not on that list.

A second notice, 72 of 9 November 2025, suspends the second paragraph of an older dual-use rule on exports to the United States, 46 of 2024, until 27 November 2026. ^[11] It is not a rare earth notice, but it is a second date to watch.

On 30 October 2025, the commerce ministry said China would suspend the October measures for one year and would study and refine specific plans. ^[13] Of the 20 to 23 September 2026 talks, it later said the two sides agreed to extend the Kuala Lumpur arrangement to 10 January 2027 and keep talking. ^[12]

It did not name rare earths. The 2025 statement did tie the October export measures to that arrangement. ^[13] So an extension of the whole package is plausible. It is an inference, not something written down.

Each case, and what follows from it:

• If Beijing extends the pause by new notice, the October package stays paused for the new period. ^[10]
• If it lapses with no notice, nothing in Announcement 70 extends the suspension, and the ministry would have to confirm when each notice takes effect. ^[10]
• If Beijing says the pause continues without a new notice, the legal text and the statement would differ, and firms would have to ask the ministry which applies. ^{[10] [12]}
• If a new notice changes the package, its text, not this article, governs.

On the second case, the IEA says full implementation would put about USD 6.5 trillion a year of downstream production outside China at risk. ^{[2] [3]} Beijing could also issue a narrower replacement. The April licences, which reach rare earth magnets containing terbium or dysprosium, continue in every case. ^{[1] [5]}

The ministry's own announcement list, read on 4 November, ends with entries dated 19 October. None changes Announcement 70. ^[14]

One think-tank report says the suspension has already been extended to 10 January 2027. We found no notice that says so, and other recent reports still give 10 November. ^[36] Reports of Beijing's plans are commentary, not law. Readers should check the ministry's site on the day, and any notice after 4 November supersedes this text.

The IEA ranks magnet rare earths among the materials most exposed to supply risk. It cites high concentration, few substitutes and critical uses. ^[3]

The projects are real, but each is small next to what is needed.

United States

MP Materials. The Defense Department agreed on 10 July 2025 to buy $400 million of MP preferred stock, worth 15% with a warrant. It set a 10-year price floor of $110 per kilogram for neodymium-praseodymium products. ^[17]

MP says the department also committed to ensure that defence and commercial customers buy all magnets from a new "10X" plant for 10 years. The plant is due to start commissioning in 2028. MP says US rare earth magnet capacity would reach 10,000 tons when it is done. ^[17]

In the quarter to 30 June 2026, MP produced 840 tons of neodymium-praseodymium. Its Fort Worth plant has made NdFeB magnets since December 2025. Magnet revenue was $16.5 million, and MP said construction of the 10X plant had sped up. ^[18]

Vulcan Elements. On 3 November 2025 Vulcan and ReElement announced a $1.4 billion partnership with the US government, which includes a conditional $620 million loan to Vulcan and $80 million to ReElement. ^{[22] [24]} Vulcan picked Benson, North Carolina, for a $1 billion plant with 10,000 tons of capacity on 18 November 2025. ^[23] The loans were conditional at announcement. This research found no report, in a search on 4 November 2026, that they have closed.

USA Rare Earth. It commissioned Phase 1a of its Stillwater, Oklahoma magnet line in March 2026. It targets 600 tons a year by the end of 2026 and 1,200 tons in early 2027. ^[25] No later update was found.

Australia

Lynas. The company produced 3,481 tons of rare earth oxides in the quarter to 30 June 2026, including 1,857 tons of neodymium-praseodymium. It also made 19 tons of dysprosium and terbium. It raised the budget of its Malaysian heavy rare earth project from about A$180 million to about A$294 million. ^[19] In July 2026 it agreed to invest about A$50 million in JS Link of South Korea to build a 3,000-ton NdFeB magnet plant in Kuantan, Malaysia. ^[19]

Iluka. Its Eneabba refinery in Western Australia was about 60% built at its August 2026 results. Iluka plans commissioning in 2027, at a cost of A$1.7 to 1.8 billion. Nameplate output is 5,500 tons a year of neodymium-praseodymium oxide and 750 tons of dysprosium and terbium oxide. ^[20] The government facility behind it totals A$1.65 billion. ^[20]

Government framework. On 20 October 2025 Australia and the United States signed a critical minerals framework with a US$8.5 billion pipeline. Australia agreed to take a US$100 million stake in Arafura's Nolans project. ^[29]

Europe

Neo Performance Materials. It opened a magnet plant in Narva, Estonia in September 2025. It starts at 2,000 tons a year and plans 5,000. German suppliers Schaeffler and Bosch signed contracts at the opening. ^[21]

Solvay. The Belgian group's La Rochelle plant is reported to be adding heavy rare earth separation. Reports in July 2026 said Solvay planned industrial-scale dysprosium and terbium output from autumn. That rests on one report, and no start had been confirmed by 4 November. ^[26]

EU law. The Critical Raw Materials Act sets 2030 goals: 10% of annual need from EU extraction, 40% from EU processing and 25% from recycling. No more than 65% of any strategic raw material should come from one non-EU country. ^[27] A European Parliament briefing in November 2025 said the October controls exposed how hard EU dependence is to ease. ^[28]

Japan, India and the seabed

Japan's research ship Chikyu raised rare earth mud from deep water near Minamitorishima island in early February 2026. ^[30] This is a test, not a supply.

India's cabinet approved a 7,280 crore rupee scheme on 26 November 2025. It aims for 6,000 tons a year of sintered magnet capacity. ^[31]

How much do these add up to?

The IEA counts all announced projects as of early 2026. By 2035, existing and announced capacity outside China would cover about 50% of mining demand, 25% of refining demand and well under 20% of rare earth magnet demand. ^[2]

Mining outside China could pass 50,000 tons of rare earth content. Refining would be under 40,000. Metals, alloys and magnets together would be about 18,000. ^[2]

The IEA's tonnages and percentages use different bases, so do not divide one into the other. Magnet making is the main bottleneck. ^[2]

The IEA estimates about USD 60 billion of investment over the next decade for diversified magnet rare earth supply. Refining needs nearly half and magnet making about a third. ^[2] It compares that with the USD 6.5 trillion a year at risk. ^[2]

On the hopeful side: Rare earths were the notable exception in 2025 to rising concentration in refining. ^{[3] [4]} Several non-Chinese plants are running. A price floor backs some projects, and the US defence rule is due to create buyers from 1 January 2027 if the date holds. ^{[16] [17] [39]}

On the cautious side:

• Output is tiny: Lynas made 19 tons of heavy rare earths in a quarter, on one press account of its filing. ^[19]
• Costs are rising: Lynas lifted its heavy project budget by more than half, on the same account. ^[19]
• Money is conditional: Vulcan's loans were conditional at announcement. ^{[22] [24]}
• Know-how is scarce: processing needs equipment and skills held in few countries. ^[3]

Iluka's 750-ton heavy plant is a 2027 start. ^[20] China is also not standing still, since a paused notice would have controlled the technology itself. ^[9]

Recycling and design could help. The IEA says recycling could cut the need for primary supply by up to 35% by 2050. ^[2] It does not say it can replace new supply before 2035.

Beijing's side deserves a hearing. It says controls on dual-use goods are lawful, follow international practice and serve non-proliferation. ^[13] It also tied its October pause to a US pause on its own affiliates rule. ^[13] What the sources support is narrower than "weaponisation": China built a chain others depend on, and it has shown that licensing can slow supply and shape terms.

Related reading: The African Theatre covers the contest for critical minerals in Africa. Greenland holds an estimated 1.5 million tons of rare earth reserves, according to USGS, but shows no production in the 2024 to 2025 data. ^[1] The Arctic angle is in The Northern Frontier.

The facts come from the official Chinese notices, read in Chinese on the commerce ministry's site; the IEA and USGS reports; the US acquisition regulation; company releases and results; and press reports where no filing was available. We separated what an official text says from what commentators say it means. A company's plans are not output.

The Chinese notices are our own translations of the original text, not official English versions. We did not read the annex of Announcement 61, so its full product list is not described. The IEA's China shares are for the four magnet elements in 2024, while USGS counts all rare earths, so the two should not be mixed.

The reports on Lynas, Neo, USA Rare Earth, Solvay, India and Japan's seabed trial are press accounts of company or government statements. We did not read the filings behind most of them. Company results are the companies' own figures and were not audited here.

The status check found no notice extending the suspension. We read the ministry's announcement list on 4 November 2026, and a later notice would supersede this text. We did not analyse Chinese customs data or rank Chinese magnet makers by share, because we found no verifiable source for firm-level shares. We held no interviews.

China made about 94% of the world's sintered rare earth magnets in 2024, 91% of the refining and 60% of the mining of magnet rare earths, by the IEA's count. ^[2] On the IEA's later measure its refining share was 85% in 2025. ^[4]

The concentration is deepest in refining and magnets, not in the ground. The April 2025 licences on seven heavy and medium rare earths and some of their magnets were still in force on 4 November. ^{[2] [5]}

The October 2025 package is suspended until 10 November 2026 by Announcement 70. Whether it is extended is unresolved. ^[10]

The commerce ministry's statement on the 20 to 23 September talks extended a wider arrangement to 10 January 2027 without naming rare earths. Its announcement list shows no notice changing Announcement 70. ^{[12] [14]}

Alternatives are real and still small. MP, Lynas, Iluka, Neo, Vulcan and USA Rare Earth have plants, deals or loans, but in 2035 the IEA expects existing and announced capacity outside China to cover well under a fifth of magnet demand there. ^[2]

As of 4 November 2026, China is dominant in refining and magnet making and less so in mining, and no single percentage describes all three. What changes on 10 November depends on a notice that had not been published when this was written.