Critical minerals: processing, not mining, is the chokepoint

Public debate about critical minerals tends to ask where the ore is, which suggests a geological contest that new discoveries could settle. The record says otherwise. The International Energy Agency finds that, excluding rare earths, the leading refining country's average share of refined supply rose to 72 percent in 2025 from 70 percent in 2023, and that Indonesia, for nickel, and China, for the other key energy minerals, accounted for over three-quarters of the growth in refined supply over those two years (IEA 2026). Mining is comparatively diverse. Refining is not, and each year of growth has gone to the same few refiners.

The asymmetry in value makes the position sharper. Rare earths are about 40 percent of the cost of a permanent magnet but under 1 percent of the value of a vehicle, and critical minerals as a group are about 3 percent of an average electric vehicle's price. A tripling of rare earth prices would add about 0.1 percent to the cost of a car (IEA 2026). The input is cheap, so buyers never paid to secure it, and the cost of its absence is the entire downstream product: the IEA estimates that full implementation of China's expanded rare earth controls would put about 6.5 trillion dollars a year of downstream production outside China at risk.

The useful frame is therefore a chain, not a deposit. Value passes from mine to concentrate, to separated oxide, to metal and alloy, to magnet, to motor. Security depends on whether an alternative exists at every link, in volume, at an acceptable price, and by a date that matters. Deposits are necessary and almost irrelevant; capacity in the middle of the chain is the scarce asset.

Separating rare earths is chemistry at industrial scale. The elements are chemically near-identical, so separating them requires long cascades of solvent-extraction stages, each tuned by operators whose skill is largely tacit, and the waste streams include radioactive thorium that must be permitted, stored, and defended politically. The business has low margins and long learning curves, which favours incumbents with scale and tolerance for environmental cost.

The IEA's cost data shows why entrants struggle. Refining capital costs outside the dominant supplier run from 20 percent to more than 150 percent higher, and operating costs about 50 percent higher on average. Diversifying magnet rare earth supply would require roughly 60 billion dollars over the next decade (IEA 2026). A producer facing those costs against a competitor able to price below them has no commercial path unless something changes the revenue side. In 2025, neodymium-praseodymium oxide sold inside China for under 60 dollars per kilogram, roughly half the 110-dollar floor later guaranteed to MP Materials (Columbia CGEP 2025).

This is the mechanism that turns a market into a chokepoint. A dominant processor can lower prices when new competition appears, then raise them or restrict access once rivals fail, and a licence regime adds the further ability to ration supply by customer. Western producers could not survive the first half of that cycle, so they never reached the point at which the second half became a risk. Mines opened in allied countries were shipped to China for refining because nowhere else could do it economically.

The concentration varies sharply by link. For rare earths, an assessment published in September 2026 puts China at about 90 percent of global processing and magnet fabrication capacity, about 88 percent of light rare earth oxide separation, and about 97 percent of heavy rare earth oxide separation (Bloomsbury 2026). The heavy elements, dysprosium and terbium, are the ones that let a neodymium magnet survive heat, and therefore the ones that electric motors, drones, and missile actuators cannot do without. The light elements are also concentrated, but the heavy elements are where the dependency is close to total.

The same pattern appears beyond rare earths. China's share of global copper smelting capacity rose from about 15 percent to 50 percent by 2025, with utilisation below 70 percent outside China against roughly 85 percent inside, and benchmark treatment charges for 2026 were set at zero dollars per tonne for the first time (IEA 2026). Smelters outside China are operating at a loss or half-idle. The IEA also notes that the dominant supplier accounts for over three-quarters of battery pre-treatment capacity and 90 percent of material-recovery capacity, so recycling does not escape the concentration.

China's export controls have escalated in steps. The IEA records that the number of mineral tariff codes under Chinese export controls has tripled since 2023. In April 2025, China began licensing seven medium and heavy rare earths and the magnets containing them. In October 2025 it expanded the regime to cover magnet-manufacturing technology and foreign-made products containing Chinese-origin material, announced controls on battery chokepoints including cathode materials, graphite anode materials, and battery equipment, and then suspended the October measures for a year as part of a trade truce. In May 2026 it curbed sulphuric acid exports after sulphur supply disruptions, a reminder that the instrument extends to process inputs as well as products (IEA 2026).

The suspension has been extended once. China's commerce ministry announced on 28 September 2026 that the suspension of the October 2025 package would run to 10 January 2027, rather than expire in November, after a Washington summit that produced no documented agreement on critical minerals (Bloomsbury 2026). The April 2025 licensing of heavy rare earths was never suspended. Beijing continues to treat these as dual-use items subject to legal and security review, which means access has become a regulated process shaped by end-use, not a plain commercial transaction (Foreign Policy 2026).

The control is not unique to China. The Democratic Republic of the Congo introduced a cobalt export quota, and Zimbabwe and Mozambique restricted lithium and graphite exports (IEA 2026). Resource nationalism is spreading wherever a producer sees that raw exports capture little of the value, which complicates any Western strategy that assumes friendly mining jurisdictions will stay open. The licence regime is a dial rather than a switch, and its main effect is uncertainty, which raises the cost of building any downstream business that depends on it.

Washington's response has moved from grants to state-backed demand. In July 2025 the Department of Defense agreed a ten-year price floor of 110 dollars per kilogram for MP Materials' neodymium-praseodymium products, structured as a contract for difference, and committed to a new magnet facility intended to take total US magnet capacity to about 10,000 tonnes, with commissioning expected in 2028 (Federation of American Scientists 2025). The floor is high enough to work as a subsidy against Chinese pricing: MP reported 1,006 tonnes of neodymium-praseodymium sales in the second quarter of 2026 and 17.6 million dollars of price-protection income in that quarter (MP Materials 2026). The Federation of American Scientists warns of a fiscal gap, since the Defense Production Act fund request for fiscal 2026 was smaller than the payments the floor implies at current prices.

The model is spreading. In March 2026 Lynas signed a binding letter of intent with the US government for roughly 96 million dollars of light and heavy rare earth oxides over four years, again with a 110-dollar floor for neodymium-praseodymium oxide (BNN Bloomberg 2026). In February 2026 the administration and the Export–Import Bank announced Project Vault, a stockpile financed by a 10 billion dollar loan and about 2 billion dollars of private capital, with manufacturers committing to purchase at predetermined prices (Nasdaq 2026). Section 232 action in January 2026 imposed no tariffs but directed negotiators to seek agreements with allies, with minimum import prices and tariffs held in reserve (White & Case 2026). Allied financing followed: the US and Australia reported that, within six months of their October 2025 framework, they had committed 2.2 billion and 1.4 billion US dollars respectively, and Alcoa took a final investment decision on a 100-tonne-a-year gallium plant at Wagerup in July 2026 (Australian Government 2026; Mining.com.au 2026).

Procurement law adds a hard deadline. From 1 January 2027, Pentagon rules bar neodymium-iron-boron and samarium-cobalt magnets whose production at any stage, from mining to fabrication, occurred in China, Russia, Iran, or North Korea. US suppliers have said they will not be ready, and the administration has reportedly weighed relief for some Chinese material (Oregon Group 2026). A demand mandate set ahead of capacity creates the pressure that builds supply only if the mandate holds, and the date has now become another negotiating variable.

The clearest evidence that the response is only partial is price. In the third quarter of 2026, Benchmark Mineral Intelligence reported dysprosium oxide in Europe at about 16.3 times the Chinese domestic price, up from 6.8 times at the end of the second quarter, with the September average near 3,500 dollars per kilogram; terbium oxide outside China rose to 6,745 dollars per kilogram. Neodymium-praseodymium, by contrast, traded outside China at about 106.5 dollars per kilogram in September, below the 110-dollar floor, as producers redirected volume into an oversupplied ex-China light market (Benchmark 2026). The IEA likewise reports that gallium and heavy rare earths in Europe have been about five times Chinese domestic prices (IEA 2026).

The split matters because it shows what policy has built and what it has not. Light rare earth oxide and some metal capacity outside China is now sufficient to push prices below a government floor, which is the outcome the floor was designed for. Heavy separation has barely begun. The IEA expects diversified rare earth refining capacity by 2035 to equal about two-thirds of expected mined supply, but planned magnet production only one-third, so the conversion step remains the shortest part of the chain.

Malaysia illustrates the middle position of a processing state. Lynas operates its separation plant at Gebeng and has been developing heavy rare earth separation there, while Malaysia maintains a ban on exports of unprocessed rare earths and, in August 2026, was reported to be weighing conditional easing tied to investment and technology transfer (Caixin 2026). For a country in that position, processing capacity is a bargaining asset that attracts competing offers from the United States and China, and the choice of whose technology and offtake to accept becomes a foreign-policy decision.

The strongest argument against the thesis is that leverage decays when it is used. Controls tripled in three years have accelerated Western investment: public finance commitments in advanced economies reached about 65 billion dollars in 2025, more than four times the 2023 level, and rare earth refining concentration declined modestly in the IEA's latest data. Stockpiling is cheap by the standard of the risk, since the net annual cost of holding stocks of eleven high-risk materials outside the dominant supplier is estimated at under 900 million dollars. Recycling could lift average recycling rates for key energy minerals from about 10 percent to close to 20 percent by 2040, and substitution in motor design reduces exposure at the margin (IEA 2026). The effect is real, but it favours the buyer slowly and unevenly, which leaves a window of years in which the dependence is intact.

Three paths follow. In managed bifurcation, allied governments sign a binding price and trade mechanism, heavy separation and magnet plants reach commercial scale by 2028 to 2030, and two price tiers settle at a stable gap. In licensed dependence, Beijing keeps its regime in place as a dial, Western projects remain subsidised and sub-scale, and users accept a premium for access. In shock absorption, stockpiles, recycling, and substitution blunt the leverage before capacity does. The signposts are observable: whether China extends or lets lapse the suspension beyond 10 January 2027, whether the Pentagon magnet cutoff is enforced or waived, whether a plurilateral agreement is signed, whether the ex-China dysprosium premium narrows, whether MP Materials' 10X plant commissions on its 2028 schedule, and whether Alcoa's gallium output reaches the market as planned.

The comparative record supports a selective conclusion. Diversification succeeds where the whole chain is financed together, with a price floor that makes refining bankable, offtake that guarantees a buyer, and a downstream magnet or component plant that creates the demand. It stalls where mines are opened without refiners, or refiners are built without buyers. By 2030 the light elements are likely to be partly diversified, while heavy rare earths, gallium, graphite, and parts of the battery chain are likely to remain dependent whatever policy does.

For governments, the priority is coordination across the chain and enough credible fiscal backing that floors survive budget cycles, together with binding arrangements among allies that spread the cost. For firms, the discipline is to map exposure to the stage, not the mineral: a supplier of magnets or motors that depends on a Chinese conversion step has the same exposure whatever the origin of its ore, and the heavy elements deserve separate treatment from the light. For investors, floors and offtake agreements make projects financeable but shift the risk to the fiscal commitment behind them, and the winners will be producers at the scarce link, heavy separation and magnet manufacturing, not those adding capacity where prices already sit below the floor.

Critical minerals are a coordination problem that looks like a resource problem. The country that controls the chemistry controls the chain, and the strategic task for everyone else is to rebuild the missing middle at a pace set by capital, skills, and politics, not by the discovery of ore.