Rare Earths, Mining and Refining

Gallium and germanium are not mined. They are recovered, at trace concentration, from bauxite and zinc ore that were dug up to make aluminium and zinc. Neither appears on a finished chip's bill of materials, yet both sit, refined to one part in ten million or finer, underneath a large share of the radio-frequency amplifiers, 5G base stations, and fibre-optic links in service today. Electronics supply chains do not start at the wafer fab. They start at a refinery gate that decides whether the fab can operate at all.

In this piece I follow the ESDM supply chain to that starting point. I set out the five minerals that recur across the book, show why China's position is stronger in refining than in mining, walk through the export-control escalation from 2023 to 2026, examine how far visibility survives up the supplier tiers, and end with India's exposure and its policy response. I keep the ranges the sources give me as ranges, because on several of these numbers the sources disagree, and picking one would be a false precision.

Five minerals that electronics cannot substitute away

A handful of minerals recur across semiconductor, display, and power-electronics supply chains with a shared structural feature. None is abundant in the way iron or aluminium is. Several are not mined as primary ores at all. Every one needs a chemical refining step, often to parts-per-billion or parts-per-trillion purity, before it can enter an electronics process.

Gallium

Gallium is recovered as a trace byproduct of the Bayer process, the alkaline route that converts bauxite into alumina, at about 0.005% within the process stream. Zone refining and fractional distillation then take it to the required purity. Commercial-grade gallium at 4N (99.99%) suffices for LED lighting and solar alloys. Semiconductor-grade gallium at 6N to 7N is mandatory for gallium arsenide and gallium nitride epitaxial growth, where a single part-per-billion iron impurity can cause electrical leakage in an RF amplifier. Global annual gallium production is estimated at roughly 550 to 650 metric tons.

Germanium

Germanium is likewise a byproduct, of zinc ore processing, typically refined to 6N. Its main electronics uses are silicon-germanium heterojunction bipolar transistors in high-speed RF and mixed-signal circuits, fibre-optic components, infrared night-vision optics, and high-efficiency solar cells for space. The application set is narrower than gallium's but concentrated in defence, telecom infrastructure, and satellites.

Rare earths

Rare earth elements are the exception among the group in that they are primary-mined, from bastnasite, monazite sands, xenotime, and ion-adsorption clays. But their electronics relevance sits almost wholly in what happens after extraction. An ore body typically contains fifteen or more elements bound together, useless individually until separated. Neodymium, praseodymium, and dysprosium go into NdFeB permanent magnets in electric motors, hard disk drives, and precision robotics, including the linear motors in SMT pick-and-place machines. Europium and terbium go into phosphors. Cerium oxide is a key input to the chemical-mechanical planarisation slurries used to polish wafers. Dysprosium and terbium, the heavy rare earths needed for magnets that keep performing at high temperature, are refined almost exclusively in China.

Lithium

Lithium's ore is geographically diverse. Roughly 58% of proven reserves sit in the Lithium Triangle of Chile, Argentina, and Bolivia, with Australia a major hard-rock source. Mining is not China-concentrated. What follows mining is: converting spodumene concentrate or brine into battery-grade carbonate or hydroxide is a separate, chemically intensive step, and that midstream capacity is where China's position concentrates. Lithium's ESDM relevance is in battery cells for vehicles, storage, and the backup power inside data centres and telecom sites.

Tungsten

Tungsten, from wolframite and scheelite, is refined to 6N mainly as a precursor for tungsten hexafluoride gas, used to deposit the contact-via plugs that connect a transistor to its interconnect stack. Unlike the others, tungsten is concentrated in China at the mining stage, not only downstream.

Electronic-grade polysilicon carries a note of its own. The figure most often cited for China's total polysilicon capacity ranges from roughly 75% to 94.7%, depending on source and vintage. I flag that as a genuine cross-source disagreement rather than resolving it. But nearly all of that capacity, whichever figure is right, is solar-grade. Semiconductor-grade polysilicon is a separate, much smaller oligopoly, about 50,000 tonnes of 2025 demand against roughly 1.38 million tonnes of solar-grade demand, concentrated in Germany, the United States, South Korea, and Japan. That is why electronic-grade silicon carries less China-concentration risk than the solar chain that shares its feedstock.

The chokepoint is downstream of the mine

The single most important structural fact here is easy to state and easy to misread. China's dominance across this mineral set is concentrated far more heavily in refining and chemical processing than in raw mining. For several materials the two figures diverge sharply enough that reading "China controls mineral X" as a mining statement leads to the wrong diversification strategy. A mine in Australia or a brine field in Argentina does not reduce exposure to a chokepoint that sits three process steps downstream.

Rare earths are the clearest illustration. China accounted for about 69.2% of global rare earth mine production in 2024 to 2025, a large share, but one with real ex-China alternatives already operating in Australia, the United States, and elsewhere. Refining and separation is a different picture. Estimates put China's share of global refining capacity between 85% and as high as 99.9%, with acute concentration in heavy rare earths such as dysprosium, terbium, and yttrium. Lynas's Malaysian facility became the first commercial heavy rare earth separator outside China only in May 2025, and remains small relative to global demand. The 16 to 30 percentage-point gap between mining and refining share is the argument, in miniature, for why critical minerals policy cannot stop at securing ore.

Lithium follows the same pattern in a less visible form. Australia and the Lithium Triangle hold most proven reserves, none of it in China. But China is estimated to hold roughly 65% to 81% of midstream conversion capacity. The sources disagree materially between the low and high figures, and I keep the range. A lithium chain that looks geographically secure at the mine can still run through a single-country chemical bottleneck.

Gallium and germanium sit at the extreme end because neither has a mine of its own. For gallium the sources really do differ. The USGS places China's share of primary, low-purity crude gallium above 98%, while a Stimson Center and Fastmarkets synthesis puts refined gallium closer to 80%. The gap probably reflects crude output at the point of byproduct recovery against value-added refined product downstream, where 6N-grade gallium draws additional, smaller capacity from Japan, Germany, Slovakia, and Kazakhstan. But that is my reading. I report 80% to 98%. Germanium's sources agree better, at about 60% to 70%, with the balance held by Teck Resources in Canada, Umicore in Belgium, and smaller US and Finnish capacity.

Tungsten is the outlier. About 80% of world mine production is in China, so diversification has to start at the mine. For gallium or rare earths, an alternative effort can in principle begin at the refinery, using feedstock already available elsewhere.

The export-control escalation, 2023 to 2026

China's use of export licensing as a supply-chain lever did not begin with one dramatic action. It escalated in discrete steps, each adding a mineral or tightening a control, and each followed by measurable price and shipment disruption abroad.

The sequence began in August 2023, when China's Ministry of Commerce mandated individual export licences for gallium and germanium compounds, including Ga2O3, GaAs, GaN, GeO2, and GeSi, with each application subject to a 30 to 60 day end-use audit. Antimony followed in August 2024. By December 2024 exports of gallium, germanium, and antimony to the United States were effectively banned, a measure that remained suspended into 2026 amid trade negotiations. The antimony ban alone produced a reported 97% drop in shipments to the US and roughly a 200% price rise.

Tungsten was added in February 2025, and Chinese tungsten export volumes fell a reported 13.75% between January and September 2025 against the same period of 2024. April 2025 brought the most consequential step: controls on seven heavy rare earths and their derivatives, which unlike several others was never suspended. In October 2025 the controls widened by five more rare earths and an extraterritorial "50% rule" reaching foreign-made products above a defined Chinese rare earth content, with a truce provision suspending that expansion until November 10, 2026. An updated Export Licensing Catalogue took effect on January 1, 2026, adding samarium, gadolinium, and lutetium compounds, silver, and other materials.

What the price data shows

The direction is consistent even where magnitudes vary by source. Reported price spikes in Europe during 2025 ran as high as 365% for gallium and roughly 400% for germanium, against unwrapped gallium exports reported near zero through much of the year. Across the affected set, industry synthesis places reported spikes at roughly 200% to 600% depending on mineral and window. Approval rates for European buyers' licence applications were reported below 25% in certain sectors at points. Practical disruption ran well beyond what the headline price suggests.

Three features stand out to me when I read the sequence end to end. Each successive control targets a mineral where China's refining share is already dominant, which fits the argument that refining, not ore, is the leverage. The escalation has been managed more than undone: individual bans are suspended or truced, but the licensing infrastructure, the application requirement, the audit, and the extraterritorial reach, has only grown. And the April 2025 heavy rare earth controls are qualitatively different from the earlier ones, since the others each came paired with a suspension or partial reversal and this one did not. A 30 to 60 day audit is not a tariff. It is a decision, renewed shipment by shipment, about which foreign buyers keep operating.

How visibility fades up the supplier tiers

Mineral risk is worse than it looks because most buyers cannot see it. A Tier 1 supplier is one with a direct commercial relationship to the brand owner. Tier 2 supplies Tier 1, Tier 3 supplies Tier 2, and so on upstream to extraction. The OEM holds a contract, service levels, and audit rights over Tier 1. Tier 1 picks its own suppliers under terms the OEM is not party to, and a component maker's chemical and material sources are proprietary information it has every incentive to protect. By Tier 3 or 4, such as a specialty chemical plant or a mineral refiner, most OEMs do not know the entity exists.

A small labelling trap deserves mention. My own two source documents number tiers in opposite directions. One runs from Tier 0 at the OEM up to Tier 4 at mining and refining. The other runs from Tier 5 at raw mineral extraction down to Tier 1 at box-build and OEM. Anyone reading across sources has to check which convention is in use, or they will misplace a chokepoint by a full layer.

Scale compounds the problem. A 2020 Global Semiconductor Alliance and Accenture study found chips and components can cross international borders more than 70 times and travel over 25,000 miles before reaching a consumer, on a just-in-time model with minimal buffer. No single actor holds an end-to-end map. Each tier sees at most one layer up and one down.

The worked trace above stacks three single-supplier chokepoints on top of each other for one finished module. The substrate row is the instructive one: three named suppliers across two countries look diversified, until you go one input deeper and find a single Japanese chemical company with an estimated 99% to 100% share of the resin film beneath all three. The same logic applies to minerals. A diversified-looking Tier 1 base can sit on a single refiner two tiers down.

The lead-time arithmetic in the source is also worth carrying, with its caveat. Stacking about 6 weeks of refining, 14 of wafer fabrication, 10 of substrate production, 6 of packaging, and 2 of assembly gives roughly 38 weeks from ore to shipped product, with a plausible range of 26 to 42. That is my repository's own synthesis of stage medians, not a published industry benchmark. Its point is directional: a refinery disruption takes months to reach the shelf, which is exactly why it arrives without warning.

Two remedies follow the classification. A single-source dependency, where no qualified alternative exists, cannot be dual-sourced, so the levers are accepting exposure, securing priority through long-horizon commitments, or funding a substitute. A sole-source dependency, chosen by contract, can be dual-sourced. Mineral refining is closer to the first category for heavy rare earths and gallium, which is why stockpiles, recycling, and new separation plants matter more than a second purchase order.

India's exposure and its response

India sits near the most exposed position a large ESDM-aspirant economy can occupy. USGS-sourced data puts its import reliance for primary electronic-grade gallium, germanium, indium, and high-purity polysilicon at 100%. It imports roughly 93% of its rare earth permanent magnets from China as of FY2024 to 2025, with reported stockpiles for only two to three weeks of disruption.

The mining gap is not mainly geology. India holds an estimated 6.9 million tonnes of reported critical mineral reserves but mines only about 2,900 tonnes a year. The sources point to unresolved exploration data, permitting friction, and a historical absence of downstream demand. The clearest illustration is the Salal-Haimana lithium deposit in Jammu and Kashmir, inferred at 5.9 million tonnes of ore, which failed two auction tenders in 2023 and 2024, attributed to thin exploration data and to the deposit being clay-hosted rather than the more proven hard-rock or brine types investors know.

Midstream capability is nascent but not absent. IREL operates a Rare Earth Extraction Plant at Chatrapur, Odisha, and a refining unit at Aluva, Kerala, with about 6 lakh tonnes per annum of processing capacity across ilmenite, rutile, zircon, sillimanite, and garnet. That base has not yet translated into commercial-scale separation of dysprosium and terbium, which drive the 93% figure. KABIL, the overseas acquisition vehicle, signed a lithium exploration agreement in Catamarca, Argentina in January 2024 and holds a 2022 memorandum with Australia's Critical Minerals Office.

The Mission's outlay of ₹34,300 crore comprises ₹16,300 crore of direct government spending and an expected ₹18,000 crore from public-sector undertakings and others. Its scope spans exploration, mining, beneficiation, processing, and recovery from end-of-life products and e-waste, which I think is the right instinct: it does not stop at the mine.

Two gaps are worth naming. My research finds no confirmed programmatic link between the Mission and the India Semiconductor Mission or SPECS, so the mechanism by which mineral output would qualify a fab or OSAT project for incentives is unsourced. India-specific customs and logistics lead-time data for critical-mineral imports is also unsourced. The evidence documents policy in detail but not the transaction-level friction a buyer would meet.

What to take away

The practical rule I take from all this is to match the lever to the location of the chokepoint. A new mine changes where ore comes from. It does not change who can refine it to wafer-grade or magnet-grade purity, and that capability is what has been slowest to replicate outside a handful of emerging facilities. Anyone planning supply around these materials should ask where the refinery is, not just where the mine is.