Rare earth is not rare.
The processing is.

A high-performance permanent magnet is mostly iron. The part that makes it a magnet worth fighting over is neodymium and praseodymium, usually written NdPr, with dysprosium or terbium added so the magnet keeps working when a motor gets hot. Those four elements are the whole story. Everything on this site exists to move them from rock to motor.

The chain has seven links, and each one is a different business with different physics, different capital intensity and a different set of companies.

Rare earths are not geologically rare. Deposits exist on every continent and several are larger or richer than anything China mines. The concentration is downstream of the rock.

Read the three numbers below in order. The share rises as you move down the chain, and it is effectively total at the heavy separation step, the one that supplies the dysprosium and terbium a hot motor needs.

That is why a mine on its own solves nothing, and why the interesting question for anyone allocating capital is not who owns a deposit but who can turn an oxide into a magnet without asking permission.

Export licensing has split the market in half. The same oxide now trades at a Chinese domestic price and a much higher ex-China price, and the gap is widest exactly where substitution is hardest.

Two markets, moving independently, for identical material.

That yttrium number is not a typo. Benchmark only began assessing yttrium in March, and the ratio reflects how little of it clears outside China at any price.

A bifurcated price is not a temporary dislocation to trade around. It is the economic basis of every non-Chinese project on this map, and it is why government price floors showed up when they did. More on that on the policy page.

Magnet demand used to be an EV and wind story. It now has a third driver. AI infrastructure, the energy transition and Western defence modernisation are all pulling on the same elements at the same time, and none of them is slowing down to make room for the others.

The physical intensities are worth holding in your head, because they explain why the arguments about substitution are really arguments about which application. An EV traction motor carries one to three kilograms of NdFeB. An offshore wind turbine carries roughly 600. An F-35 is estimated at 417. Data centres are a smaller number per unit and an enormous one in aggregate.

Against that, a new rare earth mine takes ten to fifteen years from discovery to production and a new separation plant five to eight. The demand curves are a sprint and the supply response is not, which is the entire reason recycling, unconventional feedstocks and substitution are interesting at all. Each pulls supply forward by skipping a step. Each introduces a different constraint in exchange: feedstock control, resource economics, or performance qualification.

Demand framing and project timelines drawn from Collide Capital's August 2026 note on the critical mineral and rare earth crunch, which in turn cites IEA, USGS and Goldman Sachs. Magnet intensities are widely reported figures rather than measurements taken here.

The map puts every organisation on one plate, with the companies building two or more links of the chain in the centre. Each centre tile carries a seven-segment strip showing which links it owns, which it has secured through offtake or a partner, and where the gaps are. The gaps are where the partnerships live.

Economics is a cost model you can break: fifteen assumptions, four plant archetypes, and a fully-loaded cost per kilogram that falls out the other end. Policy tracks the instruments rather than the jurisdictions. Method publishes what this dataset does not yet know, including the share of records still unverified.