Magnet rare earths: financing under supply concentration

Neodymium, praseodymium, dysprosium and terbium end up inside the permanent magnets that sit in motors, generators, drives, actuators and cooling equipment. A buyer of these materials almost never buys ore. It buys a separated oxide, a metal, an alloy, or — increasingly — a finished magnet, and each of those is a manufactured product with a specification behind it.

That distinction decides who holds pricing power, and it is the single most misunderstood feature of the market.

Rare earths are not rare, and mining is not the constraint. They occur widely. What is difficult, capital-intensive and environmentally regulated is separating a mixed concentrate into individual elements at the purity a magnet requires, and then converting those into metal and alloy. The value added between the mine gate and the magnet is where the margin, the technical barrier and the leverage live.

A mine without a route to separation sells into a market with very few buyers. A concentrate producer's product is an input to a small number of processing facilities. Its commercial position is therefore closer to a supplier negotiating with a handful of customers than to a producer selling into a market. This is why mining announcements move headlines and change very little about supply.

The buyer's real counterparty is the processor, wherever the ore came from. A buyer contracting for magnet-grade material is dependent on the processing step regardless of the mine's flag. Diversifying the mine while remaining dependent on the same separation capacity diversifies nothing that matters, and a supply chain map drawn at the mine level will look robust while the position underneath it has not moved.

Price formation happens where the processing is. Where a step is concentrated, the references a contract uses to price it — assessments, indices, published quotations — are observations of transactions inside that concentrated market. An offtake that fixes volume and floats price against such a reference has not escaped the concentration; it has indexed itself to it.

The honest way to describe the position is that the buyer is exposed to a manufacturing bottleneck, not to a resource. Manufacturing bottlenecks are solved by building capacity, which takes capital and years, and by qualifying it, which takes more.

Once the chokepoint is correctly located, several things that look like separate risks turn out to be one risk seen from different angles.

Capacity is lumpy and slow. Separation and metallisation capacity is built in large, capital-intensive increments, against permits that are difficult to obtain because the processes carry real environmental burdens. It does not respond to a demand signal within a project's development cycle. A buyer discovering a shortage cannot induce capacity into existence on any timeline that helps it.

Concentration is self-reinforcing. Where most of the world's separation runs in one place, that is also where the process knowledge, the trained workforce, the equipment supply chain and the downstream magnet manufacturing sit. A new entrant elsewhere is not simply building a plant; it is rebuilding an ecosystem, and its first output is more expensive than the incumbent's while it does so. That cost gap is the reason non-dominant capacity struggles to reach a final investment decision without contracted demand or policy support behind it — which is itself a financing question rather than a technical one.

Policy risk lands on the processing step. Because that is where the leverage is, that is where any intervention is applied — and where a licensing regime, if one arrives, will bite hardest and fastest. The mechanics of that are on gallium and germanium under export control, and they apply here without modification. The difference is only that concentration in the magnet elements is structural rather than incidental: it would be the dominant commercial fact about these materials even in a world with no export controls at all.

The exposure is usually invisible at the point it is taken. A project buys motors, drives or generators. Nobody in that procurement conversation is buying dysprosium. The dependency is two or three tiers down a supply chain nobody has mapped, and it surfaces as an equipment delivery date slipping for reasons the equipment vendor describes vaguely. A sponsor that cannot say which of its equipment packages contain magnet material, and where that material is processed, has an unpriced schedule exposure — the same class of exposure that long-lead equipment procurement finance exists to make visible on transformers and switchgear.

Long-term contracting is the standard answer to a concentrated input, and it is the right instinct. It is also routinely expected to do more than it can. A contract binds a counterparty; it does not create capacity, and it cannot bind an authority.

What a well-drafted offtake genuinely delivers is priority and predictability against a specific seller: volume commitments, a pricing mechanism, delivery obligations and a remedy if they are missed. On a liquid commodity that is close to security of supply, because a failed seller can be replaced from the market. Where processing is concentrated, three things behave differently — and the general mechanics of the instrument are on offtake, which this page assumes rather than restates.

Damages are not material. A remedy that pays the buyer for non-delivery is valuable to a trader and nearly worthless to a manufacturer whose line has stopped. Where there is no alternative source to buy from, the liquidated damages clause is a transfer of money, not a solution, and sizing it larger does not change that.

A commitment can be defeated upstream. A seller that cannot obtain an authorisation, cannot obtain feedstock, or is subject to an allocation across its own customers may be excused or may simply perform partially. The buyer's contract is with the seller, and the seller's constraint is with someone else.

Take-or-pay cuts both ways, and here it usually favours the seller. In a concentrated market the party with capacity has little difficulty selling it, so a minimum-volume obligation transfers demand risk to the buyer without transferring much supply risk in return — see take-or-pay for the mechanism. The exception is the case where it does real work: a take-or-pay given to a NEW, non-dominant processor is what makes that facility financeable, and there the buyer is deliberately paying for optionality it hopes never to need.

The implication for a buyer with real exposure is that contracting has to be done at the tier where the constraint sits. An offtake with an equipment vendor secures equipment. It does not secure the magnet material inside it unless the vendor has secured that itself, and asking the vendor to evidence that — rather than assert it — is the diligence step most often skipped.

There are two ways out of a concentrated dependency and both are slow. The first is to bring an alternative source into production and into qualification. The second is to need less of the material. Neither is available on the timescale of a crisis, which is why both have to be started outside one.

Qualifying a non-dominant source is a capital and schedule programme. Magnet material is specified into a motor or a generator whose performance, thermal behaviour and warranty depend on it. Changing where it comes from means characterisation, production trials, component testing, and approval by whoever holds the equipment warranty — often an OEM rather than the buyer. Capital is spent, and engineering capacity consumed, well before any alternative material is delivered.

How that programme is financed depends almost entirely on whether contracted demand stands behind it. A qualification programme with binding volume commitments from credible buyers is an identifiable position with a revenue path, and it can be structured as one. The same programme run on the expectation that demand will appear is a research budget, and it should be funded as one rather than presented as infrastructure.

Substitution is partial, slow, and rarely free. The realistic options reduce exposure rather than remove it — designs that use less of the heaviest and most concentrated elements, motor architectures that avoid permanent magnets at some cost in efficiency, weight or size, and recovered material from manufacturing scrap and end-of-life equipment. Each is a genuine engineering answer. Each also takes a product cycle to implement, applies to new designs rather than to installed equipment, and often trades performance for independence. Recycling in particular is a structurally sound long-term answer whose near-term supply is limited by how much equipment has actually reached end of life. Presented to a lender as a mitigant for a schedule risk this year, none of them qualifies.

What a lender tests when a project's schedule depends on a concentrated input is narrower than sponsors expect, and it is almost entirely about evidence rather than intention.

How projects and procurement positions fail on magnet materials:

• The supply chain was mapped at the mine. Diversified ore sources feeding the same separation capacity, presented as resilience. The map looks robust and the position has not moved, because the chokepoint was never at the mine.
• The dependency was never identified at all. The project bought motors, drives and generators from reputable vendors, and nobody asked what was inside them or where it was processed. The exposure surfaces as an equipment delay attributed to unspecified supply-chain conditions, at which point there is nothing to negotiate.
• An offtake was mistaken for security of supply. Volume contracted with a counterparty that cannot obtain its own feedstock, cannot obtain an authorisation, or allocates pro rata across its customers. The buyer holds a claim for damages and no material.
• Damages were sized to the contract rather than to the consequence. A remedy that compensates for a missed cargo against a delay that costs a commissioning date. On any schedule-critical input the two numbers are unrelated.
• A take-or-pay was given to a party that did not need it. Demand risk transferred to the buyer with no reciprocal supply security, because in a concentrated market the seller had no difficulty placing that volume elsewhere. The instrument does real work when it underwrites new, non-dominant capacity; given to an incumbent it is a one-way concession.
• The second source was identified but never qualified. A name on a slide, no sample material characterised, no production trial run, no OEM approval sought. It becomes a supplier at the point it becomes urgent, which is the point at which every other buyer in the market is calling it.
• A qualified source went stale. Approved once, never purchased from, and requiring requalification years later when it is finally called on — with specifications, processes and approvals all having moved in the interval.
• Substitution was presented as a mitigant on the wrong timescale. Redesign, alternative motor architectures and recycled material are real answers over a product cycle. Offered to a credit committee as cover for a delivery risk inside the construction period, they are an admission that the risk is unmitigated.
• The qualification programme was funded as if it were infrastructure. Capital raised against a facility that had no binding demand behind it, on the expectation that a policy environment favouring alternative supply would produce buyers. Where contracted demand exists, this is a financeable position; where it does not, it is a research budget wearing a project's clothes.

Continuum structures and arranges financing around procurement and qualification positions of this kind, and coordinates the parties to them. It does not trade, source, broker, take title to or hold materials, does not operate or own processing capacity, is not a bank, a broker-dealer or a direct lender, and does not hold client funds. Material properties, processing routes and magnet specification are covered by Pantheon, not here.