Gallium and germanium under export control: licence risk

Gallium and germanium sit inside compound semiconductors, optical and infrared components, and specialised electronics. They are produced in small quantities as by-products of larger metallurgical processes, which means their supply does not respond to demand the way a primary metal's does. Where they come from and how they are made is Pantheon's subject, not this one. What concerns capital is what happens to a contract when the state inserts itself between a willing seller and a willing buyer.

A licensing regime is not a ban. The material may still be sold, shipped, insured and paid for. What changes is that a shipment now requires an authorisation granted by an authority that is not a party to the contract, on a timetable it does not publish, against criteria it is not obliged to explain, and with an outcome it can revisit. Four commercial consequences follow, and they are the whole of the page.

Delivery becomes conditional. Every downstream commitment resting on that delivery inherits the condition. A contract to supply a component, a project schedule, a covenant tied to a commissioning date — all of them now sit behind an administrative decision.

Lead time becomes a distribution rather than a number. Procurement planning built on a delivery date has to be rebuilt on a range whose upper tail is open-ended, because a licence application has no guaranteed conclusion date and a refusal has no fixed appeal.

The counterparty stops being able to promise. A seller that fully intends to perform cannot commit to an outcome it does not control. Any seller that does commit is either mispricing the risk or has not read its own regime, and both are reasons for caution rather than comfort.

Substitutability collapses at the exact moment it is needed. A controlled material's alternatives are qualified sources, not equivalent ones. Switching is a technical programme, and it starts from zero the day it becomes urgent.

The documentary discipline that follows is the same discipline described, in a much harsher form, on financing restricted materials: end use and end user become the gating items, the file is organised around the parties rather than the cargo, and it has to be complete before it is presented.

Two features of a licensing regime make it a delivery problem rather than a price problem, and a third makes it everyone's problem at once.

The decision is not the buyer's, and not the seller's. Neither commercial party can accelerate it, guarantee it, or convert it into an obligation. This is what distinguishes licence risk from ordinary supply risk: a supplier that is short can be pushed, penalised or replaced, and an authority cannot.

The timetable is the exposure, not the outcome. Buyers concentrate on the possibility of refusal, which is the visible risk. The costly case is more often approval that arrives late. A refusal is a bad answer received in time to act on; a six-month silence is the same commercial damage with none of the certainty, and it consumes the window in which alternatives could have been qualified.

Concentration converts an administrative decision into a supply event. Where the processing and refining capacity for a material sits in a small number of jurisdictions — as it does for both of these, and for the reasons set out on magnet rare earths and supply concentration — a policy decision in one capital is not a local disruption that buyers route around. It is the market. There is no alternative pool of committed capacity waiting to absorb displaced demand, because capacity in materials of this kind is built against long-term demand that already exists elsewhere. This is why buyers experience licensing as a step change rather than as friction: the whole of the accessible supply becomes conditional simultaneously.

The practical translation for anyone underwriting a project: a controlled input is not a cost-line risk, it is a completion risk. It belongs in the schedule analysis, next to the long-lead equipment and the interconnection date, not in the operating budget. The same reasoning applies to any concentrated, licence-dependent input, and it is why long-lead equipment procurement finance treats the order rather than the asset as the position being financed.

Buyers negotiating under a licensing regime routinely believe they have allocated licence risk when they have not. The gap is worth setting out precisely, because it is discovered at the worst possible time.

Force majeure usually does not do what parties assume. A conventional clause excuses a party from performing when an event outside its control prevents performance. Three things commonly defeat its application here. First, a licensing requirement that existed when the contract was signed is a known condition, not a supervening event — foreseeability is exactly what these clauses are drafted to exclude. Second, a refusal that follows from the seller's own application, its own documentation or its own end-user diligence is not obviously outside the seller's control, and that argument is available to the buyer. Third, and most important commercially: force majeure excuses performance, it does not deliver material. A successful claim means neither party is in breach. The buyer's line still stops.

What can genuinely be allocated is narrower, and has to be drafted for specifically.

There are only two structural responses to a licence-dependent input, and a serious procurement plan uses both, because each covers what the other cannot. Buffer stock buys time; qualification buys optionality. Neither buys supply.

Buffer stock is a working-capital position, and it should be underwritten as one. Holding inventory against a licensing interruption means capital committed to material that generates no return while it sits, plus storage, insurance, handling and — on some forms of these materials — controlled-storage requirements of their own. It converts a supply risk into a balance-sheet one, which is usually the right trade and is rarely recognised as a financing decision. Three points decide whether it is well done:

• Sizing is a function of qualification time, not of consumption. The correct buffer covers the period needed to bring an alternative source to production quality, not the period needed to feel comfortable. Sizing to consumption alone produces a stock that runs out precisely when the alternative programme is halfway through.
• Inventory is collateral of uneven quality. Material with a deep, transparent market and standard specifications supports a borrowing base. Small-market, high-purity material specified to one buyer's process does not, whatever it cost. Lenders discount it accordingly, and the discount is the real carrying cost.
• Stock has to be positioned on the right side of the control. Inventory held where it can itself become subject to an authorisation before it reaches the plant is not a buffer.

Qualification of an alternative source is a schedule and capital problem, not a purchasing one. Bringing a second source into a controlled process means sample material, analytical characterisation, process trials, tool and yield validation, and — where the end product is itself qualified into a customer's system — requalification at that level too. It consumes engineering capacity, production time on a line that would otherwise be making revenue, and in many cases a customer's consent. The programme runs in quarters and years, and the capital it consumes is spent before a single unit of alternative supply is bought.

Two consequences follow that buyers consistently underestimate. Qualification has to start before it is needed, because a programme begun on the day supply is interrupted delivers after the buffer is gone. And requalification is not free once a source is qualified — process changes, specification drift and customer-side approvals mean a qualified alternative decays if it is never actually bought from. A source qualified and then left dormant for years is a source that will need requalifying at the moment it is called on.

Where a buyer's own product is sold into long-term contracts, the qualification programme is itself financeable against those contracts on the same logic that makes any contracted revenue stream bankable — see offtake. What is not financeable is a qualification programme with no contracted demand behind it and no evidence that the customer will accept the alternative source.

How positions in licence-controlled materials go wrong:

• The regime was treated as a price event. Procurement responded by hedging or by renegotiating price, on the assumption that supply would clear at some level. Licensing does not clear at a price; it clears at an authorisation, and no cost line covers a shipment that does not move.
• Force majeure was assumed to be the answer. Discovered at the point of claim to be either unavailable, because the licensing requirement predated the contract, or useless, because being excused from performance was never the buyer's problem. This is the single most common contractual failure on these materials.
• The buffer was sized to consumption rather than to qualification time. Inventory that comfortably covers ordinary operations and runs out well before an alternative source can be brought to production quality. The buffer's job is to outlast the programme, and if it cannot, it is buying an interval that ends in the same place.
• The qualification programme started on the day of the interruption. By which point the engineering capacity, the sample material and the customer's attention are all being competed for by every other buyer in the same position.
• A qualified alternative had gone stale. Approved years earlier, never purchased from, and requiring requalification precisely when the primary route closed. A dormant qualification is an intention, not a position.
• Allocation was not contemplated. The regime did not refuse; supply was simply cut across a seller's customer base, at a level too low to run on and too high to trigger any contractual remedy. Contracts drafted around refusal and delay say nothing about partial performance.
• Downstream and upstream relief did not match. The buyer took a supply contract with generous suspension provisions and sold into customer contracts with none, leaving it exposed on both sides of a single event.
• The project's schedule never showed the input. Completion analysis covered equipment, construction and interconnection, and treated a licence-dependent consumable as a procurement detail. Lenders test schedules by looking for exactly this omission, in the same way they test a cluster's dependencies on what lenders underwrite on a GPU cluster.
• The end-use documentation was the buyer's own failure. Of everything on this list, the licensing outcome most within the buyer's control is the quality of the end-user and end-use evidence it provides to the party making the application. It is also the one most often handled as an administrative formality.

Continuum structures and arranges financing around procurement positions of this kind — inventory and buffer-stock facilities, qualification programmes supported by contracted demand, and the working-capital consequences of a conditional supply chain — and coordinates the parties to them. It does not trade, source, broker, take title to or hold materials, is not a bank, a broker-dealer or a direct lender, and does not hold client funds. It does not advise on export-control classification or licensing, which is a matter for counsel and for the competent authority. What a material is and where it comes from is covered by Pantheon, not here.