What is tenor?

Tenor is the length of a financing: the period from first drawdown to final maturity for a facility, or the committed term for a lease. It answers a question distinct from how much and how fast — it is the duration of the commitment itself.

Three related measures are frequently confused with it and mean different things.

Amortisation profile is the schedule on which principal is repaid. A facility can have a long amortisation profile and a short tenor, which is precisely what a balloon or bullet repayment is: payments calculated as though the term were long, with the unpaid remainder falling due at a much earlier maturity. That structure lowers periodic payments and creates a refinancing event — the two are inseparable.

Average life is the weighted average time until principal is actually repaid. It captures what tenor alone misses: two facilities with identical maturities have very different exposure if one repays steadily and the other repays entirely at the end.

Economic life belongs to the asset, not to the money. It is how long the asset is worth operating, and it is the benchmark the other three should be measured against.

Tenor is also the input that determines how much a given cash flow supports. Debt service is spread over the term, so a longer tenor produces lower periodic payments and therefore a larger amount supportable at the same coverage — the arithmetic set out at [debt service coverage ratio](debt-service-coverage-ratio). This is why tenor is negotiated as hard as amount: it very often *is* amount.

Every financing sits at the intersection of three periods, and there are only a few ways they can be misaligned. Each has a characteristic failure.

The money outlasts the asset. Capital still outstanding against something no longer worth operating. The obligation continues; the thing that was supposed to service it does not. This is the failure that makes short-lived assets so difficult to finance on long terms, and it is why nobody writes a fifteen-year facility against a three-year asset regardless of how strong the sponsor is.

The money matures before the asset is finished earning. A perfectly good asset with capital falling due, requiring a refinancing. Refinancing risk is not a defect — most long-lived infrastructure is refinanced several times — but it is a real exposure, and it is at its worst when maturity falls in a market that has moved or at the same moment as everyone else's.

The contract is shorter than the money. The most common case in this sector and the most consequential. Capital is outstanding into a period where the revenue supporting it has ended and has not yet been replaced. Somebody is carrying the risk that the capacity recontracts, and at what price — the recontracting tail. Three things reduce it: renewal economics that favour the customer, capacity that can be recontracted to somebody else without moving anything, and repayment scheduled to complete inside the contracted period. The last is the only one that eliminates it, at the cost of a heavier repayment profile.

A fourth alignment is easy to overlook and fails hard: the layer beneath has to outlast the layer above. A compute contract cannot reliably run longer than the facility's power commitment, and the facility cannot outlast the site control beneath it. Power that expires first converts a performing asset into a relocation problem, and site control that expires first converts a building into a dispute.

The reason tenor is the organising question in this sector rather than a technical detail is visible as soon as [the four layers](the-data-center-capital-stack) are read as periods rather than as assets. Nothing about the table below is unusual in isolation — each row is an ordinary financing. It is the fact that all four sit at one address that creates the problem.

A single instrument spanning all four layers has one maturity, and one maturity cannot be right for durations that differ by an order of magnitude. It gets priced by whoever is least comfortable with the shortest-lived asset in it — so the accelerators set the terms for the land, the land makes the accelerators look better secured than they are, and every turnover of the compute becomes an event for the entire structure.

That is the whole argument in three sentences, and it is deliberately not developed further here. It is the position the rest of the site is built on, and it is made in full — with the mechanics of how the separation is actually executed — at [tenor mismatch: financing compute inside infrastructure](/compute/tenor-mismatch-compute-and-infrastructure).

What follows from it for this page is narrower and practical. Matching tenor to an asset means writing the money to the shorter of the asset's economic life and the contract behind it, and then deciding deliberately who carries whatever is left over. That leftover is a real position: the [residual value](residual-value) at the end of a lease, the recontracting tail after an [offtake](offtake) expires, the refinancing at a balloon maturity. Every structure contains at least one of them. The failure is not having one — it is having one nobody identified, priced, or agreed to hold.

The separation itself is executed through the ordinary machinery: a [special-purpose vehicle](special-purpose-vehicle) per layer so each can be held on its own terms, a [construction loan](construction-loan) taken out by capital matched to the completed asset, and lease or [sale-leaseback](sale-leaseback) structures where the asset's life is shorter than the building around it.

Continuum structures deals along those seams: establishing what each layer's real duration is, matching the arrangement to it, and making sure the leftover positions are held by parties that chose them.