What is the debt service coverage ratio (DSCR)?

The debt service coverage ratio is cash available for debt service in a period, divided by the debt service — interest plus scheduled principal — due in that period.

A ratio of 1.0 means the project generates exactly what it owes, with nothing left over for anything to go wrong. Every required ratio therefore sits above 1.0, and the distance above it is headroom: how far cash flow can fall short of forecast before the obligation stops being met. How much headroom is required is a function of how uncertain the cash flow is — contracted revenue from a strong counterparty needs less, merchant revenue needs considerably more.

The numerator is where the substance is. "Cash available for debt service" is a defined term in a financing agreement, not an accounting standard, and what it includes decides the answer:

• Revenue, and specifically which revenue counts — contracted only, or contracted plus some haircut portion of merchant.
• Operating costs, including the ones that are easy to forget: site costs, insurance, management fees, and the cost of an operator.
• Maintenance capital expenditure. Whether major overhauls and replacements are deducted materially changes the ratio on any asset with real maintenance requirements, and every party to the negotiation knows it.
• Tax and working capital movements.

Two projects with identical economics can report different ratios purely on definitional choices. Reading the definition before reading the number is not pedantry; it is the only way the number means anything.

The most common misunderstanding is that the ratio is a test a project passes or fails after the amount has been decided. It is the other way round. The ratio decides the amount.

The logic runs backwards from cash flow:

1. Establish the cash the project will generate in each period, on assumptions a third party will accept rather than the sponsor's own case. 2. Divide by the required coverage ratio. What remains is the debt service the project can support in that period, with headroom intact. 3. From that supportable debt service, and given a repayment profile and a term, derive the amount of capital that can be advanced.

The consequence is the one every sponsor eventually meets: what the asset cost is not an input. A project can cost whatever it costs; the amount that can be advanced against it is set by what it earns. Where those two numbers diverge, the gap is equity, and no amount of arguing about the cost base closes it.

This is also why [offtake](offtake) dominates the conversation. The numerator is contracted cash flow, so improving the contract improves the sizing directly — a firm [take-or-pay](take-or-pay) obligation from a strong counterparty enters the calculation at close to full value, while uncontracted capacity is discounted heavily or excluded. Two identical facilities, one contracted and one not, support very different structures for reasons that have nothing to do with the buildings.

A related test, the loan life coverage ratio, applies the same idea across the whole remaining life of the facility rather than period by period, and catches profiles that look adequate in each individual period while failing over the term as a whole.

One number appears at several points in a financing agreement with different consequences attached. Confusing them is a routine source of alarm — a projection dipping below a distribution threshold is a very different event from breaching a default level.

The ratio is a general test, but three features of this sector determine where it binds.

The tail after the contract ends. Coverage has to hold in every period of the facility, including the periods after the offtake expires. Where the contract is shorter than the money, the cash flow in those later periods is a forecast, and it is discounted accordingly — which frequently means the facility is sized on the contracted period alone. The cleanest response is to repay within the contracted term, at the cost of a heavier profile. This is the [tenor](tenor) question expressed arithmetically, and it is usually the binding constraint rather than the headline ratio.

Which layer is generating the cash. [The four layers of a data center](the-data-center-capital-stack) produce cash flows of very different quality. A capacity payment on generation and a long lease on a shell are the kind of revenue this test was designed for. Compute revenue is discounted much harder, because the contracts are shorter, the counterparties are frequently younger, and the recovery behind them is weak — the underwriting is set out at [what lenders underwrite on a GPU cluster](/compute/what-lenders-underwrite-on-a-gpu-cluster).

Costs that are not fixed. Power is the dominant operating cost in this asset class, and where it is not contracted on terms matching the revenue, the denominator is stable while the numerator is not. A facility with fixed revenue and floating power cost is carrying a real exposure that shows up in coverage before it shows up anywhere else.

The build has no coverage at all. During construction the project generates nothing, so this test has nothing to measure and the [construction loan](construction-loan) is controlled by completion mechanics instead. Coverage becomes the governing test at the point the takeout arrives — which is why the ratio the completed asset is expected to produce is negotiated long before the asset exists.

Continuum structures deals so that each layer is arranged against the cash flow that genuinely supports it, rather than against a blended figure that flatters the weakest part of the stack.