What Actually Sizes Project Finance Debt? DSCR, Gearing and the Constraint Stack
A credit agreement filed in 2026 gives a useful answer to the question “what sizes the debt?”
Not one ratio.
The agreement caps the term loans at the lower of a stated dollar amount and a model-derived amount. The model-derived amount itself is then tested through a 20-year sculpt, separate P50 and P99 coverage tests for contracted cash flows, much higher coverage requirements for merchant cash flows, a severe-downside repayment case and a limit on how much of the loan can be supported by merchant revenue.
That is much closer to real project finance than the familiar shortcut:
Debt = MIN(75% of capex, DSCR-sized debt)
The shortcut is useful for teaching. It is not a universal description of a credit agreement.
A better way to think about sizing is:
Every lender test produces a maximum debt amount.
Final debt
=
MIN(
leverage / gearing limit,
P50 coverage sizing,
P99 coverage sizing,
merchant-supported debt limit,
severe-downside repayment limit,
facility / commitment cap,
any other closing sizing test
)
The negotiation is about the individual tests. The model's job is to tell you which one actually binds.
Start with the signed sizing criteria, not a market DSCR
A surprising amount of project-finance modelling starts with a cell called Target DSCR and then reverse-engineers the rest.
The credit agreement works the other way around.
The lender defines the eligible cash flows, production cases, coverage ratios, amortization period, merchant treatment, caps and sometimes whole-loan downside tests. Those terms together produce the debt amount.
A 2026 Big 5 refinancing agreement filed with the SEC is a good example. Its contracted fixed-price cash flows are sized to the lesser of a P50 case at 1.25x and a P99 ten-year case at 1.00x. Merchant cash is sized separately to the lesser of a P50 case at 1.80x and a P99 ten-year case at 1.40x. The loan must also fully repay under a severe downside case. A related senior facility limits merchant-supported debt to no more than 25% of the term loans.
None of those figures should be treated as “the market.” They are one executed financing.
What they demonstrate is the architecture: lenders can make multiple independent views of cash flow bind at the same time.
A DESRI financing agreement provides a different architecture. During the PPA term it requires 1.30x coverage on P50 and 1.00x on P99, then applies a 2.00x P50 ratio for up to five years after PPA expiry. Different project, different credit agreement, same principle: debt is sized by a set of contract-defined tests.
That is the right place to start.
Coverage sizing turns CFADS into allowable debt service
The basic DSCR identity is simple:
DSCR(t) = CFADS(t) / Debt Service(t)
If the sizing target in a period is 1.30x:
Maximum Debt Service(t)
=
CFADS(t) / 1.30
Once the allowable debt-service stream is known, the facility amount is the present value of that stream at the debt rate, subject to the amortization assumptions and other constraints.
For a fixed rate and no special amortization floor:
DSCR-sized debt
=
PV of [ CFADS(t) / Target DSCR(t) ]
The formula is straightforward.
The difficult input is CFADS.
CFADS is a credit-agreement definition
“Cash flow available for debt service” sounds like an accounting subtotal. It is not.
A filed DESRI agreement defines DSCR using Operating Cash Flow Available for Debt Service over a test period divided by Debt Service, and then contains specific treatment for items such as LC loans and unreimbursed obligations.
That is typical of the drafting problem: every facility has to decide what counts above debt service and what sits below it.
The points worth checking are transaction-specific:
- cash taxes;
- working-capital movements;
- maintenance capex;
- battery augmentation or major overhaul cost;
- reserve funding and releases;
- insurance proceeds;
- liquidated damages;
- deemed-generation payments;
- merchant hedge receipts and payments;
- extraordinary / one-off receipts;
- management fees and affiliate payments.
A sponsor can show a 1.30x DSCR and a lender can show 1.24x without anybody making a formula error. They may simply be using different CFADS definitions.
That is why the sizing sheet should have an explicit bridge from project cash flow to Lender CFADS.
P50 and P99 can be simultaneous sizing tests
Another oversimplification is to say “lenders size on P50” or “merchant debt is sized on P90.”
Actual agreements can require more than one production case at the same time.
The 2026 refinancing example mentioned above uses both P50 and P99 tests. A Greenbacker facility filed with the SEC similarly sizes investment-grade-offtaker projects using both a P50 DSCR test and a P99 DSCR test, with different ratios for non-investment-grade offtakers.
For an investment-grade offtaker, that agreement used:
P50 DSCR 1.25x
P99 DSCR 1.00x
For a non-investment-grade offtaker, it used:
P50 DSCR 1.40x
P99 DSCR 1.15x
Again, those are examples from a filed facility, not current market recommendations.
The conceptual point is more important than the exact numbers.
A P50 test asks how much debt the central production case can service with the required headroom.
A P99 test asks whether a severe production outcome still clears a lower coverage floor.
Both can bind.
So the model should calculate both debt capacities independently:
P50 Debt Capacity
P99 Debt Capacity
Final Coverage-Sized Debt = MIN(P50, P99)
If the model only reports one target coverage ratio, it may be hiding the actual closing constraint.
Gearing is a different kind of limit
A leverage or gearing test does not ask what the project's cash flow can repay.
It asks how much of the project cost the lender is willing to fund.
Simplified:
Gearing Debt
=
Eligible Project Cost × Maximum Gearing
If eligible project cost is $250 million and gearing is 75%:
Gearing Debt = $187.5m
That test protects a different concern.
Coverage is about repayment from cash flow.
Gearing is about sponsor capital at risk and the relationship between debt and invested cost or value.
Some financings use both. Others use different leverage constructs, advance rates, borrowing-base rules or absolute facility caps.
The modeller should therefore avoid turning “gearing” into a mandatory project-finance feature. It is one possible member of the sizing stack.
The crossover is still useful — if you label the assumptions
The old version of this article stated that the crossover between DSCR sizing and gearing was about 9.3% CFADS-to-cost.
That number is not a market fact.
It was the result of one illustrative term sheet.
The concept is still valuable.
Assume:
Project cost $250.0m
Maximum gearing 75%
Target DSCR 1.30x
Debt rate 6.5%
Amortization period 18 years
Flat annual CFADS X
The gearing cap is:
75% × $250m = $187.5m
For a flat CFADS profile, the DSCR-sized debt can be approximated using an annuity factor:
Debt
=
(CFADS / 1.30) × annuity factor
Set the two answers equal and solve for CFADS.
Using an 18-year annuity factor of about 10.43:
Crossover CFADS
≈
$187.5m × 1.30 / 10.43
≈
$23.4m
Against $250 million of cost:
Crossover CFADS / Cost
≈ 9.35%
Under those assumptions, a project below roughly 9.35% cash yield is DSCR-constrained and a project above it is gearing-constrained.
Change the debt rate, tenor, DSCR or gearing and the crossover changes.
The useful output is therefore not “9.35%.”
It is:
Project CFADS
Crossover CFADS
Headroom to constraint switch
A project sitting 2% above the crossover can flip from gearing-constrained to coverage-constrained on one modest downside assumption.
That regime change is worth seeing.
Binding constraint tells you what actually moves the debt
Suppose two projects both close with $180 million of debt.
Project A is gearing-constrained.
Project B is DSCR-constrained.
The headline leverage is identical. The economic response to the next assumption change is not.
If gearing binds
Improving revenue may improve equity returns and covenant headroom without increasing initial debt.
Reducing capex can reduce the gearing-sized debt if the percentage is applied to cost, even while the project becomes economically stronger.
A sponsor arguing for a better PPA price may therefore be arguing about equity value, not initial leverage.
If coverage binds
Improving CFADS can increase debt directly.
Reducing capex can lower the sponsor equity cheque without reducing debt, until another constraint becomes binding.
The best commercial negotiation can therefore be different depending on which test wins.
That is why a model should publish:
Constraint Max Debt
--------------------------- --------
Gearing $187.5m
P50 DSCR $181.2m
P99 DSCR $176.9m
Merchant cap $185.0m
Facility cap $190.0m
Final debt $176.9m
Binding constraint P99 DSCR
Second constraint P50 DSCR
Headroom to second test $4.3m
The second-best constraint matters too.
A sensitivity can switch the binding test before it changes the final debt materially.
Credit quality can change the sizing ratios
The Greenbacker filed facility is useful because it shows the same portfolio using different DSCR parameters depending on the off-taker's credit quality.
That is a reminder that DSCR is not simply a technology parameter.
The ratio is compensating for the lender's view of volatility and credit risk in the cash flow being relied on.
Other features can move the treatment too:
- fixed-price PPA vs market exposure;
- hedge tenor;
- basis / shape exposure;
- counterparty quality;
- curtailment;
- operating history;
- concentration;
- resource uncertainty;
- technology / availability risk.
The ratio is the output of the lender's risk allocation, not a universal project-finance constant.
Tenor is part of sizing, not a separate documentation question
Coverage debt is the present value of permitted debt service.
A longer amortization period therefore usually increases the amount of debt the same cash flow can support.
But the lender may not allow the debt to use the whole asset life.
The sizing period can instead be linked to:
- remaining PPA term;
- a percentage of remaining PPA term;
- a notional amortization period longer than legal maturity;
- a refinancing / mini-perm structure;
- an explicit merchant tail;
- a final balloon constraint.
The Greenbacker agreement, for example, sculpts over a period tied to the remaining PPA term and imposes caps on that period. Other filed agreements use a longer notional amortization period while legal maturity occurs earlier.
That distinction matters:
Legal maturity
≠
Notional amortization end
≠
PPA expiry
≠
Asset life
A model that uses one date for all four has hidden the refinancing structure.
A severe-downside test can bind even after the DSCR tests pass
The 2026 Big 5 agreement includes a whole-loan requirement that the debt fully repay under a severe downside case combining a P99 production case with a low historical power-price assumption and escalation.
That is conceptually different from a one-period DSCR test.
The lender is asking:
Even if the central sizing cases pass, does the entire debt still amortize under a coherent severe scenario?
This is why a debt-sizing model should not stop at a table of annual DSCRs.
It can also need a scenario repayment test:
Base lender case → size debt
P99 / downside case → test annual coverage
Severe scenario → prove full repayment by required date
A facility can pass the first two and fail the third.
Merchant-supported debt can have its own cap
A lender can give merchant revenue some credit and still prevent it from supporting too much of the loan.
One 2026 filed agreement does exactly that: merchant electricity cash flows cannot support more than 25% of the senior term loans.
Another financing requires merchant energy cash flows in the forecast but caps the aggregate term-loan amount supported by merchant cash at 25% of total term commitments.
This is a useful design because it separates two questions:
How much merchant cash flow do we model?
How much of the debt are we willing to let that cash flow support?
Those are not the same question.
Part 3 of this series takes the merchant block apart properly.
Sizing DSCR is not the same thing as covenant DSCR
A facility can use one coverage ratio to size debt and another to test distributions or default after closing.
A 2026 SolarBank filing, for example, reports a 1.60x quarterly DSCR requirement and a separate 1.20x Forward Contracted DSCR.
Those are covenant / maintenance concepts, not necessarily the original closing sizing ratio.
This distinction matters because people often say:
“The deal has a 1.30x DSCR.”
That sentence can mean:
- closing debt sizing target;
- minimum maintenance covenant;
- distribution lock-up threshold;
- default threshold;
- forward contracted test;
- historical trailing test.
The credit agreement can contain several of them.
The financial model should label them by purpose.
The model should make every sizing test visible
I would build the sizing page as a constraint register.
1. Gearing / leverage capacity
2. Contracted P50 coverage capacity
3. Contracted P99 coverage capacity
4. Merchant P50 capacity
5. Merchant P99 capacity
6. Merchant-supported debt cap
7. Severe-downside repayment capacity
8. Absolute facility / commitment cap
9. Other closing test, if applicable
FINAL DEBT = MIN(all applicable capacities)
For each line, publish:
- calculated maximum debt;
- headroom to final debt;
- whether the test binds;
- which assumptions drive it.
Then sensitivities become useful.
A price downside might move the merchant test.
A capex overrun might move gearing.
A worse resource case might move P99.
A longer tenor can move coverage without touching gearing.
The model tells the team which negotiation matters instead of presenting one opaque debt number.
A worked simplified case
Assumptions only:
Project cost $250.0m
Maximum gearing 75%
P50 CFADS $24.0m
P99 CFADS $21.0m
P50 target DSCR 1.30x
P99 target DSCR 1.05x
Debt rate 6.5%
Amortization period 18 years
Absolute commitment cap $190.0m
Assume, for simplicity, flat annual CFADS and an 18-year annuity factor of 10.43.
Gearing capacity
= 75% × 250.0
= $187.5m
P50 capacity
= (24.0 / 1.30) × 10.43
≈ $192.6m
P99 capacity
= (21.0 / 1.05) × 10.43
≈ $208.6m
Commitment cap
= $190.0m
Final debt:
MIN(187.5, 192.6, 208.6, 190.0)
= $187.5m
Gearing binds.
Now reduce P50 CFADS by 5%:
P50 CFADS = $22.8m
P50 debt
≈ (22.8 / 1.30) × 10.43
≈ $182.9m
The final debt falls to about $182.9 million and the binding test switches to P50 coverage.
One 5% operating downside changed the financing regime.
That is the output I want from a debt-sizing model.
Not merely the new debt amount.
The handoff to sculpting
Once the sizing case determines the maximum debt-service stream, the next problem is how that debt service becomes interest and principal over time.
That is sculpting.
And the clean version is less circular than many models make it.
Sources: SEC — 2026 Big 5 Refinancing Credit Agreement · SEC — DESRI Financing Agreement · SEC — Greenbacker Credit Agreement · SEC — 2026 SolarBank Filing