# The Financeable Maturity Window

AI Infra Credit · September 22, 2026  
Public companion edition · September 24, 2026 · Original research and executable experiment dated September 22

## Finding

A longer repayment term can make the same operating plan financeable, but only inside a window. The short end is constrained by start-up equity and debt-service coverage. The long end can be constrained by cash flows after the customer contract ends. The price paid to the lender—and when that price is paid—moves both boundaries.

In the explicit illustration below, a 36-month loan fails the sponsor's funding budget and an assumed operating coverage test. At a 7% effective annual loan rate, a 42-month loan passes those screens and matures with the modeled customer contract. Extending to 48 months reduces initial equity further but leaves six debt-payment months exposed to weaker renewal cash flow.

When the lender's required scheduled-cash-flow yield rises to 8%, keeping a 7% loan rate and paying the difference upfront does not work at 42 months. Repricing the loan to 8%, with no upfront fee, does work in the same weak-tail scenario. This is a timing result, not free financing.

## The disclosed mechanism and the hypothetical experiment

DigitalOcean's September 10 Form 8-K describes equipment advances that may be requested until September 10, 2027, with monthly rent and full amortization by September 10, 2030. This motivates distinguishing the draw window from repayment life. The filing summary does not establish the annuity schedule, cash flows, coverage standard, equity budget or discount rates used here. We do not reconstruct DigitalOcean's leases or propose a trade in its securities. [Primary filing, Item 1.01](https://www.sec.gov/Archives/edgar/data/1582961/000110465926106704/tm2625205d1_8k.htm).

All following amounts are hypothetical **USD millions**. Month zero is the same equipment-funding date for every structure. This compares alternative terms for an identical physical plan, not borrowing early against borrowing late.

| Assumption | Value |
|---|---:|
| Equipment investment / debt funding | 100 / 90 |
| Months before first operating-cash period | 6 |
| Operating-cash periods | Months 7–54 |
| Customer-contract cash periods | Months 7–42 |
| Monthly cash available before debt service, during contract | 3 |
| Base / weak post-contract monthly cash | 3 / 1.8 |
| Maximum initial equity funding | 25 |
| Required monthly operating-cash coverage | 1.20x |
| Equity / unlevered-project reference discount rates | 12% / 12% |
| Base loan rate / lender discount hurdle | 7% / 7% |
| Reserve return | 0% |

Rates are annual effective, converted consistently to monthly rates. Each loan pays a level monthly amount from month one. Equity funds equipment cost not funded by debt, a reserve covering all pre-operating debt payments, and any upfront fee. The reserve is counted once and then spent on debt service, not charged to equity again.

Operating cash is an input after the operating costs, taxes and recurring investment the user wishes to include. It is not revenue. Monthly receipts are assumed available for that month's debt payment; collection lags, default, hedging, actual lease conventions, additional transaction costs and residual asset sales are not modeled. The assumed coverage test applies only after commissioning. It is not represented as an observed covenant.

The default policy distributes residual cash once operations begin. A separate retention case holds cash until debt repayment. Any additional support shown by the cash ledger is a **funding requirement**, not capital assumed available; a case requiring it fails the joint screen. Similarly, a calculated initial reserve is not proof that the equity needed to fund it exists.

## A two-sided participation test

Let P be debt, K equipment cost, N repayment months, a the effective annual loan rate and r=(1+a)^(1/12)-1. The level payment is:

```
A = P*r / (1 - (1+r)^(-N))         [A=P/N when r=0]
R = sum of debt payments before operations begin
E0_before_fee = K - P + R
```

For a given structure, let V_e(0) be equity NPV before the upfront fee, computed from the complete cash ledger at the chosen equity discount rate. Let h be the lender's scheduled-cash-flow hurdle. A nonnegative upfront fee F must satisfy:

```
F_min = max(0, P - sum_t(debt_payment_t / (1+h)^(t/12)))
F_max = min(equity_budget - E0_before_fee, V_e(0))
F_min <= F <= F_max
```

Do not clamp a negative upper bound to zero. An interval is only useful if monthly operating coverage passes and the cash ledger requires no additional support. An interval's existence is different from a particular quoted fee being acceptable. At an endpoint a party or funding constraint may bind.

This discounts promised lender cash flows; it is not a stochastic credit valuation, an expected-loss model or proof of lender willingness. Different parties' NPVs are not added into a measure of social value.

## Results: identify the window before extending the loan

Zero upfront fee; base operating cash of 3 throughout. The last column is the upper feasible fee before applying the other screens.

| Repayment months | Monthly debt payment | Initial equity required, before fee | Minimum operating coverage | Debt months after customer contract | Upper fee bound |
|---|---:|---:|---:|---:|---:|
| 36 | 2.7701 | 26.6206 | 1.0830x | 0 | -1.6206 |
| 41 | 2.4656 | 24.7933 | 1.2168x | 0 | 0.2067 |
| 42 | 2.4134 | 24.4803 | 1.2431x | 0 | 0.5197 |
| 48 | 2.1462 | 22.8772 | 1.3978x | 6 | 2.1228 |

The 36-month case requires 26.6206 of equity against a budget of 25 and covers debt service only 1.0830x. Extending to 42 months lowers the requirement to 24.4803 and lifts coverage to 1.2431x. The 48-month case is easier to fund initially, but runs past the contract.

Across the tested integer terms from 24 to 60 months, the base case has a joint fee interval at 41–54 months. Reducing only post-contract cash from 3 to 1.8 leaves **41–42 months**. These are scenario boundaries, not estimated optimal terms for an issuer.

At 48 months with weaker tail cash, operating coverage falls to 0.8387x. The residual-distribution policy then requires another 2.0772 of support during the six post-contract debt-payment months. Holding earlier surplus until debt repayment removes that support requirement, but the operating-cash coverage ratio remains 0.8387x. Available cash and operating income are different tests.

The same retention change reduces equity NPV from 6.9178, which includes the required future support cash calls, to 1.9780. The default residual case is not financed merely because its NPV can be calculated. Cash retention is a potentially useful liquidity design with an equity timing cost, not a cure for every contract definition.

## Pricing placement changes feasibility

Hold the 42-month term and the weak tail of 1.8 fixed. Raise the lender hurdle to 8%.

With the loan still priced at 7%, the lender needs an upfront fee of at least **1.4269**, but the sponsor can fund at most **0.5197**. The interval is empty.

With an 8% effective annual loan rate and no upfront fee:

- monthly payment is **2.4523**;
- initial equity required is **24.7135**;
- minimum operating coverage is **1.2234x**;
- debt is repaid at the modeled contract end, month 42;
- equity NPV is **4.6772**, and scheduled lender NPV is approximately zero at its 8% hurdle.

The lender receives its required time value through later payments. The sponsor avoids a larger cash charge at inception. The operating buffer is narrower, so this is not a claim that charging higher interest is generally better.

In the same weak-tail assumptions, searching 24–60 months finds no joint interval when the loan rate stays 7% and lender hurdle is 8%. Repricing both to 8% admits the 42-month term. The no-fee, par-priced 42-month rate frontier is approximately **9.2265%** effective annually. It is the output of the chosen budget and coverage constraints, not a market yield or a price for unmodeled credit loss.

## Two counterexamples keep the result honest

**Delivery delay:** increasing pre-operation time from six to nine months raises the 48-month structure's initial equity requirement to 29.3158. Its operating coverage still looks comfortable, but it fails the 25 funding budget. Extra tenor alone does not fund the cash needed to reach operations.

**Equity still needs a business:** at 42 months, removing all post-contract cash leaves debt coverage at 1.2431x and no post-contract debt. Yet equity NPV is **−7.6434** at the stated 12% rate. Protecting lenders through contracted cash does not establish adequate total equity returns.

The physical project's unlevered NPV is identical across term choices when its cash schedule and reference discount rate are held fixed. Under weak tail cash it is −0.2259 at the separately chosen 12% project reference rate, even where the parties pass their different participation tests. That arbitrary reference rate is not an empirically estimated asset hurdle. Neither financing feasibility nor a sum of party NPVs proves productive additionality.

## What to test against real transactions

A deal-specific application needs the payment schedule, the permitted reserve and distribution mechanics, actual customer collection dates, a defensible post-contract cash path, sponsor funding capacity and lender pricing for tenor and credit exposure. The constructive question is how much extension is sufficient, and which combination of coupon, upfront fee, reserve funding and contract duration makes the whole cash calendar work.

The next technical-economic sensitivity should map utilization, compute-price renewal, power cost and useful hardware life into operating cash—not treat an assumed cash flow as an observed company forecast.

## Reproduce

```sh
node verify-maturity-design.mjs
```

Files: `maturity-design-model.mjs`, `verify-maturity-design.mjs`, `maturity-design-results.json`.

The run passes **64 grouped checks** across 12 selected scenarios, monthly cash/principal identities, a separate equity-PV expression, fee endpoints, downside cash, zero-debt/zero-rate cases and the par-rate boundary. The term sweeps are deterministic comparisons, not statistical confidence or independent credit validation.

Publication note: this dated study is included in the September 24 public Compute Financing Research Kit. The baseline model and results are unchanged.