InfraComputeThe capital behind compute.
Financing / InfraCompute research

Sizing debt against operating cash

GPU capacity becomes financeable through a sequence of deposits, deliveries, acceptance and collections. Model that sequence before choosing leverage.

Explore the analysis

Test the cash available for debt service

Illustrative scenario1.47×

Illustrative debt-service coverage

0.66m cash available / 0.45m debt service.

Illustrative model resultScenario result

Assumptions & calculation

Monthly full-use revenue 1.80m × paid utilisation × rental-price factor; less fixed operating costs of 0.60m. Divide by debt service of 0.45m. Simplified model excludes taxes, working capital, refresh spending and reserves; not a contractual covenant definition.

Original InfraCompute hypothetical model. All monetary inputs use one consistent, unspecified currency; m means millions. Values are not market quotes, appraisals or recommended financing terms. Inputs are not saved.

01

Separate the funding gaps

A cluster can have enough committed capital to buy its equipment and still run short of cash before customers pay. Build a monthly sources-and-uses schedule that separates hardware deposits, delivery payments, installation, fees, operating losses and reserves. Then assign an available funding source to each use. An indicative facility is not cash, and an undrawn commitment may depend on milestones the project has yet to achieve.

American Compute’s cluster-finance guide describes the coordination problem between equity, customer commitments and debt. A funding-readiness test asks: what evidence does each provider require, who supplies it and when must it exist? This highlights circular conditions early, such as a customer requiring an operational cluster while the lender requires a signed customer contract.

02

Measure coverage after operating costs

Use cash available for debt service, after the operating costs defined in the model, as the coverage numerator. Gross rental revenue divided by loan payments can flatter the project because it leaves power, hosting, maintenance and other costs unpaid. State the convention and reconcile it to the contractual covenant definition before comparing ratios.

Utilisation, realised price and collection timing need separate assumptions. Technical uptime does not mean billable capacity; a signed reservation does not automatically mean an unconditional payment obligation. Model availability credits, ramp-up, termination rights and contract expiry. A flat utilisation assumption hides the period in which cash reserves are most exposed.

03

Draw to milestones; stress the refinancing need

Delayed draws can align debt deployment with hardware acceptance and reduce interest on amounts not yet borrowed. They can also introduce commitment fees, long-stop dates, eligibility conditions and an obligation to bridge deposits with equity. Compare the complete cash schedule rather than the headline rate. A cheaper facility that arrives too late can create an expensive funding gap.

For an illustrative stress test, move first customer collections back three months, lower realised rental prices and shorten the first contract. Recalculate minimum cash, debt-service coverage and the balance outstanding at renewal. Add junior debt only after those results are visible. Junior capital closes a funding gap but also adds a payment claim; it cannot create operating cash.

04

Keep benchmarks in their proper place

Calculate break-even utilisation from the project’s costs, realised price, financing and ramp. Published examples depend on their assumptions and cannot establish a market-wide threshold. An SPV also does not by itself establish bankruptcy remoteness or unrestricted contract transfer.

Apply the analysis

Questions for the credit paper

  1. Which costs cannot be paid from the equipment facility?
  2. What is minimum cash if acceptance or customer collection slips by three months?
  3. What debt remains when the first material customer contract expires?
Sources & scope

About this guide

Original InfraCompute analysis informed by the American Compute material below. The article structure, examples and interactive visuals are InfraCompute’s work. Source-specific findings remain attributed to American Compute; attribution does not imply affiliation or endorsement.

We reviewed the published edition and selected public references. We have not independently verified American Compute’s proprietary datasets, every market claim or confidential financing terms. Frameworks and hypothetical models describe questions to investigate; they do not establish protections in a particular financing. Legal and engineering conclusions depend on the relevant documents, system and jurisdiction.

Read our research methodology →
Next in the handbookComparing neocloud business models
Randeep Bubbra

InfraCompute Weekly

The week in compute finance, read from the filings.

Deals, terms and the clauses behind them, for the people financing data centres, GPU clouds and the power behind them. By Randeep Bubbra.

Working on a GPU or data centre financing? Book 20 minutes with Randeep.

Fridays, one email. Work email only. Unsubscribe any time. Privacy.