Reading a data center's balance sheet
ISSUE 03
A quick note before we start: the numbers below are illustrative, built to show how the structure works not a real company's financials. The point isn't the figures. It's learning where to look.
In Issue 01, the story was that AI data centers are financed against a hyperscaler's lease, not the building. In Issue 02, it was that DACH is starting to adopt the same playbook the US already runs. This issue is about what that actually looks like on paper — because the same physical data center produces two very different balance sheets, depending on how it's financed.
The conventional view
If a hyperscaler builds and owns its own facility outright, the balance sheet looks like any normal corporate one. Simple, and not very informative if you're trying to assess the risk of one specific site.
| Illustrative HoldCo Inc. | |
|---|---|
| Assets | |
| Property, plant & equipment | $1,200M |
| Cash & equivalents | $150M |
| Total assets | $1,350M |
| Liabilities & equity | |
| Corporate debt (unsecured) | $400M |
| Shareholders' equity | $950M |
Notice what's missing: no visibility into which tenant occupies the facility, how long the lease runs, or whether the site itself could service debt on its own. All the risk is absorbed into one corporate entity. That's fine if the corporate entity is Microsoft. It's a very different question if it's a smaller developer.
The SPV view
Now take the same physical facility, financed the way most AI infrastructure deals actually get structured ring-fenced into its own special purpose vehicle, with debt sized against the site's contracted cash flows.
A quick accounting note first, because it matters: you won't see both the physical facility and the present value of its lease listed as separate assets in a real filing. Under an operating lease, the lessor keeps the physical asset on the books and recognizes rent as it's earned. Under a finance (or sales-type) lease, the physical asset is derecognized and replaced by a lease receivable because that would double-count the same economic value. Most data center SPVs report on an operating-lease basis, so that's what's below.
| Illustrative Data Center SPV LLC | |
|---|---|
| Assets | |
| Data center facility (at cost) | $900M |
| Debt service reserve (restricted cash) | $40M |
| Total assets | $940M |
| Off-balance-sheet: contracted lease backlog (PV), used to size DSCR | $780M |
| Liabilities & equity | |
| Senior secured notes | $425M |
| Mezzanine debt | $70M |
| Sponsor equity | $445M |
| Key metrics | |
| Debt service coverage ratio (DSCR) | 1.65x |
| Loan-to-value (LTV, debt ÷ facility value) | 55% |
| Tenant concentration | 100% (single hyperscaler) |
| Weighted average lease term (WALT) | 13.2 years |
This is a completely different document from the conventional one, even though the lease backlog doesn't appear as a second asset. Instead, it's the input lenders use to size how much debt the facility can actually support DSCR is calculated straight off the contracted rent, independent of the balance sheet total. That's tenant-credit substitution in practice: the collateral is still the physical facility, but how much debt gets written against it is set by the tenant's credit, not the building's replacement cost.
Four numbers worth checking first
Whenever you're looking at a data center financing a private credit deal, an SPV prospectus, a REIT filing these are the lines that tell you more than the headline debt figure:
DSCR (debt service coverage ratio). How many times over does contracted cash flow cover debt payments. Below 1.2x is tight; above 1.5x is comfortable for this asset class.
Tenant concentration. A single-tenant facility is a bet on one counterparty's credit, full stop. Multi-tenant colocation spreads that risk, but usually at a lower yield.
WALT (weighted average lease term). Short leases mean the debt could outlive the contracted revenue backing it, a mismatch lenders price carefully.
LTV (loan-to-value). How much of the facility's appraised value the debt actually represents measured against the collateral value, not total balance sheet assets (adding cash reserves or a lease receivable into that denominator would understate real leverage). Investment-grade data center debt typically prices tighter below 55–60%.
The SPV structure isn't hiding the risk; it's relocating it onto a single balance sheet where it can be measured precisely. That's exactly why lenders prefer it to lending against a sprawling corporate entity with a hundred other obligations.
Once you know where these four numbers sit, a data center financing stops looking like a mysterious structured product and starts looking like what it actually is: a bet on one tenant's ability to keep paying rent, priced to the decimal.
Next issue: what happens to this balance sheet if the tenant leaves how AI infrastructure debt actually gets stress tested.
