Construction-to-permanent for a data center

A construction-to-permanent structure is not a single loan that changes character. It is two facilities with two different credit analyses, joined by a test.

The construction stage funds work. It is drawn progressively against certified progress rather than advanced at the outset, it accrues rather than amortises, and it is underwritten on whether the building will be completed on budget and on schedule. The security is the property, the construction contract, the performance security behind it and whatever support the sponsor provides. There is no cash flow to look through to, so the credit rests on completion.

The permanent stage holds a finished asset. It is underwritten on the cash flow the asset produces — rent under a lease, or fees under hosting arrangements — and on the credit of whoever pays it. Its tenor is matched to the building's life and to the contracts that occupy it, which on layer 2 means a horizon measured against a generation rather than a product cycle.

Those two analyses have almost nothing in common. One is a construction risk question; the other is a credit question about a tenant. The conversion test is where the second replaces the first, and it exists because the party funding the build wants to know it is not left holding a completed building with nobody in it.

The alternative shape is a construction facility with a separate takeout, which does the same work with two lenders and a refinancing. Whether the conversion is contractual or the takeout is a fresh transaction is a real distinction: a contractual conversion allocates the refinancing risk at the outset, and a separate takeout leaves it with the sponsor until the day it happens. Both are ordinary. Which one is appropriate depends on how much confidence there is in the conversion conditions being met, which is the subject of the next two sections.

Conversion is a checklist, and the useful way to read it is not by what each condition says but by who has to act for it to be satisfied. A condition inside the sponsor's control is a management problem. A condition outside it is an allocated risk wearing the clothes of a formality.

On most buildings, completion and occupancy arrive together and the conversion test is a formality once the contractor is finished. On a data center they can be separated by a long interval, and the reason is that the building is not commercially complete until power reaches it at the contracted capacity.

This produces a structure with a hole in it. The construction facility was sized against a build programme. The build finishes. The permanent facility cannot convert because the facility cannot be energised, and the tenant's obligations do not commence because it cannot take capacity. Nobody is at fault, the asset is physically finished, and there is a period with construction-stage capital outstanding, no conversion available and no revenue.

There are only a few responses, and they are chosen at structuring rather than discovered later:

• Size the construction period against the energization date rather than the build programme, with margin. This is the cleanest answer and the most expensive one, because it extends the period over which the more costly capital is outstanding.
• Separate the two events in the documents. Convert on completion, with the coverage tests deferred to a stated date after energization and sponsor support covering the interval. This keeps the structure intact but requires the sponsor to fund a gap it may not have budgeted.
• Bridge the interval separately, which is what the [bridge to energization](bridge-to-energization) instrument exists for. It is a distinct facility with a distinct security package precisely because it is a different risk.
• Remove the dependence. Where on-site generation can serve the load, energization stops being a queue question and becomes a delivery and commissioning question, which sits closer to the sponsor's control. That is one of the reasons behind-the-meter capacity is worth what it is worth, and it is structured as its own layer — see [financing on-site generation](financing-onsite-generation).

The failure is not choosing one of these. A conversion test containing an energization condition, with a construction period sized against the build, has allocated a third party's schedule to whoever is holding the facility when it matures.

Between drawdown and conversion, the party funding the build is exposed to an incomplete asset — which is worth materially less than the sum spent on it. Completion support is what closes that gap, and its scope is where much of the negotiation sits.

The forms recur across markets. A completion guarantee obliges the sponsor to finish the works, which is a different obligation from paying a sum of money. A cost-overrun undertaking obliges it to fund cost above budget, and its edges matter: whether it is capped, whether it covers overruns from any cause or only specified ones, and whether delay costs are included. Contingency held inside the facility does the same work up to a point and is exhausted first. Retention under the build contract keeps a portion of the contractor's money until defects are resolved. Liquidated damages for delay compensate but do not deliver a building.

The questions that decide whether the package is worth what it appears to be:

• Who is giving it? Support from an entity with no assets is documentation, not credit. Whether the obligor is the entity with the resources is the same test applied to any counterparty.
• What triggers it, and what discharges it? Support that falls away on completion leaves the energization gap uncovered, which on this asset class is the gap that matters.
• Does it survive the events that would call it? Support that terminates on the sponsor's own insolvency is worth nothing in the circumstance it was written for.
• Is the contractor's own security real? Performance bonds and parent guarantees behind the build contract are part of the completion analysis, and contractor failure part-way through a specialised build is not a hypothetical risk.

A structure with strong completion support and a weak conversion test is usually preferable to the reverse. The first funds the sponsor through a delay; the second simply records that the delay happened.

The recurring ways this structure disappoints:

• Completion and energization treated as one event. Covered above, and it is the sector-specific failure. Everything else on this list is ordinary construction lending.
• A takeout market assumed rather than committed. Where the permanent stage is a refinancing rather than a contractual conversion, the sponsor holds market risk from first drawdown to completion. That is a real position and it should be priced as one rather than assumed away.
• Conversion conditions that were never reconcilable. Coverage tests calculated on contracted cash flow, in a structure whose tenant's obligations commence on a date after the conversion deadline. The two documents were negotiated by different people and never read against each other.
• Scope creep against a fixed facility. Data-center fit-out changes with tenant requirements, and a scope that moves after the budget is set consumes contingency that was there for something else.
• Liens surviving completion. Subcontractor claims that block clean title are a conversion blocker the sponsor may not have caused and cannot unilaterally resolve.
• The building is complete and the layers were never separated. A single facility spanning shell, generation and equipment converts once, on one test, at one tenor — and reintroduces exactly the mismatch that separating the layers was meant to solve. The argument is set out in [tenor mismatch](/compute/tenor-mismatch-compute-and-infrastructure).

Continuum structures and arranges these transactions and coordinates the parties in them. It is not a bank, not a broker-dealer and not a direct lender; it does not provide construction capital or hold client funds.