Data Center Lease Structures: Turn-Key, Powered Shell

Introduction

Data center leases do not map cleanly onto office or industrial templates, and trying to read one as if it were a warehouse lease will miss the economics entirely. The taxonomy is organized around a single question: how much of the facility does the landlord actually build before the tenant moves in. At one end, the landlord delivers little more than a powered shell and the tenant fits out everything inside. At the other, the landlord builds a complete, ready-to-run turn-key data center and the tenant brings only its servers. The pricing gap between those poles is enormous: in one comparison, powered shell space leased for about $24 per square foot against roughly $163 per square foot for fully built turn-key space. Understanding where a given lease sits on that spectrum is the first step in underwriting it.

The two anchors of the spectrum are the powered shell and the turn-key lease, and they allocate capital and operational risk in opposite ways.

Powered Shell

A data center building delivered with structural shell, utility power, and connectivity in place, but with an unfinished interior. The tenant installs cooling, backup generators, and IT systems itself. These are almost always leased triple-net on 10-to-20-year-plus terms, with rent quoted per square foot.

A turn-key facility flips the arrangement. The landlord invests the heavy capital to build a complete, redundant data center, with power distribution, cooling, and building management all in place to a contracted service level, and the tenant supplies only its computing equipment. That capital intensity is why turn-key rent per square foot runs many times higher than powered shell rent; the landlord is recovering a far larger investment and taking on the operational risk of meeting uptime guarantees.

The choice between them is a make-versus-buy decision for the tenant. A sophisticated operator that wants control and long-term cost efficiency leans toward powered shell; a user that prioritizes speed and guaranteed reliability takes turn-key. Build-to-suit sits alongside turn-key as the bespoke version, where the landlord develops a facility to a single tenant's exact specification. These forms all differ from the general gross, net, and NNN lease structures of traditional real estate, even though the powered shell's triple-net character is recognizably the same idea, and they map onto the retail, wholesale, and hyperscale segments covered separately.

Hyperscale leases, used for the largest deployments, are typically structured as modified gross or triple-net agreements running 10 to 20 years, often with only one to three tenants in a facility. The defining feature is that the headline economics are quoted per kilowatt of capacity per month, not per square foot, because the tenant is buying power, not floor area. That quoting convention reduces to a single unit rate, the figure every hyperscale negotiation ultimately turns on:

Anchoring rent to critical IT load rather than square footage is what lets a powered shell and a turn-key suite be compared on the same axis: a tenant reserving capacity cares about the dollars it pays for each kilowatt it can draw, not the floor area that houses it. Rent escalators, usually a fixed annual percentage, are standard, which is part of what gives these leases their bond-like, long-duration cash flow.

The revenue inside a hyperscale lease splits into two main pieces. Roughly 70% to 80% comes from base rent for the reserved power capacity, and another 15% to 20% comes from metered power pass-through charges. That split matters because it determines how much of the landlord's revenue is fixed and contracted versus variable and tied to consumption.

The contrast with a traditional commercial lease is instructive. An office lease loads the landlord with recurring tenant improvements and leasing commissions, the mechanics covered in commercial lease mechanics, so the landlord re-spends capital every time space rolls. A hyperscale lease instead front-loads the landlord's capital into the building's power and cooling, then collects a long, escalating, largely net income stream with minimal ongoing leasing cost. That structure is why a stabilized hyperscale lease behaves more like an infrastructure annuity than an office tenancy, and why investors will pay a premium for it.

That bankability is also why these leases are now signed before the building exists. In a supply-starved market developers rarely build on spec; hyperscalers pre-lease 70% to 90% of a facility's capacity before construction begins, and increasingly they are not renting finished space at all but pre-purchasing energized capacity, with land, substation, transformers, and cooling bundled into a single contracted deliverable. The lease is the thing that de-risks the development: a signed hyperscale commitment is what lets a developer raise construction debt against a building that does not yet exist. That inverts the traditional real estate sequence, in which a landlord builds first and leases later, and it is a large part of why these contracts are better read as financing instruments than as tenancies.

The most distinctive feature of a data center lease is how electricity is handled. Power is usually passed through to the tenant at cost, often written as "plus E" (plus electricity), because utility rates and actual consumption fluctuate. This is not a minor line item: the power bill frequently dwarfs the base rent, so how the pass-through is structured is often the most negotiated part of the lease.

Power Usage Effectiveness (PUE)

A measure of data center energy efficiency: total facility power divided by the power actually delivered to IT equipment. A PUE of 1.0 is theoretically perfect; real facilities run higher. A lower PUE means less energy wasted on cooling and overhead, so the metric directly affects the tenant's all-in power cost.

PUE matters because it scales the bill the tenant actually pays. The pass-through is not just the raw cost of the power the servers consume; it grosses up that cost by the facility's overhead, so a less efficient building hands the tenant a larger invoice for the same compute:

The first term is what the IT load draws at the prevailing utility rate; multiplying by PUE captures the cooling and electrical overhead the tenant is billed for on top of it. At a PUE of 1.4 the tenant pays for 40% more energy than its servers actually use, which is precisely why the next clause is negotiated so hard.

PUE is where landlord and tenant incentives collide, and modern leases increasingly address it directly. A lease may include a PUE cap: if the facility's PUE exceeds, say, 1.4, the tenant is not responsible for the inefficiency above that threshold, which shifts operational risk back to the landlord and gives the landlord a reason to run the building efficiently. Some hyperscale tenants also negotiate to bring their own equipment, such as their own uninterruptible power supplies, which lowers the lease rate by removing that cost from the landlord's scope. Because power availability gates the whole sector, these clauses connect directly to the power constraint that defines data center development.

The lease form determines almost everything about the cash flow's quality. A long turn-key or hyperscale lease with a creditworthy tenant and fixed escalators produces contracted, predictable income that looks and trades like long-duration net lease, which is why these assets are increasingly compared to single-tenant net lease real estate. The diligence then centers on three things: the tenant's credit, since the entire value rests on a decade-plus of payments and makes the hyperscaler tenant credit the key assumption; the escalator and term, which set the growth and duration of the income; and the power pass-through and PUE provisions, which determine who bears energy-cost and efficiency risk.

The conventions travel across borders with local variation. In Europe, where capacity clusters in the Frankfurt, London, Amsterdam, Paris, and Dublin markets, leases broadly mirror the US turn-key and powered-shell forms, but longer fixed terms and index-linked escalators (tied to inflation rather than a fixed annual percentage) are more common, reflecting the net-lease conventions that prevail across European commercial real estate. A banker working a cross-border data center mandate has to translate the escalator and the pass-through mechanics rather than assume the US form applies unchanged.

Read a data center lease the way you would read a financing, not a tenancy. Its value turns on three things the square footage never captures: the tenant's credit over a decade-plus, the escalator and term that set the income's growth and duration, and the treatment of power, the pass-through and any PUE cap, where the real risk is allocated. Get those right and a turn-key or hyperscale lease is an infrastructure annuity; misread the power mechanics or the tenant credit and the same long contract is a concentrated bet dressed as stable real estate income.