The most valuable sentence in a data center lease is the one about electricity. Whether it is a condition or a promise decides who eats a two-year delay.
Data centers and digital infrastructure now show up routinely among the largest commercial real estate deals of the year, and the constrained resource is no longer capital—it is grid interconnection. That shift has moved leverage back toward whoever controls shovel-ready, powered land, and it has moved the negotiation’s center of gravity out of the rent article and into the power article.[1]
“Available power” is not a legal term.
It is a marketing term wearing a suit. A site can be described as having 200 MW “available” when what exists is a utility study, a queue position, and an interconnection agreement that has not been signed. Between that and electrons at the meter sit substation construction, transformer lead times, transmission upgrades, and a regulatory approval nobody controls.
So the lease has to define, in words, which of those milestones matter: energization of a specified capacity, at a specified date, measured at a specified point of delivery. A clause that promises “sufficient power for Tenant’s operations” has promised nothing a court can enforce without a trial about what “sufficient” meant.
The doctrinal fork
Once the power obligation is written down, it does one of two things, and drafters routinely fail to say which.
A condition is an event that must occur before performance becomes due.[2] If power at 100 MW by a date certain is a condition precedent to the lease term commencing, then failure means the obligation never matures—rent does not start, and a properly drafted clause gives the tenant a walk-away right. Nobody is in breach. The deal simply does not happen.
A covenant is a promise. If the landlord promises to deliver 100 MW by that date, failure is a breach, and the tenant’s remedy is damages—which, for a tenant that has already committed capital expenditure to servers and a customer to an availability commitment, is a lawsuit rather than a solution.
The same facts produce entirely different outcomes depending on which box the drafting fell into, and the doctrine’s default preference is not the tenant’s friend: where language is ambiguous, courts lean against reading a provision as a condition, precisely because forfeiture is a harsh result.[3] Ambiguity in the power clause therefore tends to resolve into “promise,” which means damages, which means the tenant’s downside is a claim rather than an exit.
What the market actually negotiates
The current template splits the difference along a timeline. Delivery-date slippage inside an outside date produces liquidated damages or rent credits; slippage past the outside date converts into a termination right. Service-interruption events after commencement produce rent abatement, and repeated service-level failures escalate into termination.[4] Reliability standards are quantified rather than described—the industry’s “five nines” figure permits roughly five minutes of downtime per year, which is a number a lease can be built on in a way that “commercially reasonable efforts” is not.
Two carve-outs then decide whether any of it holds. Force majeure that sweeps in “utility delays” quietly reverses the entire allocation. And a change-in-law provision determines who absorbs a new tariff, demand charge, or capacity charge—which matters because “all-in” power pricing frequently is not, and utility cost increases have a way of arriving as pass-throughs.[5]
The regulatory variable nobody priced in
Delivery risk is no longer only an engineering problem. On July 14, 2026, New York became the first state to impose a statewide pause, with an executive order suspending discretionary environmental permits for data centers of 50 MW or more for up to a year while the state writes rules on grid and ratepayer impacts.[6] Local moratorium and ballot activity has spread across dozens of states in the same period, and a separate wave of nuisance, water, and land-use litigation is testing already-built sites.[7]
For a lease signed in 2024, none of that was in the risk model. Whether it lands on the landlord or the tenant is decided by the definition of force majeure and the change-in-law clause—two provisions that were, in most negotiations, an afternoon’s work near the back of the document.
Why this is a portfolio question, not a document question
An owner with fifteen powered-land sites and a handful of hyperscale tenants does not have one power clause. They have fifteen, drafted at different points in a market that repriced this risk twice, some as conditions and some as covenants, with outside dates and abatement triggers that do not line up. The exposure is not in any single lease. It is in the distribution.
This is the same structural problem as the remedies clause in a purchase agreement: the provision that governs the worst-case outcome is the one least likely to be standardized, and the one most likely to be buried in a rider. Silence is never neutral—it selects a default, and the default is rarely the one you would have chosen.
Reading fifteen power clauses at once#
Answering which of our leases let the tenant walk if power slips two quarters? by opening fifteen documents and their amendments is exactly the work that used to consume associate weeks—and it produces an answer that is stale the next time anything is signed.
This is the problem Acrebase is built for. Acrebase is AI-powered contract intelligence for commercial real estate: it parses leases, purchase agreements, and amendments into structured data, flags unusual or missing provisions, and lets you query an entire contract library in plain language. Ask which leases treat power delivery as a condition versus a covenant, which have outside dates inside the next four quarters, which include utility delay in force majeure, and which are silent on change-in-law pass-throughs—and get answers with every extraction traced back to the exact language in the source document, because a termination right you cannot cite is a termination right you do not have.
For teams that would rather catch it during drafting than during diligence, the same review sits inside Microsoft Word. The Acrebase Word add-in puts chat, proofreading, a clause library, and redlining in the task pane next to the document, so “shall use commercially reasonable efforts to provide adequate power” gets flagged while it is still a draft. It is the working version of the argument we made about AI in CRE due diligence: the software handles extraction and comparison across the portfolio, the attorney keeps the judgment. And because power negotiations are slow, it is worth remembering what long negotiations do to deals when the interconnection queue is moving underneath you.
The practical takeaway: in a data center lease, write the power obligation as a dated, metered, capacity-specific milestone, and then say in terms whether missing it is a failed condition or a breach. If the document does not say, the doctrine will say it for you—and it will usually say “breach,” which leaves the tenant holding a claim instead of a door.
Footnotes
[1] “Hyperscale Data Center Lease Terms in 2026: What Developers Need to Know.” https://build.inc/insights/hyperscale-data-center-lease-terms-2026
[2] Restatement (Second) of Contracts § 224 (1981) (condition defined); § 225 (effects of non-occurrence).
[3] Restatement (Second) of Contracts § 227(1) (1981) (preference against constructions creating a risk of forfeiture).
[4] Orrick, “Data Centers: Leasing Considerations” (Jan. 2026) (service levels, rent credits, acceptance testing, escalation to termination). https://www.orrick.com/en/Insights/2026/01/Data-Centers-Leasing-Considerations
[5] Vinson & Elkins, “The Coming Wave of Disputes in Data Centre and AI Infrastructure” (shortfall and delivery failure, force majeure, and change-in-law claims). https://www.velaw.com/insights/the-coming-wave-of-disputes-in-data-centre-and-ai-infrastructure/
[6] N.Y. Exec. Order No. 62 (July 14, 2026). https://www.governor.ny.gov/news/first-statewide-moratorium-new-hyperscale-data-centers-launched-governor-kathy-hochul
[7] WilmerHale, “Data Centers in Court: The Emerging Wave of Nuisance, Environmental, and Land-Use Litigation” (July 13, 2026). https://www.wilmerhale.com/en/insights/client-alerts/20260713-data-centers-in-court-the-emerging-wave-of-nuisance-environmental-and-land-use-litigation
Acrebase is AI-powered contract intelligence for commercial real estate — clause extraction, risk flagging, and portfolio-wide search, plus an AI contract review add-in for Microsoft Word. Learn more at acrebase.com, install the Word add-in from Microsoft AppSource, or get in touch about pricing.