A data center can be finished as a building and still be months away from doing the work its owner purchased it to perform. Permanent power may be unavailable. A generator may fail an integrated test. A change in computing equipment may require a different cooling design. Each problem can delay the date on which the owner can deliver capacity to a customer, even if much of the physical construction is complete. The gap between building completion and operational readiness is where many legal disputes arise.
Artificial intelligence and cloud computing have accelerated data center development, but the pressure to deliver capacity has arrived alongside shortages of skilled workers, long equipment lead times, and constrained electric grids. For contractors, design professionals and insurers, these are familiar construction risks in an unfamiliar combination. The most consequential decisions may be made before mobilization, when the parties decide who controls each dependency and who bears the cost if it fails.
A Larger Pipeline, Concentrated in Key Markets
JLL estimates that roughly 100 gigawatts of new data center capacity could come online globally between 2026 and 2030, requiring up to $3 trillion in investment. Its figures also show average global shell-and-core construction costs rising from $7.7 million per megawatt in 2020 to $10.7 million in 2025, with $11.3 million forecast for 2026. Those construction figures exclude the tenant's technology fit-out, which JLL says can cost considerably more for AI infrastructure.[i]
The U.S. construction pipeline is substantial, but it is not evenly distributed. CBRE counted 7,481 megawatts under construction across eight primary North American markets in the first half of 2026, a 24.8% increase during that six-month period. Atlanta led active construction for the first time, with 2,882 megawatts underway. Northern Virginia remained the largest market by existing inventory and had another 2,420 megawatts under construction. CBRE's primary-market group also includes Dallas-Fort Worth, Chicago, Phoenix, Silicon Valley, the New York tristate area, and Hillsboro, Oregon.[ii]
Development is also reaching beyond those established hubs. CBRE identifies West Texas projects advancing toward energization and expects continued developer interest in Indiana, Michigan, Pennsylvania, and North Carolina as power constraints shape site selection.[iii] These are forecasts and market observations, not guarantees that every proposed project will be built. They do show why contractors may increasingly encounter data center work in places without a deep pool of firms experienced in delivering it. For the construction industry, the significance lies in the overlap. Multiple large projects in the same region may compete for the same electricians, mechanical trades, supervisors, manufacturers, and utility services. A delay on one campus can also affect crews or equipment committed to the next.
The Labor Shortage Can Become a Defect Claim
Kelly Services projects that permanent data center employment will reach 650,000 positions by the end of 2026, up 30% from 2023. It also expects data center-related construction jobs to exceed 180,000 through 2028. According to Kelly, competing hyperscalers and operators hire approximately one-quarter of data center personnel.[iv] These are workforce estimates, but they help explain the strain on project staffing.
That strain can surface in litigation under another name. A project that loses an experienced superintendent may later face a dispute over sequencing or supervision, A newly assembled electrical crew may complete its assigned installation but leave problems that become apparent only during testing. Extensive overtime may preserve a near-term milestone while increasing rework.
It would be a mistake to assume that a labor shortage excuses deficient work. It is more useful to recognize staffing as an early indicator of claim exposure. Contractors should evaluate whether proposed subcontractors have the personnel, supervision, and financial capacity to perform the work they are taking on, particularly when several projects will draw from the same regional workforce. Insurers assessing a project should pay attention to those commitments as well as to the subcontractor's prior data center experience.
A Completion Date is Only as Reliable as its Dependencies
Data center owners need usable computing capacity, often by dates connected to leases, customer commitments or technology deployments. Contracts may therefore attach substantial consequences to missed milestones. Yet a contractor's work is only one element of the delivery schedule. JLL reports that 57% of data center projects experienced construction delays of at least three months in 2025. It also puts average U.S. equipment lead time at 42 weeks, 83% above 2019 levels. Those figures do not establish who caused any particular delay. They do show an aggressive completion date requires more than an ambitious field schedule.[v]
Consider a facility whose electrical rooms are ready, but whose switchgear arrives late. Or one whose systems are installed, but whose utility connection is delayed. An owner-directed increase in rack density might require electrical and cooling changes while construction proceeds. More than one event may affect the critical path, and responsibility may change over time.
The contract and baseline schedule should make these dependencies visible. Milestones should distinguish work the contractor controls from owner design decisions, owner-furnished equipment, utility service, governmental approvals, and access to work areas. Notice provisions, extension rights and compensation for delay should reflect that allocation. Without those terms, the parties may spend more time debating whether the contractor missed a date than determining what actually prevented the facility from becoming operational.
Equipment Orders are Being Placed Before Design Settles
Transformers, generators, switchgear, uninterruptible power systems and cooling equipment can drive the schedule long before they arrive on site. JLL reports that some developers order materials as much as 24 months ahead of need. Early purchasing can protect a delivery date, but it also means committing substantial funds while design and tenant requirements may still change.[vi]
That creates questions which ordinary change order language may not answer well. Who authorizes an early purchase? Who pays a deposit and takes title? What happens if the owner suspends the project after fabrication begins? If specified equipment becomes unavailable, who approves a substitute and pays for the related redesign or testing? Does a manufacturer's delay entitle the contractor to schedule relief, and what proof must the contractor provide?
These issues deserve an equipment specific procurement plan incorporated into the contract. The plan should address release dates, design assumptions, storage and insurance, cancellation charges, substitution, warranties, and the records needed to establish a delay. An owner may reasonably ask a contractor to manage purchasing. That instruction should not silently transfer every risk associated with manufacturing and design changes to the contractor.
Commissioning is Where Separate Systems Become One Facility
Many construction disputes begin when a component does not work. Data center disputes may begin when every component appears to work separately, but the facility fails as a system. ASHRAE describes a progression of commissioning activities that end with integrated systems testing. That final stage tests how the facility responds under operating and failure scenarios, rather than checking equipment only in isolation. ASHRAE also observes that accelerated turnover and new AI infrastructure are putting pressure on traditional commissioning sequences, particularly where projects use liquid cooling. [vii]
Suppose backup generation starts properly, the cooling system passes its functional test, and the controls operate as programmed. If the systems do not respond together as intended during a simulated power loss, the facility may still be unable to accept the planned computing load. Responsibility could involve design, installation, equipment, controls programming or the test procedure itself. The general contractor, specialty trades, engineers, vendors, and commissioning agents may each hold part of the record.
"Substantial completion" should therefore be defined with care. The documents should indicate:
- which tests are required for each turnover milestone;
- whether permanent utility power is a condition;
- who provides test loads and operating personnel;
- who records results; and
- how a failed test will be corrected and repeated.
They should also distinguish construction completion from the owner's installation and operation of its technology equipment. Precise acceptance criteria will not prevent every technical dispute, but they give the parties a shared measure of performance.
Design Changes Can Reach Far Beyond the Data Hall
Computing requirements may change during a project that takes years to plan and build. Higher density equipment can alter electrical distribution, heat rejection, piping, structural loads, and controls. CBRE reports growing use of liquid cooling to support dense GPU installations, including conversions of existing air-cooled facilities whose ready-for-service dates may extend by months.[viii]
An owner needs to respond to changing technology. A contractor needs to know when a revised requirement changes its price or time to perform. The difficulty is that these changes may arrive piecemeal. A revised equipment schedule, a meeting instruction, then a design bulletin several weeks later. By then, procurement or installation may already be underway.
Project teams should record the date of each changed requirement and identify its effects across disciplines. Requests for information, design bulletins, change notices and schedule updates should describe the work that changed, the decisions still needed, and any procurement commitments already made. That record matters most when the project is moving quickly. Reconstructing it after a failed milestone is far harder.
The Grid Has its Own Schedule
The most carefully managed construction schedule cannot create utility capacity. JLL calls "speed to power" the leading site-selection criterion and reports that average grid-connection waits in primary data center markets exceed four years.[ix] Federal energy analysis likewise projects a steep increase in U.S. data center electricity use, from approxinmately176 terawatt-hours in 2023 to a range of 325 to 580 terawatt-hours by 2028. The forecast has a wide range, but its direction illustrates the pressure on power planning.[x]
Energization can depend on easements, interconnection approvals, substations, transmission work and utility equipment that falls outside a building contract's control. Even an on-site generation plan introduces its own design, permitting, procurement and testing obligations.
The parties should identify who is responsible for each power-related step and what occurs if its slips. A useful contract does more than state that power will be available by a target date. It ties that assumption to the construction and commissioning schedule, provides a process for revising milestones, and addresses standby costs or resequencing if the assumption proves wrong.
Community Challenges Are Reaching Active Projects
Power and land availability no longer settle the site-selection question. Communities are scrutinizing water use, noise, emissions, traffic, and potential effects on electric rates. CBRE now identifies local opposition and zoning delays as material development constraints.[xi]
Data Center Watch, an organization tracking opposition to proposed facilities, reports that at least 75 U.S. projects representing approximately $130 billion were blocked or delayed during the first quarter of 2026.[xii] Its count should be understood as the organization's tracking estimate, rather than a government census. Even with that qualification, the figure shows the scale of entitlement risk confronting developers.
A dispute in Massachusetts, offers a concrete example. Residents challenged approvals connected to the expansion of an existing data center. In August 2026, a Massachusetts Superior Court judge paused installation of four additional diesel generators while the challenge continued. The interim order did not resolve the underlying case, but it demonstrates how a permitting dispute can halt a defined part of a project.[xiii]
A contractor usually has limited control over the developer's entitlement strategy or a neighbor's lawsuit. Its contract should address the effect of permitting appeals, injunctions and owner-controlled approvals on schedule and cost. Those provisions are especially important if procurement or mobilization will begin before all approvals are secure.
Put the Risk Allocation to Work Before Mobilization
No contract can remove the technical and external pressures on a data center project. It can make their consequences more predictable.
- The first task is to identify the promised deliverable. Is the contractor providing a completed building, installed infrastructure, a commissioned facility, or capacity demonstrated at a specified load? The answer should carry through the scope, milestone schedule, testing requirements and payment provisions.
- The second is to match responsibility with control. Procurement commitments, utility service, permits, owner decisions and tenant changes should have named owners and documents decision dates. Delay provisions should account for those dependencies without excusing failures the contractor can reasonably prevent or mitigate.
- Finally, the parties should test whether the financial exposure matches the project and its insurance. Lost tenant revenue or unavailable computing capacity may greatly exceed the construction fee. Liquidated damages, consequential- damage waivers and liability limits warrant careful drafting alongside review of builder's risk, liability, professional liability, and applicable delay coverage. Coverage cannot be assumed from the facility's value or its importance to the owner; it depends on the policies and the cause of loss.
The data center boom presents substantial work for the construction industry. It also brings projects in which a late component, an unresolved design change or an unavailable utility connection can have consequences far beyond the affected trade. The firms best positioned to manage those claims will be the ones that define operational milestones clearly, preserve a reliable project record and allocate each dependency while there is still time to manage it.
_______________________________
[i] 2026 Global Data Center Market Outlook. JLL. (January 5, 2026).
[ii] North America Data Center Trends H1 2026. CBRE. (August 27, 2026).
[iii] Id.
[iv] Kelly® releases 2026 Data Center Salary Guide highlighting compensation, hiring trends driven by AI acceleration. Kelly Services. (August 25, 2026).
[v] 2026 Global Data Center Market Outlook. JLL. (January 5, 2026).
[vi] Williams, Adam. How Long Does It Take to Build a Data Center (In 2026). Data Center Scope. (August 22, 2026).
[vii] PNNL/ASHRAE/NEMA AI Data Center Energy Performance Framework: Supporting Essential Infrastructure for the New AI Economy. ASHRAE.
[viii] North America Data Center Trends H1 2026. CBRE. (August 27, 2026).
[ix] 2026 Global Data Center Market Outlook. JLL. (January 5, 2026).
[x] DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers. Department of Energy (Dec. 20, 2024).
[xi] North America Data Center Trends H1 2026. CBRE. (August 27, 2026).
[xii] Q1: Data Center Watch Report. Data Center Watch.
[xiii] Fortes et al. v. Massachusetts Department of Environmental Protection & another, Civil Action No. 2681CV01107 (Mass. Super. Ct. Aug. 10, 2026)
