Before the Servers Turn On: Managing Data Center Delivery Risk

Before a data center can go live, the project must clear a series of linked dependencies: site approvals, power and water availability, design decisions, long-lead equipment, skilled labor, construction completion, and commissioning all affect when planned capacity becomes usable capacity. A delay in any one part of that sequence can change the schedule the rest of the project depends on.  

Those dependencies are becoming harder to manage as data center demand puts more pressure on the grid. On June 18, 2026, the Federal Energy Regulatory Commission (FERC) directed the six regional grid operators under its jurisdiction to justify or reform the rules governing how data centers and other large energy users connect to the grid. The pressure behind that directive is not abstract: Berkeley Lab found that data centers accounted for about 4.4% of total U.S. electricity use in 2023 and could account for roughly 6.7% to 12% by 2028.  

For owners, developers, investors, contractors, and sureties, data center project risk typically starts well before a server is installed. Large land acquisitions, billion-dollar buildouts, aggressive delivery windows, and shifting scopes leave little room for bad assumptions. Our work at Vertex helps determine if the delivery plan can realistically support the target completion date.  

Community acceptance can shape the delivery path 

A completed data center may generate significant tax revenue with a comparatively lighter public-service footprint than residential, office, retail, or hospitality development. That dynamic may help explain why some municipalities welcome data centers and why facilities often cluster after the first project in a region wins approval.

“The tax case for data center development can be compelling, but the power and water case still has to hold up on its own merits.“

Public acceptance, however, is becoming harder to secure. In May 2026, Gallup reported that seven in 10 Americans oppose the construction of AI data centers in their local area, including 48% who strongly oppose them.  

Opponents most often point to environmental concerns, particularly water and energy use, along with quality-of-life issues and worries about possible economic effects such as higher utility bills. 

That concern is starting to show up in policy debates. Reuters reported that Maine lawmakers advanced a bill that would have paused approvals for data centers requiring more than 20 megawatts of power until October 2027 while a state-appointed council studied their impact on the local grid, electricity bills, air, and water. Gov. Janet Mills ultimately vetoed the bill, though Reuters also reported that roughly a dozen other states were separately weighing their own data center curbs. 

Water tends to be a central part of this public debate. The MOST Policy Initiative reported in 2026 that data centers rely on both energy-intensive air cooling and water-intensive evaporative cooling, and that as much as 85% of the water used can evaporate and not return directly to the local water supply. The specifics vary by facility and cooling method, but the concern is common enough to shape permitting conversations in some jurisdictions. 

Remote sites add a logistics test

As power and land constraints intensify in established markets, more projects are looking beyond them. Pew Research Center reported in April 2026 that more than 1,500 new U.S. data centers were in development at the time, with 67% planned for rural areas and 39% planned in counties that did not yet have a data center.

“Remote siting might solve one problem while introducing others – often shifting the risk into labor availability, logistics, emergency access, and local infrastructure capacity.“

Some regions in Canada are drawing similar interest, in part because of available power or comparatively fewer immediate siting restrictions. Those advantages can come with added requirements: work camps, worker housing, transportation planning, temporary infrastructure, fuel logistics, expanded staging, and imported labor may all be needed to make a remote site viable.

Large data centers in remote areas can also face heightened exposure to severe storms and wildfires, often sitting farther from emergency response resources. In these settings, contractors with experience in the local geography and climate can become extremely valuable.

Data center risk planning should start with the delivery assumptions 

Speed-to-market goals often depend on procurement and design assumptions that are still moving.

Long-lead equipment, vendor production slots, drawing maturity, trade buyout timing, and commissioning milestones can shape the data center schedule before field work begins. Early procurement steps may help secure equipment slots, while some scopes may be better held until issued-for-construction drawings are available to reduce design risk.

The schedule should reflect actual conditions on the ground: what must be procured, installed, staffed, tested, commissioned, and documented before the facility can operate.

Bring-your-own-power can create a second project

Power-generating facilities can take years to permit and build, and in some cases, the power source may take longer to deliver than the data center it is meant to serve. A facility may reach physical completion while approvals, equipment, fuel supply, testing, or transition planning are still in motion.

That timing gap is pushing more developers to consider bringing-your-own-power (BYOP). Dedicated onsite or nearby generation can give a project more control over power timing, but it also can add a parallel infrastructure program with separate permits, equipment, fuel logistics, environmental review, operations planning, and community considerations.

“In effect, a power workaround can become its own project, complete with separate permits, fuel logistics, testing, documentation, and operating risk.“

Interim generator power can add yet another layer. Running on generators while a longer-term power source is developed generally requires planning for diesel storage, fuel resupply, emissions permitting, maintenance, testing, natural gas access, pipeline capacity, and an eventual transition plan. The EPA’s Clean Air Act Resources for Data Centers points developers toward air-quality permitting guidance for equipment such as emergency generators, stationary engines, and other on-site power systems.  

Ultimately, BYOP is as much a permitting, fuel, environmental, schedule, and operations decision as a power decision. A sound delivery plan may need to account for how that second project will be permitted, procured, fueled, tested, documented, operated, and eventually integrated with the data center it supports.

Evolving design can change the project within weeks 

Data center capital commitments can grow large before the scope has fully stabilized. Site acquisition alone can require a nine-figure commitment. In January 2025, for example, Amazon Data Services acquired nearly 590 acres in Jefferson Township, Ohio, for just over $102 million as part of a planned data center campus. 

That is before one building becomes two, and two buildings become part of a broader program with parallel contracts, overlapping schedules, shared infrastructure, and competing resource demands. 

Design maturity can play a significant role in mitigating data center project risk. At a data center scale, rapid, high-value.

“At data center scale, a $20 million change order arriving weeks into the project can reset far more than the budget. It can reset the underlying delivery plan itself.“

Changes at that scale can affect more than price. They can reset buyout, procurement risk, manpower planning, subcontractor coordination, cash flow, schedule logic, and documentation while the project is already in motion.

Early advisory work can help separate what is genuinely firm from what is still provisional, so the project team knows which assumptions can shift without major impact and which ones, if they shift, could reshape the delivery plan.

Contractor readiness can become the limiting factor

Many data center projects aim to move from start to finish in under two years, even as equipment, power, and labor constraints can stretch the delivery path.

Some owners may push for triple shifts or heavy overtime to protect the schedule, but many markets simply do not have enough qualified labor to support that plan. Extended hours can also introduce fatigue, safety, quality, and supervision risks.

“Money can buy urgency, but it cannot install the work on its own. A project still needs enough skilled hands to execute it. “

Associated Builders and Contractors said the construction industry needs 349,000 net new workers in 2026, with demand for electricians capable of precision wiring rising sharply because of data center construction. Reuters also reported that about 41% of the current construction workforce is projected to retire by 2031. 

That scarcity can push owners deeper into the contractor market. When top-tier firms are already committed elsewhere, owners may end up turning to less experienced teams or to contractors with fewer systems in place to manage work at data center speed and scale. 

When that happens, the risk may show up in the controls behind the work rather than in the work itself: 

• RFIs and change orders moving faster than the project team can process them. 
• Payment cycles and cost controls falling behind field progress. 
• Subcontractor coordination becoming harder as scopes overlap. 
• Schedule updates that stop reflecting actual field conditions. 
• Documentation gaps that make delay, cost, or completion issues harder to evaluate later. 

In some markets, this pressure is also forcing unusual coordination across nearby projects. Contractors working on adjacent or associated projects may need to stagger work, share resources, or resequence scopes just to keep crews and materials moving — a dynamic that calls for program-level leadership even though each contractor remains accountable for its own commitments. 

There can also be additional impacts to nearby projects if dealing with a limited labor pool. Workers may opt for the higher paying job and companies may need to contend with higher labor costs to remain competitive. This could leave bonded subcontractors without the necessary resources or budgets to complete their projects as scoped.   

From a surety perspective, contractor readiness is best evaluated before the project is under stress. A bond claim can add investigation time at the exact moment an owner is trying to preserve momentum, so early review helps determine whether the contractor has the management depth, subcontractor relationships, financial controls, and resource base to perform at data center speed. Surety reviews look closely at where data center projects are located to help determine possible impacts to their accounts and ways to best manage those risks.  

The silent giant still changes after turnover 

A data center’s risk profile doesn’t end when the building is turned over. From the outside, a completed facility can look quiet, with only periodic waves of maintenance, service, and upgrade traffic. Inside, things can change quickly. 

Data center technology can evolve faster than construction timelines. AI chips, rack density, cooling needs, and power densities expected at the start of development may not be the same systems the project wants to install 18 months later. 

“Hyperscale operators may refresh equipment roughly once a year, which can drive fairly constant logistics and service traffic even with a lean onsite staff.“

That aligns with what is sometimes seen in the field. AI training infrastructure can require more energy and more efficient cooling than typical IT infrastructure, which can shape data center design and can make electricity access a key bottleneck for continued AI growth. 

For this reason, the project team needs to plan not just for construction, but for startup, replacement, maintenance, service access, cooling changes, and future load profiles. A facility designed around its first installation may be harder to adapt when the technology inside it changes. 

What early data center risk review should cover

Early review is more useful when it tests the assumptions carrying the delivery date before design, procurement, contractor buyout, and construction decisions are locked in.

Comprehensive data center risk planning can evaluate several key areas:

• Site, community, and operating constraints. Land strategy, permitting, water and cooling needs, remote-site logistics, fuel requirements, nearby competing projects, and community concerns. 
• Power and procurement readiness. Utility commitments, interconnection dates, generator plans, BYOP strategy, long-lead equipment, factory slots, and commissioning milestones. 
• Capital and scope exposure. Land acquisition, build-out, campus expansion, parallel contracts, and early change orders that may alter the financial model. 
• Contractor and resource capacity. Project management depth, subcontractor base, supplier relationships, financial controls, labor strategy, and field supervision. 
• Risk documentation. Records that can explain resequencing, acceleration, change orders, cost growth, completion risk, contractor performance, or delay later. 

At Vertex, our Data Center Project Advisory, surety, and commercial damages work helps teams evaluate these connected risks while there is still room to adjust course. The goal, in every case, is to understand what has to hold together before the servers turn on.  

Learn more about how Vertex helps teams evaluate schedule, procurement, and delivery risk before construction pressure builds.