Managing the Pressure of Data Center Construction

There’s a moment on a data center job where you have to model a piece of equipment before anyone knows how big it is.

The switchgear is on a lead time measured in months. The schedule won’t wait for it. So the model gets a space claim, a rectangle standing in for a lineup whose real dimensions are still sitting in a submittal that hasn’t come back.

The team coordinates around the placeholder. Then the submittal lands, and, unfortunately, the coordination happens again. Joe Maraugha, who has spent more than a decade on this work at Lighthouse Electric outside Pittsburgh, calls that the “double pass routine.”

Nobody made a mistake. The challenge is managing incomplete design information across the coordination process. The drawings will change and the equipment will change, and the teams that succeed are the ones who can keep everyone building from the same understanding while it does.

That’s increasingly how these jobs get awarded.

For contractors already asked to support this work, the challenge begins before mobilization. Project teams face demanding qualification standards, incomplete information and schedules that leave little room to recover from mistakes.

Many hyperscale data center projects are negotiated rather than awarded through an open low-bid process. Often, they’re delivered as construction manager at risk. The owner picks a builder early, during design, on qualifications and track record rather than the low number. In many CMAR arrangements, the construction manager develops a guaranteed maximum price and manages or holds the trade contracts. Those contracts are typically awarded to trade contractors already sitting on a prequalified roster.

So, there’s no open bid to sharpen your pencil against. By the time a scope is real, the list of firms eligible to perform it is already short, and getting on that list is a question of who’s seen your work.

Shawn C. Tovey, RCDD, has worked across field installation, construction supervision, operations, estimating, engineering, design-assist, QA/QC and design. Before moving fully into design, he led estimating and pre-award technical coordination for large-scale physical security and ICT programs, including hyperscale data centers. He describes the way the work arrives as almost entirely a function of who already knows you.

Joe Maraugha and David Griffin work at Lighthouse Electric, a 1,400-person electrical contractor outside Pittsburgh that has done data center work for more than a decade. Asked what they would tell a mid-size firm trying to break in, neither one said it was impossible. They said start smaller than you think.

The reason that works is a change in how owners buy the work. On large projects, scopes that used to go to one firm are increasingly broken into pieces. On the electrical side, Maraugha describes four.

• Package A. Core and shell. Large conduit, large feeders, large equipment. The heaviest scope and the one experienced firms compete for.
• Package B. Branch power and branch lighting. Smaller, more familiar work, and the usual first step for a firm that has not done a data center before.
• Package C. Data and IT scope
• Package D. Building automation and building management systems

The same splitting is happening across the other trades, and the advice that comes with it is to be honest about which piece you can realistically deliver.

The splitting is deliberate. Owners and general contractors spread labor and risk across several contractors on purpose, because that’s how they protect a compressed schedule. The same move that lowers their risk is what opens a door for a smaller firm.

There’s a second way in that does not require being the biggest firm in the room. Griffin points to joint ventures, where two contractors split a scope by what each one does best.

Their last piece of advice is the one nobody expects: turn work down. Lighthouse passes on jobs when the schedule is too compressed to hit, when the contract terms push too much risk onto the trade partner, or when the scope is a poor fit for the crews they can staff. Saying no to the wrong job is the same instinct that builds a delivery record, applied one job earlier.

Before you take a package, it’s worth knowing whether your team could deliver it.

That’s exactly what the one-page field guide at the end of this piece is built to help you answer.

Owners are unforgiving on these jobs for a simple reason. By the time construction starts, their margin for error has already disappeared. Power is delayed, critical equipment is backlogged, and experienced crews are in short supply.

Power is the biggest. Goldman Sachs Research puts global grid spending near $720 billion through 2030 just to keep pace, against a 165% jump in demand. In PJM, new generation now runs past eight years to come online, and large transformers are quoted near four years. We’ve written about the permitting crunch and the materials scramble behind it.

Labor is just as tight. Craft positions are going unfilled and a wave of the workforce is retiring, while data centers, grid upgrades, chip fabs and LNG terminals all compete for the same trades. We have made the labor case and the physical economy case at length. None of it is something a contractor can fix. The one variable still on the table is whether the team performing the work makes the squeeze better or worse, which is where the rest of this piece lives.

The constraints that dominate data center coverage are the ones a contractor can least control. The variable that affects repeat performance is closer to the jobsite: whether information reaches the people who need it before the schedule exposes the gap.

The constraints above are real, but they don’t automatically determine whether a contractor succeeds.

Jobs are often lost in ordinary breakdowns after the trailer goes up. Data center work does not forgive those mistakes the way a parking structure does, and the reason is arithmetic.

Electrical alone can run 35% to 45% of project cost, against 15% to 20% on a typical commercial building, and all of it has to sequence inside a window that keeps shrinking. There’s little float available to absorb avoidable rework or late changes.

The compression starts before anyone breaks ground. Tovey, from the estimating side, points out that a team may have only five to ten days to price a multimillion-dollar scope, and no chance to run down every answer before the number is due. So the number a contractor submits rests on assumptions that have not been tested yet, and what happens to those assumptions after the award is the first place a job starts losing money.

And the misses aren’t rare.

Rework can run as high as 9% of total project cost. Field crews lose roughly 14 hours a week, more than a third of their time, hunting for information and rebuilding what was built wrong. An FMI and PlanGrid study, which we’ve written about, traced close to half of all rework to poor data and miscommunication: crews building from information that was wrong, late or inaccessible.

That’s the same coordination performed twice, on a schedule that had no room for it once. Griffin describes how it compounds on a congested job: overlapping work areas, trades stacking on top of each other, and RFIs raised early that only come due much later. Once a team loses its grip on that, it rarely gets it back. If markups don’t keep pace with the revisions, the field just builds the wrong thing more efficiently!

At the high end, rework can exceed a contractor’s typical net margin and turn a profitable job into a loss.

Mortenson faced 40 miles of duct bank on a data center campus in Eagle Mountain, Utah, and rather than form and pour in place, moved the work indoors. Its prefabrication arm, BLUvera, ran factory production and trucked the finished assemblies to the site.

ENR reported the results: on-site labor hours down 56%, quality defects down 83%, safety incidents down 87%.

A contractor’s own discipline is necessary, but it can’t overcome a project team that hasn’t aligned around the process. Maraugha argues that prefabrication only pays off when the whole project is committed to it:

Griffin makes the same point from the schedule side. You can stage your own work perfectly and still be held up by a trade partner who did not.

The same logic runs through the handoff from estimating to operations. Tovey argues that the moment a notice to proceed lands, estimating and operations need a formal kickoff to review the bid assumptions, scrub the bill of materials, confirm quantities and scope, and turn the estimate into an executable baseline the field can build to.

Accuracy at that handoff is the difference between a crew that starts building and a crew that starts asking questions. The same is true at the other end of the job, at commissioning, which runs through five formal levels ending in an integrated systems test under simulated load and failure. Level 5 is a very bad time to learn the as-builts drifted from reality back in March!

Predictability is the product a specialty contractor sells here, though it never shows up as a line item. The four disciplines make that predictability repeatable. The prequalified roster is where it gets rewarded.

Contractors sometimes hear all this as an insiders’ club with the door held shut. In practice, rosters are often built through performance on smaller jobs, and package-level scopes can make a smaller job one piece of a much larger project. It also comes down to who is reading the drawings. Asked what makes an estimator suited to this work, Tovey did not name a credential.

Most estimating still happens off flat drawings. The teams that price this work well are the ones whose people can picture the finished install in the field, and the disciplines above are how that judgment survives contact with a schedule.

The firms that maintain reliable execution across multiple projects succeed by being the contractor whose scope doesn’t blow up the schedule, on job after job, until somebody notices. The challenge is making that performance repeatable as workload grows.

When data-center work reaches a contractor’s team, document control shapes how well that team can respond to change, protect its margin and support the rest of the project.