Construction has always had its dominant constraint. In the post-pandemic years, it was materials: lumber, steel, and critical components with lead times that blew schedules before a foundation was poured. As that eased, the conversation shifted to labor, specifically the craft worker shortage that the data center buildout has since turned into a full-scale structural problem. Both of those constraints had a common feature: they were visible, well-documented, and the industry had at least a theoretical response to them.

The constraint that's quietly taken over in 2026 is different. It doesn't show up clearly on a project risk register. It lives outside the traditional delivery process. And by the time most teams encounter it, the damage to the schedule is already done.

Power. Specifically, access to it.

What was briefly a data center problem has spread into the broader large-load project market, and the developers and GCs living it are describing it in terms that will sound familiar to anyone who has watched a supply chain problem compound.

Chris DeVito, General Manager of Owners and General Contractors at Built Technologies, has been out in front of clients and developers across US construction markets extensively over the last few months. The pattern he keeps hearing isn't about budget or labor or design complexity.

Chris drew an explicit parallel to the supply chain disruptions of 2022 and 2023. The constraint was different back then, and the industry had a playbook for it. Power doesn't have one yet.

On the demand side, Goldman Sachs Research's November 2025 GS SUSTAIN report projected global data center electricity demand rising 175% by 2030 compared to 2023 levels, the equivalent of adding another top-10 power-consuming nation to the global grid. That forecast has been revised upward three times in eighteen months as AI compute demand has outrun every model that tried to predict it.

On the infrastructure side, the equipment needed to actually deliver power to new facilities is severely constrained. Power Magazine's analysis from early 2026, drawing on Wood Mackenzie's Q2 2025 survey data, paints a stark picture of where things stand:

A project team ordering a transformer today is looking at delivery in mid-to-late 2028 at the optimistic end. That alone collapses a conventional construction schedule.

The interconnection queue problem sits on top of all of this. PJM, the grid operator covering 13 states, including Virginia's data center-heavy corridor, is currently processing connection requests from 2020. New large-load applications filed today are unlikely to get capacity auction eligibility before the early 2030s. Hydro-Québec, despite sitting on some of the most abundant and affordable hydroelectric capacity in North America, suspended all new large-load connections in 2024. Ontario followed in 2025 with a requirement for ministerial approval before any data center can connect to the provincial grid.

So you have surging demand, a six-year-plus interconnection backlog in prime markets, and physical infrastructure with two-to-four-year lead times. The real puzzle is that it took this long to become the dominant conversation on job sites.

The hyperscalers didn't wait for any of this to resolve. They built around it.

Oracle and OpenAI's Stargate campuses run on behind-the-meter natural gas, deliberately structured to avoid grid dependency from day one. Elon Musk's xAI drew the most attention in 2025 by trucking mobile generators into its Memphis Colossus facility to beat the grid queue, but the move attracted regulatory blowback in early 2026 when the EPA closed the permitting loophole that had let the turbines run without air quality permits.

The company is now dismantling them as it transitions to permanent grid power while planning a 1.2 GW on-site plant. EdgeConneX has proposed a 430 MW natural gas plant in New Albany, Ohio, solely to power its adjacent campus. According to Cleanview's 2026 analysis, roughly 50 GW of behind-the-meter generation capacity was announced by data center developers in 2025 alone.

The nuclear picture is shifting too, though more slowly than the policy language suggests. The Trump administration's four executive orders on nuclear energy, signed in May 2025, set a headline target of criticality in three advanced reactors by July 4, 2026. DOE selected eleven projects under the pilot program in August 2025, with three having since broken ground.

Whether any reaches criticality by Independence Day is openly contested. NPR reported in January 2026 that DOE had been quietly rewriting safety rules to compress the timeline, drawing criticism from former regulators. The longer-term ambition to quadruple US nuclear capacity to 400 GW by 2050 is a policy direction worth watching, even if the construction pipeline lags well behind it.

What happens when the largest companies in the world start building their own power plants alongside their core projects? The scope of what a GC is actually being asked to deliver expands considerably. We covered this shift earlier this year, and the core finding holds: a self-powered campus is a co-located multi-project, with a data hall, generation facility, and battery storage all built simultaneously, each carrying different procurement requirements, different subcontractor pools, and different schedule dependencies. The GC capability set required is materially different from what most firms were built for.

The grid constraint touches every role in the delivery chain differently. Here's where it's changing the job.

Energy strategy has to enter the project development process far earlier than it ever did. Site selection is increasingly governed by power availability before land cost, logistics access, or labor proximity enters the conversation. The owners moving fastest on this have specialist energy consultants in the room before the architect is appointed. Power strategy goes into the project brief before the RFP goes out.

The timing question is sharper than most firms have adjusted for. Transformer procurement has traditionally happened during design development. At 128-week average lead times, that's no longer viable for any project with a realistic delivery target. Procurement decisions have to be made during pre-design, before many downstream design decisions have been resolved.

That requires early supply chain engagement most firms weren't structured for three years ago. As we've tracked in our analysis of the data center pipeline, the gap between projects announced and projects actually under construction has grown sharply, and power availability is a significant driver.

When the project brief includes on-site generation, the owner's table now includes power purchase agreement specialists, utility regulatory lawyers, and energy infrastructure engineers who historically sat nowhere near a construction project team. How those groups integrate with the delivery schedule, who leads them, and where accountability sits for power infrastructure decisions are questions most teams are still resolving in real time.

Is the industry adapting? Yes, at the edges. The firms that have moved early are building new capability, new relationships, and new procurement infrastructure around this. But the broader market is still catching up to a constraint that was, until very recently, somebody else's problem.

The question for any executive with large-load projects in the pipeline is straightforward: where does power planning sit in your development sequence? If the answer is "somewhere in detailed design," the schedule is already compromised.

POWERED BY: