Electrical work accounts for 45% to 70% of what a data center costs to build, which has turned journeyman electricians into the binding constraint on the AI buildout. The same crews were previously wiring hospitals, schools, factories, and apartment blocks. If you are bidding any of those, you are now competing against hyperscaler budgets for labor, and the effect arrives as a wage line and a schedule slip rather than an invoice with an explanation attached.
Microsoft’s president has said publicly that the electrician shortage is the single biggest obstacle to the company’s US data center expansion, and that they have resorted to flying electricians in from 75 miles away or relocating them temporarily to keep projects moving. When a company with that balance sheet describes a trade shortage as its number one problem, the downstream effect on everyone else is not subtle.
Most contractors outside the mission-critical world have felt this without naming it. The electrical sub who used to be available is now booked eighteen months out. The quote came back higher than last year’s equivalent job by more than materials inflation explains. The superintendent you wanted has taken a role on a campus two states over.
Here is what is actually happening to the labor pool, and what the firms handling it well are doing differently.
Why the pressure lands on one trade
A data center is an electrical project that happens to have a building around it
The IBEW estimates that electrical systems account for 45% to 70% of total data center construction cost. Medium-voltage switchgear, paralleling switchboards, UPS systems, generator paralleling, battery storage. Every other trade on the job is sequenced behind the electrical scope, so when electricians are short the whole program stops rather than slowing.
The supply side is moving the wrong way. The US employed about 818,700 electricians in 2024 and needs roughly 81,000 more each year through 2034, while close to 30% of union electricians are at or near retirement age and about 20,000 leave the trade annually. Against that, AI data centers alone could require more than 300,000 additional electricians this decade.
There is a second filter that makes the real pool smaller than the headline number. General commercial wiring does not have to survive a commissioning test. Redundant, high-precision electrical infrastructure does, which narrows the qualified population well below the count of licensed electricians in any given market.
What it does to a bid you are working on right now
The cost arrives as a wage premium you cannot see itemized
Data center work carries a wage premium reported at up to 30%, and Turner and Townsend’s cost index found labor costs across data center sites in primary North American markets rising 8% to 12% year on year, driven almost entirely by trades scarcity. That premium does not stay inside the data center sector. It resets what an electrician in that metro expects to earn, on every job.
So a school refurbishment in Northern Virginia or a hospital expansion in Phoenix is now priced against a labor market that a hyperscaler is setting the ceiling for. As BRG’s analysis of the bottleneck points out, data center developers are not the only ones competing for these crews, with municipal water and sewer, highway work, hospital expansions and factory builds drawing from the same pool.
The schedule exposure matters more than the rate. A crew that leaves mid-project for better money on a campus nearby does not just cost you the hourly difference. It costs the sequencing, the rework from a replacement crew learning the job, and whatever your contract says about delay. That is where thin margins actually get eaten.
- Sub quotes. Electrical packages priced above general materials inflation, with shorter validity windows.
- Crew retention. Mid-project departures toward mission-critical work paying a premium.
- Field leadership. Superintendents with commissioning experience are scarcer than the crews themselves.
- Long-lead equipment. Large transformers running two years or more, which pushes staffing decisions earlier.
Where it bites hardest
This is a market-by-market problem, not a national average
More than 400 data centers were in active development as of late 2025, and JLL’s outlook has the sector adding roughly 97 gigawatts of capacity between 2025 and 2030. That capacity is not spread evenly. It clusters in a handful of metros with power, land, and tax treatment, and those are the markets where the local trade base gets stretched past what it can absorb.
The clearest illustration comes from the DMV region, where a local electricians’ union roughly doubled its membership between 2018 and early 2026 to reach 14,700 members, and demand still outruns supply. Doubling a union local inside eight years is an extraordinary pipeline achievement, and it has not been enough.
For anyone planning work in a data center metro, labor availability now belongs in the feasibility conversation next to power and permitting. Bidding a 24-month program in a market where comparable projects have slipped to 36 is a pricing decision, whether or not anyone writes it down as one.
What the firms handling it well are doing
Three responses that show up repeatedly, and one that does not work
The first is moving scope off site. Prefabricating electrical skids, cooling modules, and power rooms in a factory reduces on-site labor demand by 20% to 40% according to Schneider Electric’s analysis, which is the single biggest lever available and the only one that changes the total labor requirement rather than redistributing it.
The second is benchmarking pay against real local data rather than last year’s rates, and building recruitment pipelines that do not dry up mid-project. The third is automation on the scopes where it is already proven, which is a longer game but increasingly a live one. We covered what that actually looks like on site in our piece on where construction robots are actually working.
The response that does not work is assuming the boom will cool and the crews will come back. Retirements are the durable driver here, not the AI cycle. ABC projects the industry needs 349,000 net new workers in 2026 rising to 456,000 in 2027, and most of the 2026 gap comes from people leaving rather than new demand. Even a sharp slowdown in data center starts leaves that demographic problem exactly where it is. Our broader take on that sits in whether AI will replace construction workers.
Track the forces moving construction labor and cost
Weekly analysis for people who have to price the work, not just read about it.
Join 3000+ Readers
Four responses to the squeeze, compared
| Response | Effect on labor demand | Main limitation |
|---|---|---|
| Prefabrication and modular assemblies | Cuts on-site labor demand 20% to 40% | Requires design decisions early and factory capacity secured |
| Pay benchmarking against local market data | Improves retention, no change to total supply | Directly compresses margin on fixed-price work |
| Building an apprenticeship pipeline | Adds real supply over several years | Too slow to fix a project starting this year |
| Automation on proven scopes | Removes hours from specific repetitive tasks | Narrow applicability; electrical rough-in is mostly not yet automatable |
Estimates put the additional requirement at more than 300,000 electricians this decade for AI data centers alone, against a base of about 818,700 employed in 2024 and roughly 20,000 leaving the trade each year. Federal projections have the trade growing 6.6% between 2023 and 2033, faster than any other construction category tracked, and still short of demand. (Source)
Yes, through wages and availability. Municipal utilities, highways, hospitals, and factory builds draw from the same electrician pool, so a hyperscaler campus in your metro resets what crews expect to be paid on every job in that market. (Source)
Turner and Townsend’s data center cost index found labor costs in primary North American markets rising 8% to 12% year on year, attributed almost entirely to skilled trades scarcity, with a reported wage premium of up to 30% for mission-critical work. (Source)
Only partially. ABC’s analysis attributes most of the 2026 gap to retirements rather than new demand, so a slowdown in data center starts would relieve the wage pressure without fixing the underlying demographic shortfall. (Source)
Prefabrication is the clearest lever, cutting on-site labor demand by 20% to 40% by building electrical skids, cooling modules, and power rooms in factories. It requires committing to design decisions much earlier than most teams are used to, which is the main reason it gets skipped. (Source)
It increasingly does. Analysts now argue labor availability belongs in underwriting alongside power, land, water, and fiber, because a market whose project pipeline outpaces its skilled trade base will produce schedule slippage regardless of how good the site is on every other measure. (Source)
Related reading
Will AI Replace Construction Workers?
Construction Robotics in 2026: 4 Workflows Where Robots Are Actually Working
What Is Construction Technology (ConTech)?
BRG ThinkSet, The Data Center Labor Shortage
Buildforce, The Electrician Shortage in 2026 (BLS and IBEW data)
Westside Construction Group, ABC Workforce Projections for 2026
Build Inc, Data Center Labor as a Site Selection Constraint (JLL capacity data)
DCGeeks, Data Center Workforce Shortage (Turner and Townsend cost index)
