Your Electrician Quit on Friday. By Monday He Was Wiring a Data Center for 30% More.

The panel box on Lot 7 has been open for three weeks. Romex dangles from the joists like party streamers nobody cleaned up. The framing crew finished ten days ago. The insulation crew is scheduled for Monday. The electrician who was supposed to rough-in this house quit two Fridays ago. He’s pulling cable at a Microsoft data center campus outside Richmond now.

He got a 30% raise to do it.

The Numbers Behind the Exodus

The U.S. construction industry is short 439,000 workers, according to the Information Technology and Innovation Foundation. That number has a specific accelerant: more than 400 data centers are currently under development by Amazon, Google, Microsoft, and their peers, each one demanding thousands of construction workers and paying substantially above market to get them.

The Wall Street Journal reported in November 2025 that data center construction jobs pay up to 30% more than equivalent roles on conventional projects. The BLS pegs the national median electrician wage at $62,350 per year, or about $30 an hour. A 30% premium lands at $39 an hour. In hot data center corridors like Northern Virginia, central Texas, and the Phoenix metro, the gap widens further. Electricians working data center projects report premiums of 25–30% with overtime pushing annual earnings past six figures.

Your residential builder cannot match those numbers. Not without eating margin on a product category where net margins already run 6–9% on a good day. The data center general contractor has a client—a tech company with a $50 billion annual capex budget—who treats construction labor as a rounding error. Your builder has a buyer with a $425,000 mortgage pre-approval who will walk if the price goes up $8,000.

Who Actually Left

It isn’t the apprentice. The first-year wire puller doesn’t have the credentials for data center work, which typically requires journeyman-level licensing and often specialty certifications in medium-voltage systems, fiber optics, or fire alarm. The worker who leaves is the experienced journeyman or foreman—the person your sub most needs and least easily replaces.

The IBEW local serving the Washington, D.C., Maryland, and Virginia corridor has doubled its membership to 14,700 since 2018. That growth tracks the Northern Virginia data center boom almost perfectly. Ashburn, Virginia, processes an estimated 70% of the world’s internet traffic. The electricians who wire those facilities used to wire houses in Loudoun County. Some still do. Increasingly, they don’t.

The Amtec Staffing benchmark report from February 2026 names the specific trades being siphoned: electricians, welders, HVAC technicians. These are the same workers who rough-in and trim out a residential build. When the report says the data center boom is “pulling the highest-skilled trades workers away from other projects,” the “other projects” includes the house you’re waiting on.

What It Costs Your Build

Here’s the math your builder won’t show you.

A typical residential electrical subcontractor runs a crew of three to four on a subdivision. Lose one journeyman to data center work, and you’ve lost 25–33% of your roughing capacity. The remaining crew absorbs the load, but sequencing breaks: the framing-to-rough-in handoff stretches from three days to eight. Insulation, drywall, and every downstream trade shift with it.

The delay cost uses a $670-per-day figure derived from carrying costs, superintendent time, and weather exposure on a median-price new single-family home. A 10–20 day slip per house across 20 lots generates $134,000–$268,000 in absorbed cost. Matching the data center wage premium for a three-person crew costs roughly $75,000 per year. (The math: average construction hourly earnings of $40.55 × 30% premium = $12.17/hr additional × 2,080 annual hours × 3 workers = $75,942.) That’s one-quarter to one-half the delay cost. The math says match the wage. The margin says you can’t.

The Compounding Problem

Ninety-two percent of construction firms hiring in 2025 reported difficulty finding qualified workers, according to the AGC’s annual workforce survey. Forty-five percent said labor shortages were the leading cause of project delays. The Associated Builders and Contractors estimates the industry needs 349,000 net new workers in 2026, rising to 456,000 in 2027.

Meanwhile, one in five construction workers is over 55. They’re not leaving for data centers. They’re leaving for retirement. Immigration enforcement has affected 28% of construction firms, constricting a labor pool where foreign-born workers represent roughly a quarter of payroll employment. Data center poaching is a third drain running simultaneously with retirement and immigration disruption.

Three labor crises hitting the same workforce at the same time.

The Counterargument

Data center construction may be cyclical. Tech capex runs in waves—the dot-com build-out peaked and crashed, the cloud expansion of 2015–2019 leveled off. If AI infrastructure spending follows a similar curve, the current boom could plateau by 2028 or 2029, flooding the market with skilled workers who return to commercial and residential work. The IBEW’s doubling membership suggests the pipeline is expanding. BlackRock’s $100 million Future Builders initiative, announced in March 2026, targets 50,000 new skilled trades workers over five years—explicitly tied to AI infrastructure demand.

The argument deserves its full weight: the labor market may self-correct. Higher wages attract more entrants. Union apprenticeship programs are growing. Investment capital is flowing into trades training at a scale not seen in decades.

But “eventually” doesn’t help the builder who needs an electrician in April. The houses under construction right now are absorbing the cost of a correction that might arrive in three years.

What a Builder Can Actually Do

Some builders are locking in electrical subs with multi-project commitments—guaranteeing a year of work in exchange for crew priority. Others are investing in pre-fabricated electrical assemblies: panel boards pre-wired in a shop, delivered to the site, bolted in. Pre-fab shifts the skill requirement from a journeyman pulling wire through joists for three days to a helper hanging a pre-built panel in four hours. It doesn’t eliminate the electrician—the final connections and inspection sign-off still require a licensed hand—but it reduces the hours of journeyman labor per house by 30–40%.

AI-powered scheduling tools (Buildots, ALICE Technologies) can optimize sequencing to minimize the damage of a missing crew, reordering tasks so other trades fill the gap without sitting idle. But those tools assume you have the workers to reschedule. When the constraint is bodies, not sequencing, the algorithm hits a wall.

What This Analysis Couldn’t Verify

No published survey tracks how many electricians have left residential construction specifically for data center work. The wage premium data is aggregate—data center construction versus all other construction, not residential-to-data-center comparisons. The 30% figure from the WSJ and ITIF reporting represents the upper range; some workers may be moving for 15–20% premiums plus overtime availability rather than base rate alone.

The $670-per-day delay cost comes from our own prior reporting on project management delays and applies to median-price homes in mid-range markets. In starter-home developments with lower carry costs, the figure is lower. In custom builds with bridge financing, it’s substantially higher.

Regional variation is enormous. In markets far from data center corridors—the Upper Midwest, the interior Southeast—the poaching effect is weaker. In Northern Virginia, central Texas, and the Phoenix-Mesa area, it’s acute. National statistics obscure what is fundamentally a geography story.

Finally, I could not independently confirm whether pre-fab electrical assemblies reduce journeyman labor by the 30–40% figure cited by manufacturers. That number comes from vendor claims. Independent verification from a general contractor or trade association would strengthen it.