Gene Lantrip has been building homes in Abilene, Texas, for decades. He knows how to read a subcontractor schedule, manage a lumber yard relationship, and absorb the routine indignities of residential construction without losing his mind or his margins. What he cannot do is match the paycheck that a 4-million-square-foot AI data center is waving at his electricians from the other side of town. Nobody can.
"My subcontractors don't have the people," Lantrip told the Texas Tribune earlier this year. "My electrician lost two of his lead men and several of his helpers to the data center." The data center in question is Stargate, backed by OpenAI, Crusoe, and Oracle. Twice what Lantrip's subs can offer. His homes now take two months longer to finish than they did before Stargate broke ground.
On a six-month build, that is a 33 percent schedule overrun caused not by weather, not by permit delays, not by any of the familiar friction that builders learn to absorb over a career, but by a building that has nothing to do with housing and everything to do with the race to train the next generation of large language models.
The Numbers That Should Alarm Residential Builders
Federal Reserve data on private fixed investment tell a story that has no precedent in six decades of American economic data. Spending on computing equipment, driven overwhelmingly by the AI data center buildout, has surged from approximately $150 billion to nearly $400 billion in under two years. Single-family residential investment has stayed essentially flat at around $400 billion over the same period. Two categories of investment that spent the entire postwar era running on separate tracks, with homebuilding comfortably ahead, are now converging. That has never happened before.
Monthly Census Bureau data put it more starkly. Data center construction spending rose 23 percent year-over-year in May 2026, and data centers now account for 8 percent of all private nonresidential construction spending. That might sound modest until you consider what is happening to everything else: manufacturing construction dropped 22 percent year-over-year in May. Anirban Basu, chief economist for the Associated Builders and Contractors, put it bluntly: "Beyond data centers, there's not really much moving construction forward."
Single-family housing starts fell 11 percent through April 2026, according to Dodge Construction Network and the ENR 2Q 2026 Cost Report. By May, they had slid to an eight-month low of 893,000 units at a seasonally adjusted annual rate. Completions dropped to 872,000 — the lowest in six years — while the National Association of Home Builders projects a national shortage of roughly 1.2 million homes and rising.
Not enough homes, too many data centers chasing the same electricians and framers and concrete crews who would otherwise be building them.
The Electrician Is the Bottleneck
Electrical subcontractors account for between 45 and 70 percent of data center construction budgets, according to the International Brotherhood of Electrical Workers. A single Microsoft data center in Chicago required 2,177 tons of copper, per the Copper Development Association. These buildings are electrical projects that happen to have walls. That is all they are.
That demand has rewritten what electricians expect to earn. Dramatically. Three electricians under 30 working on a data center site in Plano, Texas, reported annual compensation between $240,000 and $280,000 to CNBC. Data center projects are paying 32 percent more than residential work on average, per Randstad's latest skilled trades report, and the gap between residential and commercial electrician pay has widened to 18 percent as of 2026.
Half a million new construction workers: that is what the industry needs in 2026 alone, according to Randstad, at a moment when data centers are absorbing a disproportionate share of every electrician, welder, and concrete finisher who enters the labor market. Demand for construction workers overall has climbed 30 percent in four years, with electrician demand up 18 percent and welder demand up 25 percent.
Meanwhile, the other end of the pipeline is constricting. A third of U.S. construction firms reported being affected by immigration enforcement in the last six months, according to the Associated General Contractors of America, and nearly a quarter saw subcontractors lose workers directly. Thirty percent of U.S. construction workers are immigrants, per the National Immigration Forum, and the pool is simultaneously shrinking from enforcement on one end while data centers vacuum it up on the other. Something has to give.
What This Costs Per Home
Nobody publishes this number, not NAHB, not the Census Bureau, not a single trade journal that covers the intersection of residential construction and the data center boom. We calculated it using publicly available data and conservative assumptions for markets within commuting distance of active data center construction.
| Cost Channel | Inputs | Per-Home Impact |
|---|---|---|
| Electrical wage premium | 200–250 labor hrs × $8–12/hr premium to retain workers against data center poaching | $1,600–$3,000 |
| Schedule delay carrying cost | 2-month avg delay × 7% builder financing on $420K median new home | ~$4,900 |
| Utility rate pass-through | $11.24/mo Dominion Energy increase (VA SCC, Jan 2026), attributed to grid buildout for data centers | $135/year |
| Total first-year impact | $6,635–$8,035 |
Start with the wage premium, which is the most conservative line. We assumed builders raise electrical rates $8 to $12 per hour to retain workers, which is less than the full $10-per-hour jump that industry sources describe in the most affected markets, because not every community faces Abilene-level pressure. The schedule delay, however, is direct testimony from a builder in the field: two extra months of carrying cost on a construction loan at prevailing rates, passed through to the buyer as a higher closing price.
Utility costs are the most insidious line. They never stop. The Virginia State Corporation Commission approved that $11.24-per-month base rate increase for average residential customers in January 2026, driven substantially by grid infrastructure buildout for data center demand. SB 253, which would have shifted more costs onto data centers, was estimated to save households $5.52 per month. That delta represents how much residential ratepayers currently subsidize the AI buildout through their electric bills.
Now scale it. If 15 percent of the 893,000 single-family starts sit in data-center-proximate markets, approximately 134,000 homes are absorbing $6,600 to $8,000 each in excess costs that exist because of a building type that produces zero housing units. Across those homes, the annual toll runs between $890 million and $1.08 billion.
It Is Not Just Labor
In Northern Virginia, data center developers pay $6.3 million to $8 million per acre, according to CBRE's H2 2025 data center trends report. SDC Capital paid $615 million for 97 acres in Leesburg in November 2025, land that had been assembled for roughly $57 million four years prior. In the Ashburn area, a developer approached all 143 homeowners in the Regency subdivision about a buyout valued at approximately $4.4 million per acre, or $576 million total, to raze the houses and build a data center campus.
"I don't think a residential developer can compete," said Arif Gasilov of the Gasilov Group. He is right.
He is being polite, and the understatement obscures a structural shift in how American land gets allocated: when a single asset class can pay ten times what a homebuilder can for the same dirt, the invisible hand does not negotiate, it simply redirects every vacant parcel within commuting distance of a fiber trunk line away from housing and toward servers. Natelli Communities, a family-owned Mid-Atlantic developer with more than 40 years of building master-planned residential neighborhoods, has pivoted toward data center development. CEO Tom Natelli invested in Quantum Loophole, which developed a gigawatt-scale data center park in Maryland, and then proposed campuses in North Carolina and Calvert County, Maryland, though both hit community resistance. When a residential developer pivots to data centers because the economics are better, that is not diversification — it is a market signal that should concern anyone waiting for housing supply to catch up with demand.
PulteGroup, America's third-largest homebuilder, found a different answer. PulteGroup partnered with Nvidia and startup Span to install small "fractional data centers" on the exterior of newly built homes, using each home's underutilized electrical capacity to run distributed AI compute. Homeowners get discounted electricity and internet. Clever. PulteGroup gets to stay in the housing business without losing the data center fight entirely.
The Strongest Case Against This Argument
Data center construction is not permanent. These are finite projects with two-to-three-year timelines, and the labor market will eventually adjust. Probably. Bernstein Research estimates 35 to 40 percent of announced global data center capacity is at risk of delay or cancellation because of power grid constraints, transformer lead times stretching to 80 to 100 weeks, and cooling infrastructure bottlenecks. Maybe the surge slows, and maybe the labor market finds a new equilibrium before the damage is permanent.
Data centers also produce tax revenue, bring permanent operational jobs, and the workers they employ need housing, which supports the very builders they are currently disrupting; Abilene's Stargate facility alone will employ maintenance staff, security, and technicians for decades after construction ends, generating the kind of local spending that lifts a housing market.
More fundamentally, the real constraints on housing are mortgage rates and zoning, not labor competition. Single-family starts were declining before the data center surge hit full stride, driven by 7-percent-plus mortgage rates and consumer affordability pressure. "Rising mortgage rates between January and April have continued to price out first-time homebuyers," noted Sarah Martin of Dodge Construction Network. Data centers made a bad situation considerably worse. They did not create the housing crisis.
Fair enough. But tell that to the builder in Abilene whose electrician walked out Tuesday and whose buyer will not close until September.
What You Should Do About It
If you are a builder within 60 miles of active or announced data center construction: lock subcontractor pricing earlier in the project cycle, before bids reflect the premium. Build six-to-eight-week schedule buffers into electrical rough-in phases. Investigate prefabricated electrical panel and wiring harness systems that reduce on-site labor hours. Every hour you eliminate from the field is an hour you do not have to compete for. Track your local utility rate filings, because grid infrastructure being built for data centers is being financed, in part, by your customers' monthly electric bills.
If you are buying a home in one of these markets, your closing price includes a surcharge nobody will itemize on the settlement statement. It is embedded in the builder's carrying costs, the electrician's higher rate, and the grid infrastructure your bill subsidizes. Depending on your market, it is somewhere between $6,600 and $8,000 that has nothing to do with the house you are standing in.
Limitations of This Analysis
The per-home cost calculation relies on data from specific markets, particularly Abilene, Texas, and Northern Virginia, which represent the most acute pressure points and may not generalize to all data-center-proximate communities. The two-month schedule delay is testimony from a single builder; delays elsewhere could be shorter or longer depending on local labor supply, project scale, and which trades are most affected. Our 15 percent estimate of single-family starts in proximate markets is a rough approximation based on Census data for metro areas with the largest active data center pipelines. Utility rate impacts vary by state and regulatory structure; the Virginia SCC figure is among the most documented but may not apply where rate-setting works differently. The wage premium calculation assumes builders must raise rates to retain workers, but some may absorb longer schedules without adjusting pay, which shifts costs from direct labor to carrying overhead. We did not model the second-order effects of land competition, which Gasilov identifies as a separate and potentially larger cost channel. Immigration enforcement data from the National Immigration Forum and AGC reflects conditions through early 2026 and may already be stale given the pace of policy change.