Data Centers Grew 46% This Year. Housing Construction Shrank 3.2%. That's Not a Warning Sign — It's Your Moat.

The construction industry just told you exactly where its money is going — and it isn't houses.

The Census Bureau released its June construction spending report on August 3. Buried in the topline number — total spending down 3.2% year over year — is a split so wide it should change how you think about the next three years of housing supply. Data center construction grew. Everything that competes with your rental for a buyer's dollar shrank.

This isn't a "why are housing starts weak" story. Every housing blog runs that one. This is a story about where the labor, the capital, and the concrete went instead — and why that's the best supply news existing landlords have gotten all year.

Nobody's connecting these two numbers because they live in different sections of the same report. Put them next to each other and the picture is obvious: the industry that builds your competition is starving itself to feed AI infrastructure.

Every cycle has a story about where construction capacity goes when it doesn't go into homes — condo conversions in the 2000s, industrial and warehouse space after 2020. This cycle's version is bigger and stickier than either of those, because data center leases run 15 to 20 years and the campuses are being built to match. That's not capacity that comes back to housing next quarter when rates ease. It's capacity that's been reassigned for a decade or more.

The industry is choosing data centers over your competition

Here's the mechanism, not just the headline. Construction crews, framers, electricians, and concrete crews are a fixed pool. When a hyperscale data center campus breaks ground, it doesn't add workers to the market — it pulls them off other job sites. Marketplace reported that workers are relocating from markets like Arizona, where power constraints have slowed data center construction, to booming build zones like Dallas, chasing the wages data center general contractors are paying to hit deadlines.

The scale of that pull is the part most investors haven't priced in. DataBank's own construction outlook notes that peak crew sizes on a data center site once topped out around 750 workers. Its Red Oak campus is on track to hit 4,000 to 5,000 workers by early 2026 — "the size of a small city," built to run servers, not house families. Every one of those trades could otherwise be pouring foundations for single-family subdivisions or framing garden-style apartments. Instead they're wiring server halls.

That's not a coincidence showing up once. It's showing up in the spending data every single month this year.

The Numbers

The Census Bureau's June 2026 release put total construction spending at a seasonally adjusted annual rate of $2,166.5 billion — 3.2% below June 2025, and 3.5% below the same six-month stretch last year. Residential construction came in at $877.1 billion SAAR, down another 0.3% from May. That's the fourth straight soft print for housing.

Now look at what's absorbing the capital that isn't going into homes. Construction Dive reported that data center construction spending rose 7% in June alone and 46% year over year — even as private nonresidential spending excluding data centers fell 0.6% for the month and is down 7.9% year over year. ABC chief economist Anirban Basu's own numbers show contractors with data center work in their backlog are carrying 11 months of it, 2.5 months more than firms without that work. That's a queue. Firms are turning down other jobs to protect it.

Translate that into plain English: the same trades that would otherwise be adding rental supply are booked out for the better part of a year building server farms, and getting paid enough to relocate across state lines to do it.

Common Mistakes Investors Make Here

• Reading "construction spending is down" as bearish for housing values. It's bearish for future supply, which is bullish for the rent roll and resale value of what you already own.
• Assuming the labor shortage is about immigration policy or retirements alone. Those are real, but they're being amplified by a sector actively bidding trades away from residential jobs right now, not in some abstract future.
• Ignoring where the data center campuses actually are. Dallas, Northern Virginia, and parts of the Southeast are absorbing the heaviest construction-labor pull — those are also markets where you'd expect new-home competition to shrink fastest.
• Waiting for a "construction rebound" before underwriting rent growth. If trades are locked into 11-month data center backlogs, that rebound isn't a 2026 story. Underwrite the gap, not the recovery.

Where I Land

I'd rather own rental property in a market where builders can't get a framing crew than one where they can. Every dollar and every worker going into a server hall is a unit of future competition that doesn't get built to compete with my listing. The "construction is collapsing, that means trouble" read has it backwards — it's not the credit cycle killing housing supply, it's a bidding war for labor that housing is losing on purpose, because AI infrastructure pays better and faster. I'm not betting on a supply rebound before 2028; the backlog data says the trades are already spoken for. By Q2 2027, I'd expect the metros with the sharpest permit declines and the heaviest data center buildout today — Dallas leads that list — to show rent growth running at least double the national average, because nobody built the apartments that were supposed to compete with the ones already standing.

How to Use PropGPT for This

"Show me building permit volume for the last 12 months in [metro], broken out by zip code, and flag any zip code where permits are down more than 15% year over year but population or household growth is still positive." This finds the specific submarkets where the supply pipeline has gone quiet while demand hasn't — the exact combination this story predicts will show up as rent growth first.

"Pull comparable rent trends for [metro] over the last 18 months and cross-reference them against new construction permit data for the same zip codes — where is rent growth accelerating while permits are falling?" This turns the macro thesis into a screen you can run market by market instead of trusting a national headline.

"List the housing stock age distribution for [metro] — what percentage of the rental stock is over 20 years old, and how does that compare to current permit volume for new units?" Older stock plus a thin permit pipeline is the setup for pricing power on renovated or well-maintained units with no new competition coming.

"Compare [metro A] and [metro B] on population growth, building permit trends, and known large-scale data center or industrial construction projects announced in the last 12 months." Use this before you commit capital to a market — it tells you whether local construction capacity is being pulled toward something other than housing.

The Bottom Line

The construction industry isn't broadly "slowing down" — it's re-routing. Every dollar and every crew that goes into a data center campus is a dollar and a crew that isn't building the duplex or the garden apartment complex that was supposed to undercut your rent in 2027. The Census numbers and the industry backlog data both say this is already happening, not a forecast. If you own rental property in a market where that labor pull is strongest, you're not facing a supply glut — you're facing less competition than the headlines suggest. Run the permit-versus-population screen on your market this week before you assume the "housing crash is coming" story applies to you.