One building type is holding up the whole nonresidential market. Data center construction spending reached $58.1 billion year-to-date through May, more than four times the figure logged over the same months in 2025. May by itself contributed $7.9 billion, the eighth-highest month on record for the category.
Set that against everything else and the split is stark. Nonresidential construction ran at roughly $730 billion on an annualized basis in the spring and barely moved month to month, slipping a touch from the prior reading. Offices, retail and a lot of institutional work are flat to soft. Data centers are not.
What the spending split is telling contractors
The concentration is now impossible to ignore. A market that looks merely stable in aggregate is actually two markets: a booming one tied to AI and cloud capacity, and a stagnant one tied to almost everything else. Contractors with the certifications and bonding to chase hyperscale work, the kind going up at campuses like Stargate in Abilene, are booked. Firms anchored in commercial or smaller institutional jobs are competing harder for fewer starts.
That has knock-on effects. Power and cooling subcontractors, switchgear suppliers and concrete producers near major data center clusters are running tight, which pushes lead times and pricing up for any other project drawing from the same pool. The boom doesn’t stay neatly inside its own sector.
The risk in leaning on one engine
Single-sector dependence cuts both ways. Data center demand rests on a handful of well-capitalized owners and a set of assumptions about AI compute that could shift. If even a few of the announced campuses slip or shrink, the growth that’s masking weakness elsewhere thins out fast. For now the pipeline is deep, with tens of billions in near-term projects still queued, but a market this lopsided is more fragile than the headline total suggests.
The takeaway for planners is less about celebrating the number and more about reading what’s underneath it, as the latest data center spending report makes clear. Healthy totals built on one category aren’t the same as broad demand.