The Data Center Pipeline Is $78.7B of "Not Yet" — and It's Still Pricing Your Bid
$78.7B in data center construction is still unstarted preconstruction — and AGC says the labor pool it's absorbing is tightening bids on projects with no data centers in sight.
Two numbers from this month's data center reporting are worth sitting with together, because most coverage only runs one of them. The first: $84.1 billion in data center construction has already broken ground this year through July — nearly three times 2025's pace, and roughly a quarter of all nonresidential building starts in the country (ConstructConnect). The second: another $78.7 billion across 85 projects is sitting in preconstruction, tracked for a possible year-end start, with no guaranteed groundbreaking attached to any of it.
That second number is the one worth pricing your 2027 bids against.
What actually happened in July
July alone added $2.7 billion across 23 data center projects, the largest a $1.5 billion CyrusOne campus in Texas. ConstructConnect's own comparison shows average project cost and footprint both climbing sharply against last year — $1.06 billion and 701,000 square feet per project in 2026's year-to-date figures, versus $352.8 million and 454,400 square feet in 2025. The same report puts the resulting cost-per-square-foot at $818.20, which it describes as 57% above the comparable 2025 figure — a steep number, and one that's worth treating as a directional signal from the source rather than a figure to re-derive yourself, since the underlying per-project average cost and footprint don't cleanly reduce to that exact rate. Directionally, the point stands regardless of which exact per-square-foot number you trust: this is a market getting more expensive per unit, not just bigger in aggregate.
North Carolina leads all states at $10.4 billion in year-to-date spending, with Texas close behind nationally and a Midwest cluster — Indiana, Illinois, Michigan — each clearing $10 billion.
Why a GC with zero data centers in backlog should still care
Here's the transmission mechanism, and it's not abstract. AGC's chief economist Ken Simonson, in a September 3 release on the association's latest workforce survey, put it directly: "The need for people to work on new data centers is keeping labor conditions tight even as demand for many other types of projects remains relatively soft." Workforce shortages remain the top reason cited for construction project delays industry-wide, and nearly a third of firms report federal immigration policy affecting their workforce on top of that.
Read plainly: your electricians, your controls subs, your commissioning agents, and your MEP trades are pricing their labor against a market where a hyperscale campus two counties over will pay whatever it costs to hit an energization date. You don't need a single data center in your pipeline to feel that bid. It shows up as a tighter labor market and slower trade availability on a warehouse, a hospital wing, or a mid-rise you're bidding with nothing to do with AI capex.
The skeptical read
Nearly as much data center spending sits unstarted as has already broken ground. That's not a criticism of the sector — preconstruction pipelines are normal — but it's a real number to hold against today's tight-labor assumption. This demand is concentrated in a small number of hyperscale buyers whose capital plans are tied to AI infrastructure cycles that can move fast in either direction. If even a fraction of that $78.7 billion slips, delays, or doesn't convert, the trade labor it's currently absorbing loosens up in specific regional markets — and a bid priced today against a 2027 labor assumption built on the current squeeze could be pricing the wrong market by the time it breaks ground.
The move isn't to ignore the data center numbers because they don't touch your backlog. It's to treat the preconstruction column, not just the broken-ground column, as the number that tells you how long the current labor premium actually lasts.