Data Center Construction Costs Are Spiking: Who Pays?

The average U.S. data center project now costs $1.18 billion, up from $507.4 million a year ago — a 2.3x jump that has nothing to do with projects suddenly getting twice as ambitious. According to ConstructConnect’s October 2026 construction-starts tracker, $98.5 billion in data center construction broke ground between January and August 2026 alone, and the average facility grew only 43% larger over the same period. The rest of that cost increase is unit-price inflation: copper, switchgear, transformers, and steel getting more expensive faster than anyone budgeted for. For a buyer signing a build-to-suit agreement or a guaranteed-maximum-price (GMP) contract this year, that gap between “bigger building” and “bigger bill” is exactly where risk gets quietly reassigned — usually onto whoever didn’t think to ask who holds it.

The numbers behind the spike

Cushman & Wakefield’s 2026 Data Center Development Cost Guide puts construction costs up 21% per megawatt since late 2024, against roughly 5% general inflation over the same stretch. All-in greenfield costs in the U.S. and Canada now run $8.9 million to $23.3 million per MW, excluding the GPUs and AI chips going inside — and power infrastructure alone accounts for 21% of that spend, ahead of core-and-shell construction and sitework at 17%.

The equipment driving the increase has repriced hard since December 2021: switchgear is up roughly 60%, copper 46%, generators and motors 33%, cooling equipment 32%, and transformers 29%, per the same Cushman & Wakefield guide. Turner & Townsend’s 2025 Data Center Cost Index shows why that matters so much: electrical systems alone make up 48% to 54% of total construction cost, with mechanical systems adding another 22% to 33%. When the line items that dominate the budget are the ones inflating fastest, there’s no averaging your way out of it. ConstructConnect’s own trimmed average cost per square foot climbed to $867 in its October report, up from $818 just one month earlier in September.

Why equipment, not labor or land, is the real driver

Large power transformers — the single longest lead-time item in most builds — now take 30 to 36 months to deliver, up from 12 to 18 months before the current shortage, according to a May 2026 DOE-funded national laboratory analysis. Prices for the units themselves are up 75% to 77% since 2019, with distribution transformers up 78% to 95% over the same period. We’ve covered the transformer bottleneck itself in detail, but the manufacturer-side numbers underline why it isn’t easing soon: Siemens Energy’s Grid Technologies division reported a record €51 billion backlog in August, GE Vernova’s backlog sits around $176 billion, and Eaton’s Electrical Americas segment reported rolling orders up 41%. Siemens has committed to a 50% capacity expansion by 2030 — which is the right direction, but four years away from a buyer trying to energize a facility in 2027.

Switchgear tells a similar story at a shorter time horizon: medium-voltage switchgear lead times run 38 to 63 weeks, pad-mounted transformers 68 to 113 weeks, and generators 60 to 100 weeks. None of that is labor scarcity or permitting delay — it’s a global electrical-equipment supply chain that was sized for a pre-AI-boom world and hasn’t caught up.

The contract risk nobody priced in

This is where it stops being an industry statistic and starts being a buyer’s problem. Escalation clauses in most construction frameworks and master service agreements were written for materials like lumber and rebar — commodities with price history going back decades. They typically don’t cover large power transformers, generators, or switchgear, which is precisely the equipment now driving the overrun. A general contractor who signed a GMP against 2025 unit rates and is still procuring equipment in 2026 isn’t managing a margin problem anymore; it’s a solvency problem, and solvency problems get passed to whoever is on the other side of the contract, one way or another — through change orders, schedule slippage, or a contractor walking away mid-build.

The fix isn’t exotic, but it requires buyers to ask for it explicitly before signing, not after a transformer quote comes back 40% over budget. Base pricing on current equipment quotes with explicit validity windows rather than last year’s unit rates. Split the contract structure so shell-and-core work stays fixed-price while electrical and mechanical scopes — the categories actually driving the inflation — move to cost-plus, target-cost, or provisional-sum terms. And settle, in writing, who owns long-lead procurement risk: if you as the buyer are free-issuing equipment to save time, that risk is now explicitly yours, not the contractor’s, whether or not the contract says so in plain language.

What this means if you’re negotiating capacity right now

Data centers accounted for 24% of all U.S. nonresidential building starts through August 2026, per ConstructConnect — which means data center developers aren’t just competing with each other for electricians, transformers, and switchgear; they’re competing with every other category of commercial construction for the same shrinking pool of both. That scarcity shows up differently depending on how you’re buying. If you’re leasing colocation capacity, rising build costs flow through as higher $/kW pricing on new contracts, typically with less room to negotiate than a year ago. If you’re doing build-to-suit or self-build, the fixed-price-versus-cost-plus split in your contract is the single biggest determinant of whether next year’s copper or transformer price move lands on your balance sheet or your contractor’s.

Either way, the diligence question worth asking a developer or GC before you sign is less about their headline price and more about their procurement position: do they have an existing framework agreement with Siemens, GE Vernova, Eaton, or another major OEM that puts them ahead of the queue, or are they sourcing equipment on the open market alongside everyone else who started this year? A developer with locked-in allocation from a prior framework deal can often hold a price that a developer buying spot cannot — and that difference is worth more in 2026 than almost any other line item in the proposal.

The nuance: this isn’t slowing the building boom

None of this is cooling the pipeline. Cushman & Wakefield puts the North American data center pipeline at roughly $2.3 trillion, including $492 billion under construction and $1.8 trillion planned or precommitted, and projects frontier-market spending to rise nearly sevenfold, from $91 billion under construction today to $631 billion planned. ConstructConnect is forecasting a further 30% rise in power-infrastructure construction starts in 2027. Hyperscalers with existing equipment framework agreements and deep balance sheets are largely absorbing the cost increase rather than slowing down. The nuance for smaller buyers — enterprises, neoclouds, and regional colocation tenants without that negotiating leverage — is that the inflation is real, but it isn’t evenly distributed, and the gap between who has supply-chain leverage and who doesn’t is widening faster than the headline numbers suggest.

The takeaway for anyone actually signing a data center contract this year is simple: the headline $/MW or $/kW number in a proposal is now less informative than the procurement and risk-allocation terms behind it. Ask for current equipment quotes, not 2025 benchmarks; push for cost-plus treatment on the electrical and mechanical scopes actually driving the inflation; and confirm your counterparty’s position in the transformer and switchgear queue before you confirm your own delivery date. For a structured way to work through those questions before you sign, see the TechInfraHub Data Center Buyer’s Toolkit.

Written by

Raajeev Ratra

Data Center Infrastructure Expert | 15+ Years in DC Design, Operations & Project Management

Raajeev is a seasoned data center professional with hands-on experience in hyperscale facilities, colocation design, power & cooling infrastructure, and global DC operations. He shares practical insights to help engineers and IT leaders build better infrastructure.

Connect on LinkedIn →