Data Center Site Selection: Power, Cost & Compliance Checklist (2026)

In 2020, the site selection question was “where is land cheap and fiber close?” In 2026, it’s “where can I actually get 300 MW before 2030?” That single shift has rewritten the entire criteria list operators use to evaluate a location — and a surprising number of teams are still running their evaluation process as though it’s five years ago.

This checklist walks through what actually decides a site in 2026, in the order that reflects real deal-breaking power, not marketing brochure order.

Power Comes First — Not Land, Not Tax Breaks

As AI-driven demand accelerates and campuses scale toward gigawatt levels, the grid is struggling to keep pace. Capacity is tightening, interconnection timelines are stretching, and developers are increasingly finding that a site they want can’t get power in a reasonable timeframe at any price. Many developers now select the power source first and build the rest of their site decision around it — a full reversal of the old land-first approach.

Here’s the number that should reframe how you think about incentives specifically: in a widely-cited Mortenson Construction survey of data center owners, only about 3% ranked tax credits and local incentives as the single most important site-selection factor — power, land, and network latency all ranked far higher. If you’re spending more evaluation time on incentive packages than on power delivery timelines, you’re optimizing for the wrong variable.

What to actually verify on power, before anything else:

• Confirmed interconnection queue position and realistic timeline (not the utility’s optimistic estimate)
• Whether the local grid has genuine headroom, or whether your load would require new generation capacity to be built
• On-site or near-site generation alternatives if grid timelines don’t work (Small Modular Reactors and on-site gas generation are increasingly part of this conversation for gigawatt-scale projects)

Land — Matters Less Than You’d Think, But Land Quality Still Matters

Land cost is typically a low-single-digit share of total project cost — it’s genuinely one of the smaller line items once you account for power, construction, and equipment. But land quality deserves real diligence:

• Contiguity for phased expansion — can you actually grow on this site, or are you locked to your first-phase footprint?
• Developable acreage after setbacks and easements — the parcel size on paper isn’t the buildable size
• Clean title with control instruments that survive a multi-year development timeline — a cheap parcel with a pipeline easement running through your buildable envelope isn’t actually cheap
• Geotechnical conditions and floodplain exposure — underwrite for a 20-30 year operating life, not current conditions. Insurance markets are already repricing seismic, hurricane, wildfire, and flood risk; your site evaluation should price it in too, not treat it as someone else’s future problem

Water — An Increasingly Contested Resource

Water access for cooling has moved from a minor checklist item to a genuine site-selection constraint in water-stressed regions. This matters more for evaporative and traditional cooling approaches than for facilities committed to liquid-to-chip or immersion cooling from day one — worth factoring your cooling architecture decision into your site evaluation, not treating them as sequential, unrelated decisions.

Fiber and Network Latency

Confirm genuine carrier diversity, not just proximity to a fiber route. Single-carrier markets create real operational risk — if your only option goes down, you have no failover path. Map the actual carrier landscape for each candidate site and flag anywhere fiber diversity falls short of what your operation actually requires, rather than assuming “there’s fiber nearby” is sufficient diligence.

Tax Incentives — Real, But Not the Deciding Factor

Incentives remain genuinely significant money — Indiana’s package tied to a major cloud provider’s data center investment is estimated at $8.2 billion, and Virginia’s ongoing sales tax exemption is projected at roughly $1.6 billion annually. But given the 3% figure above, treat incentives as the last criterion you optimize, not the first.

Two things worth modeling explicitly before you rely on any incentive package:

1. Durability — some states are actively reconsidering these programs. At least nine states have considered repealing data center incentives in the past year alone, while a comparable number have introduced legislation to expand them. The regulatory environment is genuinely in flux, not settled.
2. Clawback provisions and conditions — job creation commitments, capital investment minimums, and reporting requirements typically come attached to any incentive package. Model your tax position at year one, year five, and year fifteen — the incentive period is only part of the real long-term cost picture, and abatements expire on a schedule that doesn’t always align with your operating timeline.

The Public Opposition Factor — A Genuinely New Variable

Worth naming directly since it’s reshaping the incentive conversation: a March 2026 Gallup poll found 71% of American adults oppose construction of an AI data center in their local area — a higher opposition rate than Gallup has ever recorded for a nearby nuclear power plant. This is pushing back against incentive packages from both fiscal conservatives (forgone tax revenue) and environmental advocates (grid strain, utility bill increases) simultaneously. Community opposition risk is now a real site-selection variable, not just a PR afterthought — a technically ideal site can still become a multi-year entitlement fight if local sentiment isn’t accounted for early.

Quick Reference: Evaluation Priority Order

1. Power — confirmed capacity, realistic interconnection timeline
2. Land quality — not just cost; contiguity, title, geotechnical/flood risk
3. Water access — especially if not committed to liquid/immersion cooling
4. Fiber diversity — genuine carrier redundancy, not just proximity
5. Community/regulatory risk — entitlement path, local sentiment
6. Tax incentives — real value, but model durability and clawback terms; don’t let this drive the decision

Regional Snapshot: How the Landscape Has Shifted

The scale of what’s being planned has changed the entire calculus. JLL’s 2026 Global Data Center Outlook projects roughly 97 GW of new data center capacity globally between 2025 and 2030, and Goldman Sachs Research estimates data center power demand will rise 160% by 2030. Against that backdrop, individual state incentive fights have become proxies for a much larger structural question: which regions can actually deliver power at the speed and scale this buildout requires.

Some states are already visibly repricing their position. Florida eliminated its sales tax exemption for facilities under 100 MW in 2026, with no grandfathering for projects already in progress — a real signal that incentive terms you evaluate today may not hold for a project that takes years to reach operation. Georgia Power has projected that data center demand could drive up to 80% of the state’s future added generation capacity, a figure large enough to turn what used to be a quiet local tax exemption into a genuine statewide policy fight. If you’re evaluating a site in a state where data center demand represents a similarly outsized share of grid growth, expect the political and regulatory environment around your project to be more volatile than a purely technical evaluation would suggest.

The practical implication: states with genuinely cheap, reliable power are increasingly able to compete without generous tax packages, while states that were only competitive because of incentives now face pressure to either keep paying or lose the project pipeline entirely. Understanding which category your target state falls into tells you a lot about how durable any incentive offer actually is.

Building a Real Site Comparison Matrix

Rather than evaluating candidate sites one at a time in isolation, build a single comparison matrix across your realistic shortlist, scoring each site against the priority order above. A few practical notes on doing this well:

• Score power on a realistic timeline, not a best-case one. Ask utilities directly for their actual current interconnection queue position for a project of your size, not a generic estimate. Queue positions and timelines are increasingly the actual gating factor, not a formality.
• Get land quality assessed by a geotechnical engineer before you’re deep in negotiations, not after. A site that looks clean on paper can reveal expensive complications (poor soil bearing capacity, unmapped floodplain risk) that materially change your build cost once you’re already committed.
• Have a specific person own the “community sentiment” evaluation, not just the technical team. Local government relations, prior data center reception in the region, and active advocacy group presence are qualitative but genuinely predictive of entitlement timeline risk.
• Model at least two incentive scenarios — the package as currently offered, and a reduced/expired version — so your project’s viability doesn’t depend entirely on an incentive holding for its full term in a policy environment this active.

Common Site Selection Mistakes

Treating the first attractive incentive offer as the deciding factor. Given the 3% figure above, an aggressive incentive package on a site with real power constraints is a worse choice than a smaller incentive on a site with confirmed power availability — yet incentive size is often what gets the most internal attention during evaluation, precisely because it’s the easiest number to compare across sites.

Underestimating entitlement timeline risk in politically sensitive markets. A technically ideal site in a region with active grid-strain concerns or vocal community opposition can take significantly longer to entitle than a technically adequate site in a more receptive market — and that timeline difference often matters more to project economics than a marginal difference in land or power cost.

Assuming today’s grid capacity assessment holds for your actual construction timeline. Given how quickly interconnection queues are moving in 2026’s high-demand environment, a power availability assessment that’s a year old may no longer reflect current queue positions — reconfirm before finalizing, not just at initial evaluation.

Frequently Asked Questions

Is it still worth pursuing tax incentives given only 3% of operators rank them as the top factor? Yes — incentives remain real money worth capturing. The point isn’t to ignore them, it’s to sequence your evaluation so power, land quality, and risk factors are confirmed first, with incentives optimized afterward rather than driving the initial site shortlist.

How much has power become the actual bottleneck, compared to five years ago? Substantially. Where site selection in 2020 assumed power would generally be available if you paid for it, 2026’s constrained grid capacity means power availability itself can eliminate an otherwise-ideal site regardless of budget — a fundamentally different planning problem than five years ago.

Should community opposition realistically change where we build? It should at minimum change your entitlement timeline assumptions and community engagement strategy. Given the scale of recorded opposition, sites in areas with active grid strain concerns or vocal local opposition carry real entitlement risk that a purely technical/financial site evaluation can miss.

How far out should we model tax incentive durability? At minimum year one, five, and fifteen, given how actively state incentive policy is being reconsidered in 2026 — treating an incentive as a fixed, permanent input to your financial model is no longer a safe assumption in this policy environment.

Evaluating multiple site candidates and need help modeling the real tradeoffs? Book a consulting session or explore our Data Center Efficiency Suite for power and cost planning tools.

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 →