How data center site selection works: power, land, and speed to energization

Data center site selection has been turned upside down. For most of commercial real estate, land is the starting point: what is available, what does it cost, what will the zoning allow. For data centers in the AI era, land is often the last thing that matters. The first question is power, how much a site can get, and how quickly.

This guide walks through how developers and hyperscalers actually evaluate a data center site, in the order the decisions get made. For the running news on this beat, see our data center topic hub.

Power comes first

A single large data center campus can draw hundreds of megawatts, as much electricity as a small city. Utilities cannot always deliver that capacity on a developer’s timeline. So the opening question on any site is not “can I buy this land,” it is “can this site get power, and when.”

That has two parts. The first is generation and transmission: is there enough electricity in the region, and enough wire to move it to the site. The second is the interconnection queue, the backlog of projects waiting for the utility or grid operator to study and approve their connection. In constrained markets that queue can add years. A site with cheap, abundant land and a five-year power wait is often worth less than a smaller, pricier site that can energize in eighteen months.

Speed to energization is the real currency

Because AI demand is moving fast, the operators racing to deploy capacity value one thing above almost all others: speed to energization, the time from breaking ground to the day the building can actually run at load. Everything that shortens that timeline, existing substation capacity, a utility willing to move, pre-approved interconnection, adds value. Everything that lengthens it subtracts value.

This is why you see developers pay premiums for sites near existing power infrastructure, and why brownfield sites with inherited electrical capacity, former industrial plants, retired generation, can command attention that their location alone would never justify.

Then land, fiber, water, and zoning

Once power pencils, the traditional factors re-enter:

• Land and layout. Data centers want large, flat, contiguous parcels with room to expand. Assembly and expansion optionality matter.
• Fiber connectivity. Proximity to major fiber routes and network points still shapes where certain workloads can sit.
• Water and cooling. Cooling method drives water demand, which is increasingly a local political issue as well as an engineering one.
• Permitting and community. Local land-use approvals can accelerate or kill a project. Some jurisdictions court data centers for the tax base; others have moved toward moratoriums over power rates, noise, and water. Understanding the local posture early is essential, our guide on how local land-use hearings work covers how to read that process.

What to check first, in order

If you are screening a site for a data center use, work the list in the order the market actually weights it:

1. Available power capacity and the interconnection timeline.
2. Speed to energization, including existing substation and transmission proximity.
3. Utility posture and power cost trajectory.
4. Land size, expansion room, and buildability.
5. Fiber and network access.
6. Water and cooling feasibility.
7. Local permitting environment and community stance.

Get the first three wrong and the rest never matters. That inversion, power and speed ahead of land, is the single most important thing to understand about this beat, and it is why data center development is reshaping where capital goes across the US. Track the moves on our data center hub and in The Wire.