Site Selection for Data Center Developers: What to Answer, in Order

Site selection goes wrong most often when questions are answered out of order. A team that negotiates price before confirming power, or files for rezoning before checking for a local moratorium, spends time and credibility on a site that was never going to work. The sequence below is the one we use when we screen sites for developers.

The order is not rigid. A known moratorium or a mapped floodway can end a site at step one. But the order reflects which variables take longest and which are hardest to change.

Every site search should start with a written requirement set. Without one, teams compare sites against different assumptions and cannot rank them.

• Total megawatts at full build-out and the ramp schedule by phase.
• Buildable acreage needed for buildings, substation, generators, stormwater and buffers.
• Redundancy target: single or dual feeds, N+1 or 2N on-site systems.
• Cooling approach and what it implies for water and power density.
• Latency or fiber needs, which set how far from a metro the site can be.
• Schedule: the date the first megawatts must be energized.

Requirements differ sharply by project type. Compare them in hyperscale, edge and colocation site needs and, for high-density campuses, AI data center site requirements. A realistic load ramp schedule is worth writing early, because utilities respond to it directly.

Power is the gating variable. Start with nearby transmission voltage, substation locations and any public signals about capacity, then move to utility conversations. Our guides on the large-load interconnection process and power timelines and queues explain what to expect.

Ask the questions that determine whether a site is real: which substation or line would serve the load, what upgrades are likely, who pays for them, and what the utility needs from you to start a study. Upgrade costs and schedules can change a site’s ranking entirely. Read the commercial side too: minimum charges, collateral and contract length under a large-load service agreement affect project economics as much as the connection itself.

If grid timing does not fit, decide early whether on-site or bridge generation is part of the plan. Grid power vs. on-site generation lays out the trade-offs.

Once power looks credible, test the land. Subtract floodplain, wetlands, steep slopes, easements and setbacks, then lay the campus plan on what remains. Gross vs. buildable acreage shows how quickly a large tract can shrink. If the site needs several owners, read parcel assembly before the first approach, because sequencing and confidentiality matter.

Then test the approval path. Check current zoning, whether data centers are a listed use, and how the jurisdiction has handled similar requests. Screen for moratoriums and local restrictions early; they can appear quickly and often target exactly the kind of project you are planning. Noise, height near airports and generator air permits all affect the site plan, not just the schedule.

Supporting infrastructure rarely decides a site on its own, but it can add cost and months to a schedule. Check it before control, while it can still change the ranking.

• Fiber: how many independent routes reach the site, from which directions, and how far the nearest long-haul or metro route runs. Diverse paths matter more than any single provider.
• Water: whether the cooling design needs a large, steady supply, who would provide it, and how wastewater would be handled. Closed-loop and air-cooled designs reduce this need but raise power use.
• Access: road frontage, turning radii and load limits for transformers and generators, and whether the county or state will require road improvements.
• Gas: if on-site generation is in the plan, the distance to a transmission pipeline and the capacity of a lateral.

A site that is strong on power but weak on fiber or water is often still workable, at a cost. Price that cost early rather than discovering it in diligence, when the land price is already set and the utility clock is running.

Match the form of control to what remains unknown. When power and zoning are still open, an option with a long, paid study period limits exposure. When most risks are retired, a purchase contract with a shorter diligence period may cost less overall. Buy, option or lease compares the structures.

Diligence then confirms what screening suggested: title and an ALTA survey, a Phase I environmental assessment, wetland delineation, geotechnical borings, a utility study and the entitlement path. Use the due diligence checklist to scope it, and a feasibility study when a decision needs a single written basis for an investment committee.

Write the control agreement around the diligence plan, not the other way around. If the utility study is expected to take most of a year and the rezoning cannot be filed until it returns, the option term and extensions need to cover both, with room for delay. Tie extensions to milestones the landowner can see, such as a filed load request or zoning application. That keeps the relationship workable when timelines slip, which they usually do.

• Treating distance to a transmission line as a proxy for available capacity.
• Comparing sites on gross acreage instead of buildable acreage.
• Signing short, inexpensive options that expire before the utility study returns.
• Filing publicly before talking to local officials and nearby residents.
• Assuming a water-intensive cooling design without confirming supply and discharge.
• Leaving title, easements and mineral rights until after the land price is fixed.

Our guide to why data center sites fail covers these patterns in more depth, with the checks that catch them early.