Every week, another parcel gets pitched as a data center site because it's big, flat, and cheap. Land is the easiest box to check—and the least important one. What makes a site viable is whether it can deliver power at scale, on a timeline a tenant will believe.
Power comes first, and it's more than a number on a one-line. Real capacity means confirmed interconnection—an executed agreement or a defensible queue position, not a conversation with a utility account manager. It means understanding whether the local grid can actually deliver the contracted megawatts at peak, what upgrades the utility will require, and who pays for them. Two sites can both claim "100 MW available" and be years apart in reality.
Then the site itself has to hold what you'll build on it. That means contiguous, buildable acreage with workable geotechnical conditions, drainage, and access for heavy haul—transformers arrive on trucks that don't fit down every county road. Zoning and permitting posture matter just as much: a jurisdiction that wants the project can save a year; one that doesn't can kill it quietly.
Fiber decides what the site can become. A mining tenant needs modest connectivity; an AI tenant needs diverse routes from multiple carriers. Water availability shapes cooling options the same way. Neither has to be perfect on day one—but the path to both should be real, priced, and on the map before the site is marketed.
Finally, timeline is a screening criterion in its own right. A perfect site that energizes in four years loses to a good site that energizes in eighteen months. Every factor above ultimately converts into time—and time is what tenants are buying.
The discipline is in killing sites fast. Most opportunities fail one of these tests, and the cheapest moment to find out is before capital is committed. TON screens land, power, interconnection, fiber, and permitting as one integrated evaluation—so the sites that move forward are the ones that can actually be built.