We find the ground where power can be delivered this decade.

The scarce input in data-center development is no longer capital, land, or even megawatts in the abstract. It is a date — the day a site can actually take load. Almost every other variable is negotiable. That one is not, and it is set years before anyone breaks ground.

Large loads seeking a conventional grid connection in most of the United States join a queue measured in years. The consequence is a widening gap between two things that look identical on a parcel map: ground that is near power, and ground that can be energized. The first is abundant and cheap. The second is scarce and almost never priced correctly, because the people selling land price it as land.

What on-site generation actually changes

Behind-the-meter generation — gas, geothermal, solar, or a combination — does not make power cheaper. Frequently it makes it more expensive per megawatt-hour. What it does is take the schedule out of someone else’s hands. For a developer whose model depends on a commercial operation date, paying a premium per megawatt-hour to remove multi-year schedule risk is not a concession. It is the entire transaction.

This is not theoretical. Counties in the interior West have already approved integrated developments pairing a data-center campus with dedicated on-site generation and non-potable water supply, and have sustained those approvals through resident appeal. The template exists. What is scarce is ground where it can be executed.

Why the constraint has not reached land prices

Three frictions. First, the information is hard to assemble: pipeline capacity, interconnection posture and county political appetite live in different places, in different formats, and much of it is not published at all. Second, the parties holding the best ground — multi-generational ranch holdings, absentee owners, family trusts — are not listing it and are not reading data-center trade press. Third, and most underrated, the binding constraints are frequently political rather than physical. A county with excellent physical characteristics and a hostile electorate is not a site. That fact is invisible on every map product on the market.

That last item is the honest state of the field. When an interstate pipeline operator’s own presentation to a state legislature leaves a question mark where a capacity number belongs, no dataset on the market contains that number either. It is obtained by telephone or it is not obtained.

What follows

If time-to-power is the binding constraint, the correct unit of analysis is neither the parcel nor the county. It is the convergence — ground where a fuel source, a transmission path, adequate non-potable water, tolerable terrain, and a county that has demonstrably said yes all occur in the same place. Those convergences are rare, identifiable in advance, and currently owned by people who do not know what they have.

What this analysis cannot tell you

Much of the data required to underwrite these sites is modeled, not measured. Pipeline utilization is inferred from public filings and operator testimony rather than metered. County posture is a judgment formed from ordinances, ballot outcomes and permit decisions — evidence, not a number. Any framework reporting these with false precision, or quietly omitting a factor it could not resolve, is producing confidence rather than information.

Our practice is to label modeled inputs as modeled, to state the source and date of every material figure, and to refuse to produce a verdict where too much of the evidence is missing. A screen that declines to answer is more useful than one that guesses well.