Cheap acreage stopped being the constraint: data center siting now runs through the county board

The inland land-and-power grab has hit a governance ceiling: Denver, Oklahoma City and Aurora, Illinois have all approved moratoriums on data center construction, and South Dakota has enacted a law affirming local governments' authority to restrict or ban the facilities outright ^[7][9]. For anyone doing capacity planning, that shifts the binding constraint from the price of land to the willingness of a county board to grant permission ^[3].

The migration itself is not in dispute. Northern Virginia remains the country's data center hub, but hyperscalers have moved rapidly into Texas and Midwest states including Wisconsin, Michigan and Missouri, chasing plentiful and far cheaper land as AI demand outran existing infrastructure capacity ^[1][2]. What the developers appear to have underestimated is the political response. Gallup polling cited by DataCenterDynamics finds seven in ten Americans oppose such construction, with resource use topping the list of stated reasons and water and energy each named by nearly one in five ^[4][5]. Taken together, those two resource complaints account for roughly two in five of the reasons given ^[16].

Some of that is grounded. Grid strain, water consumption and noise pollution around data center construction are well documented, and household electricity rates have climbed in several markets alongside data center loads ^[6]. The politics do not require the evidence to be settled, only the perception to be shared.

Look at the instruments being used, because they are more durable than a single rejected rezoning. Oklahoma is weighing a statewide three-year pause on large facilities ^[8]. In Wisconsin, three Manitowoc County towns passed resolutions opposing a prospective project, after which the county board adopted its own 18-month moratorium ^[10]. That is four enacted moratoriums named across municipal and county government, plus one state law confirming local veto authority and one statewide pause under consideration ^[17]. A three-year pause is longer than most AI capacity plans currently extend.

The DCD opinion piece frames this as the start of a third phase of the build-out: the first concentrated capacity in metro network hubs where users live, the second was the heartland land-and-power grab, and the third will foreground local grid stability, water stewardship and active community engagement ^[11]. Its author argues the fix is not only better messaging but a different supply of capacity: the existing fleet of mid-market and colocation facilities that are already built, already permitted and already connected to the grid ^[12]. With new interconnection requests facing multi-year queues, that piece argues the standing fleet is the fastest capacity available anywhere, and that the "legacy" label is a misnomer given latent power and compute that intelligent orchestration could route toward AI workloads ^[13][14].

Treat that last part as a thesis rather than a demonstration. Already-interconnected brownfield capacity is real, but the sentence that follows in most versions of this argument is a product pitch, and the piece does not quantify how much latent headroom exists or at what cost.

What to watch: whether Oklahoma's three-year pause is adopted and whether other legislatures copy South Dakota's approach of pushing authority down to the level where opposition is loudest ^[8][9]. Watch household rate trends in markets absorbing new load, since bills are the mechanism that turns diffuse concern into voting behaviour ^[6]. And watch whether developers start disclosing water and noise figures before hearings rather than after, since the piece's own view is that the conflict resolves through genuine engagement rather than generic economic-benefit messaging, and that most of these projects do eventually come online ^[15].