That shift was the frame for a Power Lunch segment on CNBC Television dated September 4, 2026, in which former North Dakota Senator Heidi Heitkamp and Daniel Castro of the Information Technology and Innovation Foundation joined the program to debate the backlash to data centers, the support for them, the moratoriums being adopted in some jurisdictions, and the local tax benefits that cut the other way.
The conversation was a debate, not a finding, and it is worth holding it as that. But the terrain the two guests were arguing over is real and it is granular: towns and counties, not statehouses and certainly not Congress, are where most of these projects live or die.
Why the fight lands at town and county level
The reason is structural rather than ideological. Almost every lever that determines whether a data center gets built is held locally.
Zoning is local. Whether a parcel is designated for heavy industrial use, whether a special exception is required, whether a rezoning triggers a public hearing, all of that runs through a planning commission and a governing body that answer to residents within driving distance of the site. Special use permits are local. Conditional approvals with noise limits, setback requirements, and screening conditions are local.
Property tax assessment is largely local too, and this is the piece that makes the politics so unstable. A large computing facility can represent an unusually concentrated addition to a jurisdiction’s tax base. That is the promise offered to a county with a school district under pressure or a road maintenance backlog. It is also why abatement negotiations become the actual substance of the deal. The public argument may be about water or noise. The document that matters is often the payment schedule.
Then there is the grid and water politics, which people tend to describe as regional or national but which arrive as local grievance. Residents do not experience transmission planning. They experience a substation proposal on a road they use, a rate case that shows up on a monthly bill, and a groundwater question raised by a neighbor with a well. Whatever the state or regional authority technically controls, the anger is delivered to the county commission because that is the body within reach.
A moratorium is the tool that fits this situation. It does not require a jurisdiction to decide anything about artificial intelligence. It requires only a majority willing to pause while an ordinance is drafted. That is a low bar, procedurally, and it explains a good deal of the 2026 pattern of local and state pushback without needing any grand theory about public sentiment toward technology.
The tradeoff the panel was circling
The Power Lunch discussion set two positions against each other, and the tension between them is the entire local debate compressed.
On one side is the case that these facilities deliver measurable fiscal benefit to communities that have few other options for expanding revenue, and that a moratorium is a decision to forgo that. On the other is the case that the benefits are not evenly distributed, that construction employment is temporary while operating headcount is modest relative to footprint, and that the costs in land, water, power capacity, and neighborhood disruption are borne by people who may not see the revenue reflected in anything they recognize.
Neither position is obviously wrong, which is why the debate keeps recurring rather than resolving. A county with declining commercial assessment and a stressed school budget is genuinely looking at a different calculation than a growing suburb with existing tax capacity and constrained water. The same project, presented in two jurisdictions, produces two different answers, and both answers can be defensible on the numbers each place is actually looking at.
What complicates it further is that the fiscal upside is frequently negotiated away before the vote. Incentive packages, phased abatements, and infrastructure cost sharing all shift the arithmetic. A community that approves a project on the strength of projected revenue may find that the near term figure is much smaller than the headline, arriving years later than assumed. That gap between the pitch and the schedule is where a lot of local trust gets spent.
What local arithmetic can and cannot settle
County budget maths is very good at answering one question. Does this specific project, on this specific parcel, under these specific terms, improve or worsen this jurisdiction’s fiscal position over a defined horizon.
It is not built to answer the questions that national coverage keeps attaching to it. It cannot determine how much aggregate compute capacity the country needs. It cannot allocate transmission investment across regions. It cannot decide whether the current pace of buildout is proportionate to demand or whether some portion is speculative. Those are systemic questions, and no planning commission has the mandate or the information to resolve them.
But the systemic outcome is being produced anyway, as the sum of thousands of local answers. If enough counties in a corridor adopt moratoriums, capacity relocates to jurisdictions with weaker tax bases and more accommodating boards. That is a national siting pattern assembled from purely local reasoning, and nobody designed it.
This is the part worth sitting with. The debate that Heitkamp and Castro were having on Power Lunch is happening in hundreds of rooms simultaneously, in vocabulary borrowed from assessment schedules and comprehensive plans rather than from technology policy. The people in those rooms are not deciding the future of AI infrastructure. They are deciding whether a particular deal pencils out for a particular place.
Servers land where the local arithmetic allows them, not where national rhetoric points.