Permission Is the New Power

For most of the AI boom, the questions about data centers were technical. How much power will one need. How much water. How quickly can the local utility connect it. Those questions still matter, and they are hard. But over the past year a different constraint has moved to the front of the line, and it is one almost no financial model accounted for. The neighbors.

By the middle of this year, organized local opposition had blocked data center projects worth more than 130 billion dollars. That figure deserves a second read. It is not a measure of projects that failed on economics or engineering. These were plans with capital behind them and power at least tentatively identified. They died on consent. A town, a county board, a zoning commission looked at what a large facility actually brings and decided the trade was not worth it.

A town, a county board, a zoning commission looked at what a large facility actually brings and decided the trade was not worth it. The promise is abstract, a bigger tax base and a few dozen jobs. The burden is concrete, and it lands on the nearest residents: years of construction, a permanent load on the grid, millions of gallons of water a day, backup generators, and constant noise, all for a building most of them will never enter. More communities are doing that math and deciding the town gives up more than it gets.

It helps to understand what those communities are actually looking at, because from a distance it is easy to dismiss as reflexive opposition to anything new. It is not. A hyperscale data center is one of the most demanding neighbors an industrial site can have, and it gives back less locally than almost any other large investment. Start with the jobs, because that is the promise that used to win these fights. A facility that draws a gigawatt and costs billions to build might employ a few dozen people once it is running. The construction phase brings work, but the permanent footprint on local employment is thin. Against that, a community weighs a continuous industrial load on its infrastructure, and increasingly it does the math and finds the exchange lopsided.

Then there is water, which has become the sharpest edge of the whole fight. In an open, evaporative system, roughly 70 to 80 percent of the water, a site draws, is lost to the air and does not come back. The volumes are not trivial. A single hyperscale facility can consume 1 to 5 million gallons a day, as much as a town of tens of thousands. And here is the number that should reframe the whole debate: nationally, all US data centers together use less than 1 percent of the country's water. This was never a national shortage. It is a local one. About 40 percent of US data centers sit in areas of high or extreme water stress, so a footprint that looks tiny on a national chart, lands hard on the specific aquifer next to the building. That is the reality communities are reacting to, and they are reacting to it now.

It is fair to ask whether technology solves this, and increasingly it can. Closed loop cooling recirculates sealed water and consumes only about 5 to 10 percent of what it draws, against the 70 to 80 percent an evaporative tower loses. Some operators are moving to recycled wastewater instead of drinking water, and a few have unveiled designs that use almost no water at all by pairing liquid cooling with air-based heat rejection. The direction of the industry is clearly toward using less fresh water. But this is still the exception. Most large data centers today, and most of the ones being permitted right now, still evaporate fresh water, often from potable supply, because the low water option usually costs more electricity. There is a genuine tradeoff underneath the fix: saving water tends to cost power, and saving power tends to cost water. A community voting on a project this year is voting on the cooling design being proposed this year, not on where the technology may land in five.

There is even an engineering trap worth naming, because it connects water back to everything else. The obvious answer to a water fight is to go closed loop or air cooled. But moving heat without evaporation is less efficient, so it generally costs more capital and more electricity. The move that calms a community over water can worsen the very power draw that strains the grid, which is the constraint the last few years have been about. The two problems are coupled. Solving one by brute force can deepen the other, and a serious plan has to hold both at once, rather than trading one headline for another.

Add the rest of the package. The heat these facilities reject. The constant low noise of cooling systems that neighbors describe as a hum that never stops. The truck traffic during construction and the diesel backup generators that get tested on a schedule. The tax abatements that communities are often asked to grant, which further shrink whatever local benefit was on offer. None of these items alone kills a project. Together they change the tenor of a public meeting, and public meetings are where these decisions increasingly get made.

Here is the part that matters for anyone planning at scale, and it is a discipline point more than a political one. A project can pencil out flawlessly on power, land, and fiber and still die in a room full of residents. The cost of underweighting the community does not appear anywhere on a pro forma. It shows up later, and in expensive forms. A two year delay while a permit is litigated. A referendum that reverses an approval. A site abandoned after real money has already gone into it. Social license, the plain willingness of a place to have you there, has quietly become as hard a constraint as available megawatts. And it behaves like every soft constraint I have watched over the years. It is treated as free, because it does not carry a number, right up until it becomes the single thing standing between a plan and a working facility.

I have spent a long time around infrastructure projects, and the pattern is consistent. Projects rarely fail on the part that everyone was measuring. They fail on the factor that was nobody's explicit job, because it never entered the model as a line. Community consent is the textbook version of that factor. It cannot be bought late, and it cannot be rushed. A company that arrives in a town with the permits half filed and a public relations plan is already behind one that showed up early, listened, designed the cooling to recycle or avoid fresh water, offered something the community actually wanted, and earned the room before it needed it.

That difference is becoming a real sorting mechanism. Projects are already migrating toward the jurisdictions that want them and away from the ones that fight. States and counties are actively positioning themselves as data center friendly, with streamlined permitting and prepared sites, while others quietly make themselves difficult. Over time this redraws the map of where compute can physically land, and it introduces a problem downstream. If the welcoming jurisdictions are a small set, the industry concentrates there, which strains those same local grids and water systems faster, which eventually turns even friendly communities skeptical. The permission problem does not stay solved. It moves.

The politics are escalating too. As local governments say no more often, states have started to intervene, some by preempting local control to keep projects moving and some by tightening the rules. Preemption can clear a path in the short run, but it tends to harden opposition, because nothing radicalizes a community faster than being told it has no say over what gets built next to it. The places that end up with durable data center growth will likely be the ones that made the local bargain work, not the ones that overrode it.

The takeaway is not that data centers are unwelcome, and it is not that water is an unsolvable crisis. It is that the era of assuming a community would be grateful is over, the tools to be a good neighbor exist but are not yet standard, and the companies that adopt them early will build while their competitors are stuck in hearings. The megawatts were never the only thing standing between a plan and a running facility. The permission was always part of the equation. We are simply being forced, at last, to price it.