The Real Problem Isn’t Data Centers: It’s Bad Policy.

America’s latest adversary isn’t a person but a windowless, humming, usually beige structure that’s now being met with moratoria across the nation, including at the federal level, and is fueling many a 2026 campaign. Data centers are, much like Hansel, incredibly hot right now—and not in a favorable sense.

In one respect, the public pushback against data centers in the United States isn’t surprising: their construction and deployment have surged across the country, creating—without getting too lyrical—a highly visible focal point for American worry about AI, Big Tech, U.S. politics, a volatile economy, and other anxieties. A recent New York Times op-ed captures the mood on the left well: these colossal blocks gulp water, drain electricity, hum around the clock, and enrich Silicon Valley billionaires at the expense of ordinary people. Yet opposition isn’t limited to the left: you’ll also find similar sentiments among MAGA supporters on Facebook and NextDoor, and—my own recent drive near Raleigh confirmed—large “NO DATA CENTER” signs along many rural roads. 

Still, the common objections to data centers largely crumble under scrutiny. The facilities aren’t flawless and raise genuine concerns, but they also bring substantial benefits that the backlash tends to overlook. If data centers do create problems, the root cause is mainly government policy, especially subsidies and energy bottlenecks. That means improved policy can address the real issues of America’s data center buildout—without bans or grave economic fallout.

Although we throw around the phrase “data center” quite a bit, there are actually many varieties—not merely large or small, but differing by ownership and technology. For a concise primer, check out this recent Forbes feature covering the fundamentals and why the issue has suddenly gained traction. (NetChoice’s explainer series adds even more.) As highlighted there, data centers have existed for decades, and thousands are already operating in the United States. They’re essentially any “purpose-built building that runs vast numbers of computers reliably, securely and around the clock,” and there’s probably at least one buzzing nearby right now. (Relax—don’t panic!) 

What’s changed lately is the scale and tempo of data center deployment in the United States. The newest, largest facilities—often labeled “hyperscale”—spanning millions of square feet—are fueling AI and driving one of the most substantial waves of capital investment in American history:

As we’ve sadly learned from the U.S. housing cycle, such rapid, large-scale development was bound to provoke pushback. With data centers, however, the objections are intensified by the fact that many of these enormous sites are being situated in quiet rural areas that likely didn’t anticipate the noise and disruption that accompanies constructing and operating any colossal project—regardless of what happens inside them:

Tack on the usual incentives that scammers and politicians use to prey on Americans’ insecurities about these large facilities and AI at large, and you get a recipe for broad hysteria—and, if recent national polls and social feeds are any guide, that’s precisely what we’re witnessing.

(I kid you not, one guy recently posted on NextDoor a catalog of all the data centers already operating in North Carolina’s Triangle region—including tiny ones at nearby universities!—and urged neighbors and local officials to rally against them. Sigh.)

Most of the hysteria, however, is built on anecdotes and questionable science that drown out rational analysis and exploit people’s insecurities and gaps in knowledge.

Let’s start with electricity, arguably the trickiest data center issue going forward. Several regional and nationwide studies have found little basis for the claim that data centers have already driven up local electricity bills. In fact, a fresh peer-reviewed study indicates that data centers actually helped push average U.S. retail electricity prices lower, on net, from 2015 to 2024, because their steady, high demand allowed grid operators to spread fixed costs over more consumption. The result is counterintuitive but aligns with 2025 research showing that high-demand regions—like North Dakota and Virginia, which saw a surge in data centers—experienced lower prices overall, while places such as California, where demand waned, saw price increases:

“More power customers,” notes the Washington Post, “means more ways to share the fixed costs”—reducing rates for everyone.

As Roger Pielke Jr. discusses on Dispatch Energy, that doesn’t guarantee the same electricity trends will hold in the future. The hyperscale push is enormous, and demand for power is forecast to exceed supply. This mismatch is already bearing out in some communities’ energy bills, albeit modestly, and will likely appear elsewhere where policy blocks the supply side from responding. But, as Pielke emphasizes, this mismatch is addressable via policy—not an immutable “data center” problem. More on policy momentarily.

The case for water is weaker. As the Washington Post editors noted last year, one viral claim—that a single Google data center in Chile used more water than an entire town—was later admitted by its author to be “off by a factor of 1,000.” They added that all U.S. data centers combined use less than 0.5 percent of the nation’s freshwater, and that “Google’s thirstiest facility, in Iowa, uses about the same quantity as a typical golf course.” 

The golf-course analogy is telling, because it reveals a common tactic used by opponents to spark local unease: throw out alarming numbers with little real-world comparator. Jim Pethokoukis of the American Enterprise Institute recently did such a comparison, showing data center water usage to be comparatively minor:

AI researcher Andy Masley has done admirable work showing why a lot of “data center water usage” statistics in various studies include non-potable water, indirect consumption from nearby power plants, and water that is recirculated rather than consumed (thus not really affecting local supplies). Accounting for these can dramatically inflate water-use figures and public outrage, even when the data center itself isn’t drawing much water and nearby residents aren’t affected.

Looking ahead at actual consumption, Masley estimates that the water directly consumed onsite by all U.S. data centers by 2030 will amount to roughly 8 percent of U.S. golf industry use and a mere 1 percent of water used for American corn production. It’s also far surpassed by industrial uses like steel plants. Even with indirect use included, the 100 billion gallons recorded in 2023 barely registers against the annual totals for golf (about 550 billion gallons) and lawns (about 3 trillion gallons). There could be some localized strain in constrained areas, but the outrage would likely dissipate if a golf course or a cornfield were the source of the “problem.”

Masley has also debunked several viral tales about rural data centers polluting drinking water, often finding the reports grossly misrepresented or outright fraudulent. In general, data centers aren’t major sources of water pollution in host communities, largely because most cooling occurs in closed systems. (And, again, their real use is eclipsed by other perfectly acceptable activities like farming.)

Masley has since extended the same analysis to land use and waste heat, reaching similarly benign conclusions. Air quality concerns also fall short. The Taxpayers Protection Alliance has meanwhile debunked recent online assertions that data centers generate unacceptable ambient noise.

In each case, it’s not that data centers do nothing, but that their effects are modest or trivial next to many other economic and leisure activities we tolerate every day—and often benefit from more than data centers do (no offense to golf!). And like other concerns, these effects can be addressed with ordinary regulatory levers, not a blanket ban.

Here’s a helpful summary chart from Pielke that offers useful comparisons, helping to debunk much of the online data-center hysteria:

Another major flaw in most data center opposition is the tendency to overlook the benefits they bring to local communities and to the nation as a whole. The easiest starting point is tax revenue, with Northern Virginia illustrating the point most clearly. A recent Chamber of Progress study found that in 2024 the region gained more than $2 billion in extra tax receipts thanks to its booming data center market, largely from property taxes:

Localities in Virginia have used these funds for schools, affordable housing, and other government services. The report also found that without this data center tax income, Loudoun County would have needed to nearly double its residential property tax rate, adding around $5,800 to a typical homeowner’s yearly bill. Instead, homeowners enjoyed a property tax cut. 

The report then surveys the private-sector activity spurred by both the construction and ongoing operation of data centers in Northern Virginia:

That economic lift isn’t unique to Northern Virginia. A Georgia Tech study of data-center openings nationwide found that surrounding counties typically saw meaningful rises in employment, wages, and business formation after a center opened. Urban areas—already endowed with deep labor markets and business networks—tended to see larger gains, but rural regions still benefited through lower unemployment and higher tax receipts. 

Both the Wall Street Journal and Bloomberg have chronicled a genuine surge in data-center construction jobs across many regions, with welders, plumbers, and electricians now earning six-figure salaries, along with signing bonuses and other perks (free steaks, too!). This construction boom has driven demand for local workforce housing and related services, as well as nationwide demand for electrical, cooling, and power equipment manufactured in the U.S. and abroad. All told, Pielke notes, “the AI-infrastructure buildout contributed roughly 0.8 percent of U.S. GDP by early 2026.” That’s a substantial share.

Those willing to sell land for data centers are also finding themselves financially rewarded.

Zooming out further, the economic case becomes even stronger. As Gene Marks notes in USA Today, “Today’s data centers are the modern version of railroads. They form the backbone of everything that powers our economy. Railroads moved goods; data centers move information and enable contemporary commerce.” AI is a major contributor to this shift, yet—as the Cato Institute’s Jennifer Huddleston and Christopher Gardner explain—it isn’t the sole driver of demand:

Blocking data centers would stifle this important commerce on a national scale. It would also push the activity overseas, taking jobs, tax revenue, and influence with it.

None of this implies data centers are flawless, and it would be simplistic to claim the pushback is purely manufactured (though some of it is). When real problems exist, the culprit is almost always policy rather than an inherent flaw of data centers or AI. That means the issues can be resolved with policy, too.

The most obvious remedy is to eliminate the subsidies various states and jurisdictions shower on data-center projects and their billionaire owners. Virginia’s data center sales and use tax exemption alone cost more than $1 billion in 2024 and is projected to nearly double in 2025, benefiting Google, Amazon, and other mega-companies. Indiana’s incentives are nearly as large. Given the unrelenting demand for AI and other internet services, these centers will be built without subsidies. Like stadium handouts and corporate relocation incentives, these subsidies amount to states playing one another for the best deal—at taxpayers’ expense. End them all. The industry can thrive without public subsidies, and every subsidy dollar undercuts the tax-revenue benefits and fuels a legitimate grievance from those advocating moratoria—perhaps their only real complaint.

The next obvious step is to loosen the supply side of the electric market, where the principal economic risk likely resides. As Pielke notes, multiple studies show local power prices wouldn’t rise if supply could meet new demand, yet—

Other factors, such as tariffs on transformers, electrical steel, solar panels, and other essential equipment, add to these burdens. Fix those, and some of the demand-side pressure from data centers will ease.

For the rest, Cato’s Travis Fisher and Huddleston offer an even better option: allow data-center developers to contract directly with private generators or to build their own power through “consumer-regulated electricity” (CRE) reforms:

Allowing these wealthier firms to finance their own power makes a lot of sense, yet regulatory barriers and related constraints make that option impractical in most places. That is the core problem, not data center demand. Fortunately, as Fisher notes, a few states have begun experimenting with CRE, and the White House’s voluntary Ratepayer Protection Pledge, endorsed by most hyperscalers, signals a similar direction. But more progress is needed—and quickly.

Other data-center concerns, such as noise, pollution, and land use, have straightforward solutions: let localities enforce the laws already on the books and make site-specific decisions about whether to permit data centers in their jurisdictions. NIMBYism fuels much of the resistance, and I believe most of it is misplaced. The bright side is that this is a matter of local control, not a nationwide or federal moratorium that would hobble an entire industry. In the former case, a data center might relocate to friendlier regions and negotiate site-specific terms with developers, as Loudoun County did. America’s great Laboratories of Democracy can exercise their authority.

As Huddleston notes, turning a local land-use dispute into a federal ban would discard the baby with the bathwater, imposing economic harms that go well beyond a few NIMBY-affected communities.

Most of what people detest about data centers is either factually wrong or aimed at the wrong target. Beyond the points we’ve covered, there are real—indeed sometimes alarming—concerns about AI safety, security, and disruption, and I’m increasingly worried about incumbent AI giants leveraging regulatory power to erect new barriers to competition. The rational response to those genuine AI issues is to address them directly, not to throttle the essential infrastructure that makes the technology—and much more—possible. The latter path would be costly at local, national, and even global levels.

Meanwhile, the real, fixable problems with data centers trace back to policy failures: subsidies, sluggish permitting, outdated power rules, and more. Correct those, and the concerns largely vanish. Nationalizing the issue would only worsen matters.

In many respects—both the good and the bad—the current data-center debate echoes the fracking discussion from about twenty years ago. That era also featured a disruptive industrial expansion with valid local costs that were often exaggerated or fabricated, prompting calls for a nationwide ban. The U.S. avoided such a blanket ban, leaving decisions to states and communities. Some bans did occur, but the decentralization ultimately preserved a transformative technology with broad economic and geopolitical upside.

Perhaps data centers can follow a similar, second-best fate, because a nationwide prohibition could yield consequences far greater than the fracking controversy.

Maybe, just maybe, we don’t remember the good old days as clearly as we think:

Efficiency:

Studies show foreign tourism plummeted after “Liberation Day.”

Beyond AI, U.S. manufacturing is… unimpressive: