A Rothbardian Case Against Bad Data Center Policy

Data centers are quickly emerging as one of the major political issues of the day.

These large warehouses full of computer servers processing, storing, and distributing digital information are not new. They’ve been a prevalent part of the world’s digital infrastructure since the internet—and cloud storage, and remote processing, especially—became ubiquitous in modern life.

However, the sudden and extremely productive advancements in consumer-facing and widely accessible professional AI tools in recent years have driven up demand for computing power—demand that, by all indications, will only increase further in the near future.

That has led AI and intermediary companies to scramble to build new data centers. The problem, of course, is figuring out where to put them. Generative AI requires a lot of computing power. The new AI-directed data centers tend to be very large—housing thousands of GPU-based servers—and require a lot of electricity for power and water to keep computers from overheating.

The technical and spatial requirements of a building housing thousands of computer servers also virtually guarantee it will be an eyesore. And many living near them have reported that the constant noise and light emitted by these data centers can be a nuisance.

Across the country, new data center projects have met fierce opposition from local residents who hate how the large buildings will change their neighborhoods and fear that a sharp increase in electricity demand will spike their power bills.

In fact, according to a new Gallup poll released last week, 71 percent of Americans oppose the construction of data centers in their local communities.

Presumably, then, the answer is to build data centers in remote areas, far away from any residential areas that might be bothered by them. However, most of the land in the US that is remote enough to qualify is in the west, where temperatures are frequently high, and the available water supply is already a pressing issue.

For example, a plan to build one of the world’s largest data centers in a remote valley in western Utah is facing significant backlash from Utahns who are citing concerns about how the facility would affect the state’s already-stressed water supply.

Environmentalists have joined in to oppose the development of what they view as large, energy-intensive infrastructure covering up the country’s beautiful natural landscape while draining nearby springs, streams, and aquifers.

And both groups have been joined by the burgeoning movement of neo-Luddites who want the government to slow, stop, or reverse the adoption of AI due to various concerns about how the technology will affect labor markets, democracy, and society overall. These types may genuinely agree with the so-called NIMBY-types (for Not In My Back Yard) or the environmentalists. But the core motivation for them stems from the fact that blocking the construction of data centers is a very effective way to put a heavy constraint on AI development.

A lot has already been written about the flaws and fallacies leading many to believe AI will trigger an employment apocalypse that will make everyone but a small sliver of the country much poorer. There are also reasons to actually expect positive political developments as AI begins to automate the exact kind of administrative, clerical, and bureaucratic work that has defined the so-called managerial class and made their positions necessary. And the environmental threat posed by data centers is often overblown by exaggerated projections based on earlier, less efficient forms of the technology that are now obsolete.

Where the opposition to data centers does have some merit, however, is on the NIMBY front.

There’s a perception that the libertarian or free-market position on this issue is that communities must accept giant data centers being built right next door because developers have a right to their private property, and the infrastructure will be excellent for economic efficiency. But at least when it comes to the Rothbardian position, it’s a bit more nuanced.

Rothbard published a long essay in 1982 that sought to lay out the proper libertarian, property-rights-based position on air pollution. The fact that legitimate property owners can emit harmful airborne particles or even excessive light or noise that negatively affect nearby property owners has long been pointed to as a hole in any theory of absolute property rights that justifies, or even requires, government intervention.

In the essay, Rothbard completely dismantles this talking point and demonstrates that a system based on absolute property rights and libertarian ethics would actually handle air pollution far better than our current politicized setup.

But while much of the essay focuses on the emission of noxious chemicals that cause bodily harm and property destruction—which Rothbard does consider a crime that justifies forcibly collecting restitution, provided it’s proven beyond a reasonable doubt—he also devotes some time to noise and light pollution.

Obviously, it is unreasonable and entirely unrealistic to argue that everyone is entitled to complete and total silence and darkness from all their neighbors at all times. However, there is also such a thing as excessive noise and light pollution that does interfere with an owner’s use or enjoyment of their property, which, after a certain point, is not categorically different from the other unjustified forms of pollution Rothbard was writing about.

To get around this apparent conundrum, he falls back on an application of the homesteading principle that has precedent in common law.

Imagine if an airport were built near a large expanse of empty land. The jets taking off and landing would be making a lot of noise. But, those sound waves would be traveling over empty land. What, then, if someone bought or homesteaded a parcel of land next to the airport and built a house on it? Could that homeowner then turn around and sue the owners of the airport for the excessive noise?

No, they could not. Because of its prior claim, the airport had homesteaded the right to emit a certain level of noise. And the homeowner, by choosing to enter into a situation where that level of noise was already present, had effectively granted the airport an easement right to create however much noise it was generating at the time the home was built.

However, if the airport began to increase noise levels beyond what they had homesteaded, or the situation was reversed, and the homes had been built first, and an airport came in later and started emitting noise that interfered with the homeowner’s use of their property, that would constitute a form of aggression that the airport owner would be liable for.

The same goes for any other type of pollution. And, in a more just system than what we have now, the same would go for any potential data center that generates excessive noise and light that interfere with nearby property owners’ use of their own property.

This is all to say that, while there are absolutely many ways that the current legal system gives the government—or coalitions of community members and special interests using the government—the ability to unjustly block the development of private property in ways that violate rights and make society as a whole poorer, it is also entirely possible that some of the opposition to data centers being built in and around residential neighborhoods is justified from a property rights standpoint.

But, beyond that, it’s also important to remember that many of the problems feeding the escalating debate over data centers stem from the absence of market forces.

The fact remains, as I mentioned at the beginning, that there is a lot of demand for AI right now. And that, in turn, translates to a heavy demand for computing power. But consumers have long made it clear that they prefer the bulk of computer processing to take place out of sight and out of mind.

People want their internet and AI tools to run quickly on their laptops and phones, and, therefore, for the heavy processing their digital tasks require to take place elsewhere. That requires data centers. And unless the government begins criminalizing AI tools or imposing heavy limits on how much we can all use the technology, the data centers need to go somewhere.

What’s frustrating is that this problem would, in all likelihood, be handled in a quiet and relatively painless way if there were a functioning market process at work with all the various components.

Again, people don’t want these data centers in or around their neighborhoods. And, if there were a totally free market in land, it would probably be prohibitively expensive to develop large data centers on land that could be used for multiple homes or residences that people would want to buy. The opportunity cost would be high.

Market forces, especially when grounded in a system of absolute property rights, would nudge developers away from residential neighborhoods and towards areas where large tracts of land are cheap—such as around other “unseemly” but necessary facilities like junk yards, factories, or landfills. But government zoning is currently able to bypass or distort this effect.

More remote areas could be a good option for exceptionally large data centers. But they can also be expensive to get water and power to. Whether it was worth the cost would be a pretty straightforward determination if there were a functioning market for electricity and water. But there isn’t.

In addition to the fact that much of the empty land in the US is government-owned—meaning its use is determined by political officials rather than the wants and needs of end consumers—the supply of water and power is also highly politicized. That means the cost of powering and cooling a data center often does not fall entirely on its owners.

The resource requirements for even large data centers are often somewhat exaggerated by opponents. However, especially when it comes to water, the pain is real for surrounding communities. But that’s not because of how much water the data centers need; it’s because government has constrained the supply of available water by granting monopoly privileges and policing water usage. Similar problems plague the electric grid, where grid size and capacity are primarily determined by government officials rather than consumer need.

In other words, the supply-constraining and incentive-distorting effects of government control over land, water, and electricity lie at the root of much of this escalating controversy with data centers. And as long as the public and political focus remains fixed on fake solutions that either leave those constraints in place or pile on even more, the problems will persist.