More data center controversy

 I saw an illustration comparing a nearby data center to the town of Abilene, Texas.  The power consumption of the two was comparable.  It might be that power consumption at such places would be a first priority for the owners.

Very likely. Electricity is often the single largest operating expense for a modern data center, especially one heavily devoted to AI training or inference.

The comparison to Abilene is striking because Abilene has around 125,000 people. A large AI-oriented data center campus can indeed consume power on the scale of a medium-sized city. The newest facilities being discussed are measured not in megawatts but in gigawatts—power levels once associated mainly with heavy industry.

Because of that, data center operators have a huge incentive to reduce energy costs. They pursue it in several ways:

    Locating where electricity is cheap. Texas is attractive partly because of relatively low power prices and a business-friendly environment.

    Improving cooling efficiency. A substantial fraction of a data center's power goes not into computation but into removing the heat created by computation.

    Designing more efficient chips. Every watt saved per processor becomes enormous when multiplied by hundreds of thousands of processors.

    Running workloads when power is abundant. Some computing tasks can be shifted to times when renewable generation is high or prices are lower.

The interesting wrinkle is that what is best for the data center owner is not always what is best for the local grid. A company can save money by locating where power is inexpensive, but if enough large facilities arrive at once, they can strain transmission lines, raise local electricity prices, or force construction of new generation capacity. That's where many of the current controversies arise.

Your earlier point about subsidies fits into this. If a data center would come because power is cheap and land is available, then the economic case for additional tax breaks can be questioned. Local officials often see the announcement of a billion-dollar facility and thousands of construction jobs, while critics ask whether the public is giving up more in tax revenue and infrastructure costs than it receives in return.

Historically, towns worried about factories because of smoke and water pollution. Today, many communities worry about data centers because of electricity demand, water consumption for cooling, and the need for new power plants. It's a different kind of industrialization, but in some respects it raises familiar questions: who gets the benefits, who bears the costs, and how should those costs be allocated?

It's not quite the issue of building a prison or a nuclear facility next door.

No, not quite. A prison or nuclear facility tends to provoke an immediate "not in my backyard" reaction because people associate them with danger, stigma, or falling property values.

A data center is more ambiguous. Most people looking at one would see a large, rather boring industrial building. It doesn't emit smoke, doesn't handle hazardous chemicals on the scale of a refinery, and generally operates quietly from a distance. In that sense, it's easier for a community to accept.

The opposition usually arises when people learn the scale of the resource consumption. A resident may initially think, "It's just a big warehouse full of computers." Then they hear it will consume as much electricity as a city, require new transmission lines, draw millions of gallons of water for cooling, or receive substantial tax incentives. The debate shifts from fear to questions of fairness and priorities.

There is also a psychological factor. People generally understand what a factory produces. They can see cars, steel, or furniture coming out the other end. A data center's output is intangible. If a facility uses as much power as Abilene, many residents naturally ask, "What are we getting in return?" The answer may be cloud services, AI training, financial transactions, video streaming, or internet infrastructure, but those benefits are dispersed globally rather than visibly concentrated in the local community.

That may be one reason data centers generate a different kind of uneasiness than traditional industry. The concern is less "this thing is dangerous" and more "this thing is enormously consuming resources, yet its purpose feels remote from our daily lives."

As AI expands, that latter concern seems to be growing. Even people who are enthusiastic about technology sometimes wonder whether dedicating city-sized amounts of electricity to AI systems is the best use of those resources. That's a policy question rather than a technical one, but it is becoming a common topic in public discussions about new data center projects.

Such structures would be easy to ornament and landscape.

Exactly. That's one of the things that makes a data center unusually amenable to being made a good neighbor.

The building itself doesn't need windows, storefronts, or much street-facing activity. That gives the architects considerable freedom to put the functional box behind attractive landscaping, berms, trees, ponds, and perhaps a pleasant architectural façade. A few hundred yards of trees can make an enormous difference when the thing behind them is basically a giant windowless box.

And unlike a prison, there is no particular reason for the security perimeter to look forbidding. It can be designed almost like a corporate campus or industrial park. Even the cooling equipment and substations can be screened architecturally.

There is a funny contrast here: the physical presence of a data center can be quite benign while its invisible footprint—electricity, water, transmission capacity—is enormous. You could make a two-hundred-acre data center campus look like a beautiful park and still have an electrical load comparable to a town.

That actually strengthens the case for separating the aesthetic question from the resource question. If the electricity and water costs are properly borne by the operator, I don't see why the building itself has to be an eyesore.

I don't think anything is as capable of being a boring eyesore as an insurance company home office and accompanying parking, but measures to mitigate that payed off for us over and over.

Oh, absolutely. The classic insurance home office: a big rectangular box, acres of asphalt, and a heroic amount of beige. 😄

But your point about mitigation is important. Aesthetic improvements can have a very practical return even when they don't add a penny of productive capacity. Landscaping, setbacks, trees, attractive entrances, screening parking lots—all of those can change how the surrounding community experiences the same building.

And there's a feedback effect. If the neighbors regard the building as a reasonably attractive part of the community, they're less likely to resent every other aspect of its presence. An ugly building can make people notice its traffic, lighting, parking, and noise; a well-designed one tends to disappear into the landscape.

That could be particularly valuable for data centers. They have so little need for the things that normally make an office building visually interesting—windows, pedestrians, storefronts—that landscape architecture could do most of the work. Put the enormous box well back from the road, plant mature trees, berm the site, hide the equipment, make the entrance attractive, and suddenly the public's experience might be "a nice wooded industrial property" rather than "that gigantic computer building."

And I suspect your experience is the more general lesson: you don't necessarily have to eliminate an unattractive or intrusive use; you can make its externalities much less noticeable. If the mitigation costs are modest relative to the value of the facility, it's often money very well spent.