The Insight: The Data Center Dilemma

Massive data centers are popping up nationwide, fueling the AI boom and pleasing investors, but also riling communities. Do we need them? Are they really that bad? Is there a better way?

The Insight explores the “why” behind biased headlines and polarized politics, powered by your questions and balanced answers from our multipartisan news team.

1. How long have data centers existed, and what is their purpose?

By most accounts, the first-ever data center was the “mainframe” building built to house the first-ever computer, ENIAC (Electronic Numerical Integrator and Computer), at the University of Pennsylvania in 1945. As the internet developed, the number of rooms housing computer servers swelled, then exploded in the 90s. While those may have technically been data centers, today “data center” describes a computing mega-facility that houses the servers on which the internet as we know it runs.

According to DataCenterMap.com, a database that has existed since 2007, the US is home to 4,320 of the world’s 11,535 data centers. That’s over eight times more than the UK and Germany, which have the second most, at over 500 each.

In May alone, the US saw the construction of around 85 data centers. As Pew Research Center (Center) has documented, the majority of data center constructions are currently slated for rural areas, with over a third of them planned for counties that don’t yet have them.

Simply put, these data centers exist to provide the physical infrastructure necessary for the internet to run, and with the ongoing artificial intelligence (AI) boom, the demand for them has grown exponentially.

The traditional internet data centers, however, are different from the larger data centers being built to support AI processes. Traditional data centers handle general web traffic and cloud storage, whereas AI data centers are factories that specifically train and run the models. AI data centers use specialized processing units to process multiple massive, complex data sets at the same time. These specialized units, called GPUs, consume more electricity and produce more heat than the units in traditional data centers. This means AI data centers require more power to run and more water to cool them than other centers.

Want to go deeper on this issue?

Go beyond reading and discuss the issue with your fellow Americans. Participate in an AllSides Roundtables discussion next Tuesday to learn about the data centers issue through the eyes of your neighbors.

2. Are data centers central to the US’ tech race against China?

According to DataCenterMap, China only has 368 data centers, though it’s still the US’s top rival in the AI race. The Wall Street Journal’s Editorial Board (Lean Right), which has shown a strong pro-data center bias in recent months, has framed data center construction as fundamental to establishing AI dominance over China.

A feature published last week by Al Jazeera (Lean Left) argued that China could have an edge in the AI race because of its ability to produce cheap energy.

Multiple outlets on the right, including Daily Wire (Right) and New York Post (Lean Right), have raised concerns about Chinese foreign influence campaigns allegedly trying to stoke anti-AI sentiments among Americans in an attempt to gain an advantage in the race for AI dominance.

3. Why do data centers raise local electricity costs? What have governments done to address this?

A Bloomberg (Lean Left) analysis found that electricity costs for residents near data centers have increased by up to 267% as compared to costs in 2020. Another report from Carnegie Mellon University found data centers were driving up electricity bills by about 8% nationally and as much as 25% in some regional markets.

The main reason AI data centers raise local electricity costs is that they create such a large demand for electricity that utility companies often have to take on expensive grid updates to meet the demand. Some utility companies may build new power lines, substations, and power plants specifically to serve these large data centers, with the cost of these developments being passed along to everyone, not just the data center.

Additionally, some of the cost increase can be attributed simply to supply and demand. When AI data centers cause the demand to increase faster than utilities can increase supply, wholesale power prices increase, affecting regular customers’ bills.

Some states have tried to regulate this: California, Oregon, Texas, Virginia, and Tennessee are a few examples. Tennessee Gov. Bill Lee signed a law requiring data centers to pay for their own electricity infrastructure. Texas published draft rules that would impose financial obligations and disclosure requirements on data centers, and a New Jersey lawmaker proposed a bill requiring AI data centers to supply their own power from clean energy sources.

4. How much water do data centers use? To what extent do they drain water supplies?

Water is used in data centers to cool down the servers, which generate excess heat. As explained by “water intelligence” firm KETOS, the rate of water consumption varies widely depending on the cooling technique in use. Most existing data centers draw water from potable sources like evaporative cooling and consume around 70-80% of the water they withdraw. Some newer data centers recycle treated wastewater and therefore consume nearly 0% of the water they withdraw — but this technique uses more energy, a tradeoff most data center operators have so far been unwilling to make.

So, exactly how much water are we talking about here? In 2022, a scientist at a federally funded lab affiliated with the University of California told NPR (Lean Left) that a “mid-sized” data center consumes the same amount of water per day as 1,000 households. As for the water consumption of larger, “hyperscale” data centers, in 2025 a nonprofit research institute pegged the figure at 1.8 billion gallons per year, equivalent to a town of 10,000 to 50,000 people.

On the national level, these figures are a drop in the bucket. Per the KETOS explainer, all US data centers combined consume about 17.4 billion gallons of water per year, while the US Geological Survey estimates the overall rate of water consumption in the contiguous United States (excluding Alaska, Hawaii and island territories) at around 83 billion gallons per day. So data centers most likely account for less than 0.1% of overall US water consumption. Compare that to electricity generation, which accounts for about 3.5%, or agriculture, which according to the University of Michigan’s sustainability center represents a whopping 80-90% of US water consumption.

However, a data center’s impact on the local water supply can be much more severe. About 40% of them are located in areas of high or extreme water stress, like the American Southwest, according to Business Insider (Lean Left) and Newsweek (Center). In these areas, a water-inefficient data center can withdraw a sizable chunk from the potable water supply.

5. How do residents feel about the data center boom?

Public sentiment toward data center development is increasingly skeptical. According to Gallup (Center), roughly 71% of Americans oppose constructing data centers in their local areas. Reports from Data Center Watch (Not Rated) showed over $64 billion in projects were delayed or canceled between May 2024 and March 2025 due to bipartisan opposition. Fortune (Center) quoted a political risk analyst at the Data Center Watch initiative who said the impact was larger, with at least 48 data center projects representing $156 billion in investment being blocked or stalled by local opposition in 2025 alone.

Community concerns generally focus on noise, light and air pollution, water consumption, land usage, rising utility bills and a broader sense that residents are bearing the cost of these centers through tax breaks given to large tech companies.

Environmental activist and consumer advocate, Erin Brockovich, interviewed several community members about their concerns. Brockovich, who created a map of major AI data centers in the US, suggested one that’s less spoken about: the lack of transparency. One example of this includes a data center in Fayette County, Georgia, which was recently found to have used 30 million gallons of unreported water over 9-15 months. Brockovich cited residents who complained about “back-door deals,” “NDAs” and decisions being made without community input.

A writer for Vox (Left) proposed that growing anti-AI sentiment is itself a key driver of local data center opposition. It’s virtually impossible to mobilize against AI in digital space, she reasoned, but blocking construction of its physical infrastructure is an indirect way to struggle against it.

Not everyone views the data center boom as something to worry about. Some local officials, economic analysts, tech companies, real estate developers and some residents see meaningful opportunities with these centers if communities negotiate wisely and work together with their constituents.

An opinion in Fast Company (Lean Left) argued that data centers boost local labor markets, and can drive a shift toward green energy, attract business, create community partnerships and revitalize public infrastructure. The writer reported that Virginia’s data center industry added over 26,000 operational and construction jobs in 2023 and delivered $31.4 billion in economic output.

An Amazon Web Services (AWS) Economic Impact Study found that Amazon’s investment in data center development supported thousands of jobs across the US and generated an estimated $65.15 billion at local businesses through construction, operations and supply chain infrastructure.

6. Who should decide where data centers are built?

At the heart of the struggle over data centers spreading across the nation is a simple fact: the benefits of data centers are diffuse — anyone with an internet connection benefits from them — but their costs are locally concentrated. It’s a textbook example of a negative externality: a cost created by the business activity of some people but paid by other people. This sets up a clash between companies who want to monetize the benefits and local organizers who want to avoid the costs.

On paper, any community united in opposition to a data center has a clear upper hand. Localities in the US have exceptionally strong authority over zoning codes, so they can prevent practically any kind of building from being built in a particular place if they want to. However, the companies behind data centers have huge supplies of patience and money; if they’re stymied in one location, they can probably find another with residents that put up less resistance.

Since it sets regulations across all states and localities, the federal government could step in and correct for data centers’ negative externalities — for example, it could require data centers to offset any local increases in power consumption or capping the proportion of local water supplies they can withdraw. Such regulations would be much more powerful at the federal rather than local level because data center developers wouldn’t be able to just pack up and go somewhere the regulation doesn’t apply. But this would also help developers avoid a scenario where they’re completely shut out by unanimous local opposition across the country.

There’s just one problem: most Americans distrust the federal government’s ability to get AI regulation right. A Stanford University research center conducted a global survey on AI attitudes in 2025 and found that 31% of Americans trusted their government to regulate it, the lowest of the 35 surveyed countries.

Why are attitudes toward AI so much more positive in places like Asia and Africa than in the US? The director of a tech research group told Christian Science Monitor (Center) he sees these outcomes as downstream of the messages governments and private sectors are sending to the public about AI. According to him, US government entities and tech companies have neglected to put forth a vision of positive benefits AI could have for society; they’ve arguably done the opposite by warning of mass job displacement. Meanwhile, other nations portray AI as the solution to major problems faced by their citizens, such as growing professional services sectors or caring for the elderly.