The Good News About America’s Data-Center Backlash: The Myths Are Finally Being Challenged

The good news is that even a newspaper generally identified with the liberal side of American political debate, The Washington Post, is beginning to challenge some of the myths, exaggerated claims and bad arithmetic driving opposition to AI data centers.

That matters. Data centers are rapidly becoming essential infrastructure for artificial intelligence, cloud computing, communications, scientific research, financial systems, government services and much of the modern digital economy. Americans should certainly insist that these facilities be built responsibly. But public policy should be based on measurable risks and engineering facts—not frightening numbers stripped of context and then amplified millions of times through social media.

That is essentially the argument explored in The Washington Post’s September 2, 2026, Impromptu episode, “The myths fueling America’s AI data center backlash.” Host Megan McArdle speaks with writer and former high-school physics teacher Andy Masley about claims that data centers are draining water supplies, heating surrounding communities and dramatically increasing electricity bills. The Post itself summarizes the problem succinctly: some of the most alarming environmental claims are based on “outdated estimates, missing context or questionable math.” (The Washington Post)

The Washington Post — “The myths fueling America’s AI data center backlash”

That is an important admission because the national argument over data centers has increasingly become an argument about perception as much as engineering.

How Big Numbers Become Frightening Numbers

One of Masley’s most useful contributions to the debate is remarkably simple: a number without a denominator tells us very little.

A report that a facility could consume millions of gallons of water sounds frightening. But the relevant questions are: millions compared with what? How much water does the regional system supply? Is the number describing withdrawal or actual consumption? Is the water potable or reclaimed? Is the estimate for normal operation, construction or electricity generation somewhere else? And how does the facility compare with other industrial, agricultural or municipal users?

Masley has repeatedly argued that failure to provide those denominators has distorted the debate. His analysis estimates that all U.S. data centers directly consumed roughly 17.4 billion gallons of water in 2023; his broader analysis also distinguishes direct data-center consumption from the additional water associated with generating their electricity. (Andy Masley)

His criticism becomes particularly interesting when applied to social media.

Masley examined videos produced by More Perfect Union, an organization with a very large online following, and argues that enormous water numbers are repeatedly presented without the context necessary for viewers to judge their significance. One example he analyzes describes 7.2 million gallons per year—a number that sounds enormous—while Masley calculates that it would represent only about 0.02 percent of nearby El Paso’s water use. (Andy Masley)

That does not prove that every proposed data center is appropriate. It demonstrates something more important: scale must be established before risk can be established.

Social Media Amplifies Emotion Better Than Engineering

This is where the Washington Post discussion reaches beyond data centers.

Social-media systems are extraordinarily effective mechanisms for distributing information. They are not necessarily mechanisms for determining whether that information is true.

A statement such as “this facility will consume millions of gallons of water” can travel quickly because the number sounds enormous. A technically meaningful explanation—what percentage of regional capacity that represents, whether the water is consumed or returned, whether recycled water can be substituted, what cooling technology is involved and what enforceable limits apply—is much less emotionally powerful.

The result is an information asymmetry.

Fear can be communicated in ten seconds. Engineering context may require ten minutes.

And algorithms optimized for engagement have no inherent reason to prefer the ten-minute explanation.

The Washington Post episode is therefore important not merely because it challenges particular data-center claims. It recognizes the mechanism by which questionable claims can become conventional wisdom: alarming statistics are repeated without denominators, qualifications or engineering context until repetition itself begins to substitute for evidence. The Post explicitly frames its discussion around both separating legitimate concerns from myths and understanding why misleading claims spread so readily. (The Washington Post)

There is already a striking documented example of how serious these numerical errors can become. WIRED reported that journalist Karen Hao acknowledged that a figure in her book Empire of AI concerning a proposed Google data center near Santiago, Chile, was wrong by a factor of approximately 1,000, apparently because of a unit error. Masley was the person who brought the error to her attention. (WIRED)

That is precisely why technical literacy matters.

But We Should Not Replace One Myth With Another

There is an equally important caution.

The correct response to exaggerated claims about data centers is not to claim that data centers have no significant environmental or infrastructure consequences.

They do.

Electricity is the clearest example. Lawrence Berkeley National Laboratory’s latest national analysis estimates that data centers could consume about 11.8 percent of total U.S. electricity in 2030, with scenarios ranging from approximately 9.5 to 15.3 percent. Its earlier analysis estimated that data centers consumed approximately 4.4 percent of U.S. electricity in 2023. (LBL ETA Publications)

That is substantial infrastructure demand.

The International Energy Agency likewise projects that U.S. data-center electricity consumption will rise by roughly 240 terawatt-hours between 2024 and 2030. Because data centers are geographically concentrated rather than evenly distributed across the country, the IEA specifically warns that integrating those loads can create localized grid challenges even when national percentages appear manageable. (IEA)

Masley himself acknowledges legitimate problems. Among the concerns he says withstand scrutiny are additional carbon emissions, particular cases of local air pollution, noise when facilities are built too close to residences, and circumstances in which large loads can contribute to higher electricity costs. (Andy Masley)

That distinction is crucial.

“There are risks” is not the same proposition as “the technology is dangerous.”

The American response should therefore be neither blind acceptance nor reflexive prohibition. It should be engineering, regulation, measurement and correction.

America Already Knows How to Do This

Data centers are not the first large technological infrastructure America has had to integrate into communities.

We have spent more than a century developing regulatory, engineering and economic mechanisms for managing electric generating plants, factories, telecommunications networks, pipelines, airports, highways, water systems and other large infrastructure.

The appropriate question is therefore not simply:

“Does this data center use electricity or water?”

Of course it does.

The useful questions are:

How much? From what source? At what time? At whose expense? What additional generating, transmission, water or wastewater capacity is required? Who pays for it? What happens during drought or grid emergencies? What are the measurable noise and emissions limits? What happens if those limits are exceeded? And what financial responsibility does the developer assume for infrastructure constructed primarily to serve its facility?

Those questions transform an emotional argument into an engineering and public-policy problem.

And engineering problems can be managed.

The Correction Process Is Already Underway

The industry and its regulators are not standing still.

Berkeley Lab maintains an entire data-center energy research program devoted to efficiency, cooling, monitoring, controls, water use and integration with the electric grid. Its work specifically includes technologies and operating practices intended to reduce resource consumption while maintaining reliability. (Systems & Energy Technologies Analysis)

The International Energy Agency reports something equally important: electricity consumption per AI task is declining rapidly as computing efficiency improves. Total demand nevertheless continues to rise because AI usage and increasingly compute-intensive applications are growing even faster. (IEA)

That is exactly the kind of distinction routinely lost in social-media arguments. Both statements can simultaneously be true: AI computation can become dramatically more efficient while total data-center electricity consumption increases.

Cooling technology is changing. Computing efficiency is improving. Utilities are developing new rate structures for enormous loads. Developers increasingly face requirements concerning grid interconnection, generation, transmission and infrastructure costs. New nuclear, natural-gas, renewable and geothermal resources are being considered specifically because of growing computing demand. The Department of Energy is explicitly treating data-center expansion as both a challenge and an opportunity to expand and modernize America’s electricity infrastructure. (The Department of Energy’s Energy.gov)

In other words, identifying a problem today does not mean projecting that problem unchanged indefinitely into the future.

Technology responds.

Markets respond.

Utilities respond.

Communities respond.

Regulators respond.

And engineering responds.

The Real Lesson From The Washington Post

That may ultimately be the most important message in this Washington Post discussion.

The data-center debate should not be divided into people who “support AI” and people who “protect communities.” America can—and must—do both.

A poorly located data center should be relocated.

A facility producing unacceptable noise should be required to mitigate it.

A project requiring substantial new electric infrastructure should not quietly transfer unreasonable costs to residential customers.

A facility requiring substantial water in a water-constrained region should demonstrate where that water will come from and how the community will be protected.

Air emissions should be measured and regulated.

Promises made during permitting should become enforceable obligations where appropriate.

But none of those requirements logically leads to the conclusion that America should stop building the computing infrastructure required for artificial intelligence.

Indeed, the newest Berkeley Lab analysis suggests how large that infrastructure challenge actually is: its reference case estimates approximately 649 terawatt-hours of U.S. data-center electricity consumption in 2030. (BI Energy Systems Division)

America therefore faces a genuine infrastructure challenge—not an apocalypse.

Misinformation Has Consequences Too

There is another risk that receives too little attention: bad information can itself produce bad public policy.

If exaggerated claims convince communities that virtually any large data center will exhaust their water supply, poison their air or catastrophically increase electricity rates, communities may reject projects regardless of their actual engineering characteristics.

That has consequences.

Artificial intelligence is becoming infrastructure for scientific research, medicine, manufacturing, engineering, cybersecurity, defense, transportation, finance and virtually every other technology-intensive sector. The United States and China are already the two largest centers of data-center electricity consumption growth; the IEA expects them together to account for nearly 80 percent of global growth through 2030. (IEA)

The United States cannot maintain technological leadership while simultaneously deciding that the physical infrastructure underlying advanced computing should simply be built somewhere else.

Nor should technological competition become an excuse for irresponsible construction.

Those are false alternatives.

The American answer should be to build it—and build it correctly.

From Fear to Evidence

The Washington Post deserves credit for putting this argument before a broad audience.

The important development is not that every criticism of data centers has suddenly been disproved. It hasn’t.

The important development is that a major national publication is explicitly asking readers and listeners to distinguish legitimate concerns from myths, and to examine the outdated assumptions, missing context and questionable arithmetic behind some of the most alarming claims. (The Washington Post)

That is exactly the debate America needs.

We should scrutinize every major infrastructure project. We should protect neighboring property owners. We should protect water resources. We should prevent residential electric customers from improperly subsidizing enormous private loads. We should regulate noise and emissions. And we should insist that developers disclose enough technical information for communities to make informed decisions.

But we should do those things based on measurements, engineering and evidence—not viral posts designed primarily to frighten people.

America has repeatedly encountered the unintended consequences of technological change. Our strength has not been avoiding technology. It has been developing institutions capable of identifying problems, correcting mistakes, establishing standards and allowing innovation to continue.

The rapid development of AI infrastructure should be approached in exactly the same way.

Enable the technology. Protect the community. Correct the problems. Continue to innovate.

That is a much stronger response than allowing misinformation—or fear of technology itself—to determine America’s technological future.

References

1. McArdle, Megan, host. “The Myths Fueling America’s AI Data Center Backlash.” Impromptu, The Washington Post, September 2, 2026. Interview/discussion with Andy Masley. The program examines claims concerning data-center water consumption, electricity prices, environmental effects, and the role of questionable calculations and missing context in the growing public backlash against AI infrastructure.
   The Washington Post — The Myths Fueling America’s AI Data Center Backlash
2. The Washington Post Editorial Board. “The Right Response to the Data Center Backlash.” The Washington Post, May 13, 2026. Editorial examining rising public opposition to data centers, electricity costs, water-use claims, infrastructure requirements, closed-loop cooling, local tax revenue, and the strategic importance of continued U.S. data-center development. (The Washington Post)
   The Washington Post — The Right Response to the Data Center Backlash
3. Ovide, Shira. “Data Centers Have United Americans of Both Parties in a Shared Hatred.” The Washington Post, July 18, 2026. Examination of the rapid growth of bipartisan opposition to data centers, including concerns about electricity, water, land use, community involvement, permitting, economic benefits and public distrust. (The Washington Post)
   The Washington Post — How Data Centers Became a Symbol of Americans’ Rage
4. Smith, Sarah Josephine, Alex Hubbard, Alexander Newkirk, Mohan Ganeshalingam, Billie Holecek, Dale A. Sartor, Michael Mills, and Arman Shehabi. United States Data Center Energy Usage Report: 2025 Update. Lawrence Berkeley National Laboratory, June 2026. The report estimates that U.S. data centers could account for approximately 11.8 percent of total U.S. electricity consumption by 2030 in its reference case, with sensitivity scenarios ranging from 9.5 to 15.3 percent. Its reference case projects approximately 649 TWh of data-center electricity consumption in 2030. (Energy Technologies Area)
   Lawrence Berkeley National Laboratory — United States Data Center Energy Usage Report: 2025 Update
5. Shehabi, Arman, Sarah Josephine Smith, Alex Hubbard, Alexander Newkirk, Nuoa Lei, Md AbuBakar Siddik, Billie Holecek, Jonathan G. Koomey, Eric R. Masanet, and Dale A. Sartor. 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory, December 19, 2024. LBNL-2001637. DOI: 10.71468/P1WC7Q. This congressionally requested study provides historical estimates of U.S. data-center electricity consumption and scenarios for future demand through 2028. (LBL ETA Publications)
   Lawrence Berkeley National Laboratory — 2024 United States Data Center Energy Usage Report
6. International Energy Agency. Energy and AI. Paris: International Energy Agency, April 10, 2025. Comprehensive assessment of the relationship between artificial intelligence, data centers and energy systems, including electricity demand, generation requirements, emissions, grid effects, efficiency improvements and energy security. (IEA)
   International Energy Agency — Energy and AI
7. International Energy Agency. “Energy Demand from AI.” In Energy and AI. Paris: International Energy Agency, 2025. The IEA projects global data-center electricity consumption of approximately 945 TWh in 2030 in its base case. It estimates U.S. consumption will increase by roughly 240 TWh between 2024 and 2030 and identifies the United States and China as accounting for nearly 80 percent of global data-center electricity-demand growth during that period. (IEA)
   International Energy Agency — Energy Demand from AI
8. International Energy Agency. “Energy Supply for AI.” In Energy and AI. Paris: International Energy Agency, 2025. Analysis of the electricity-generation resources likely to supply expanding data centers, including natural gas, renewables, nuclear energy and coal, and differences between U.S., Chinese, European and other regional electricity systems. (IEA)
   International Energy Agency — Energy Supply for AI
9. International Energy Agency. “AI and Climate Change.” In Energy and AI. Paris: International Energy Agency, 2025. Analysis of the direct and indirect emissions associated with data-center electricity consumption and projected changes as global computing demand expands. (IEA)
   International Energy Agency — AI and Climate Change
10. Lei, Nuoa, Jun Lu, Arman Shehabi, and Eric R. Masanet. “The Water Use of Data Center Workloads: A Review and Assessment of Key Determinants.” Resources, Conservation and Recycling, vol. 219, 2025, article 108310. Technical review of the factors determining data-center water consumption and an important scientific source for understanding why water-use estimates depend on cooling technology, location, climate, workload and electricity generation. (Data Center Energy)
    Lawrence Berkeley National Laboratory — Data Center Publications
11. Taft, Molly. “You’re Thinking About AI and Water All Wrong.” WIRED, 2026. Examination of claims concerning AI and data-center water consumption, including journalist Karen Hao’s acknowledgment that a water-use figure concerning a proposed Google data center near Santiago, Chile, had been overstated by approximately a factor of 1,000 because of a unit error. The article also discusses Andy Masley’s role in identifying the problem and the importance of evaluating water consumption in its local and technological context. (WIRED)
    WIRED — You’re Thinking About AI and Water All Wrong
12. Lawrence Berkeley National Laboratory. “Data Centers.” Center of Expertise for Data Center Energy. Research and publication program addressing data-center energy efficiency, cooling, water consumption, grid integration, load flexibility, electricity-rate design and related infrastructure issues. The program provides useful technical background for evaluating both the costs of rapidly expanding computing infrastructure and technologies capable of mitigating those impacts. (Data Center Energy)
    Lawrence Berkeley National Laboratory — Data Center Research and Publications