The Data Center Scare Is a Sham

Critics on both left and right are misleading the public about AI infrastructure. The numbers do not lie, but the activists do.

Here is a number for you: eighty. Not eighty percent, not eighty million. Eighty times. That is how much more water American corn farmers use every single year compared with every AI data center operating on the face of the earth. Eighty times more. Hold that number in your mind the next time a county commissioner in Fairfax or Frederick stands at a microphone and announces that a proposed data center is going to drain the local aquifer. The claim deserves the same respectful skepticism you would give a man selling oceanfront property in Iowa.

The anti-data-center movement has become one of the more curious political coalitions of our time. On the left, you have climate activists convinced that every server farm is a planetary emergency. On the right, you have populist critics who confuse their distaste for Silicon Valley’s cultural politics with sound infrastructure policy. Both camps are generating impressive amounts of heat and precious little light. The arithmetic, when examined honestly, does not support the hysteria.

Eighty. That is how many times corn farming out-consumes AI data centers for water. Where are the protests at the Iowa State Fair?

THE STATISTICS ARE BEING WEAPONIZED

Science communicator Hank Green spent considerable time recently dissecting the methodology behind AI water-use statistics, and his conclusion was blunt: the numbers can be made to say almost anything depending on what you choose to count. Count only the water touching the servers and you get a tiny figure. Count the water flowing through the power plants supplying the electricity, add in semiconductor manufacturing, fold in cooling infrastructure, and suddenly you have a number that sounds catastrophic. Critics naturally gravitate toward the largest possible figure. Supporters cite the smallest. Neither side is necessarily lying about the arithmetic. Both are often lying about the context.

The U.S. Geological Survey data is instructive here. Electricity generation, including all the power plants that illuminate American homes, factories, offices, and shopping malls, accounts for roughly forty percent of total freshwater withdrawals in the United States. Most of that water is returned to rivers and lakes after use. The distinction between water withdrawn and water consumed is real and significant. Critics who blur it are not educating the public. They are manipulating it.

Bruce Mehlman, the bipartisan policy analyst whose June 2026 research memo circulated widely among policymakers, offered an equally clarifying comparison: California almond orchards consume somewhere between three and five and a half million acre-feet of water annually. All North American data centers combined? A fraction of that. By Mehlman’s estimate, almonds alone use approximately four to seven times more water than every data center on the continent. Nobody is staging protests outside Blue Diamond.

California almonds consume four to seven times more water than all North American data centers combined. The math is not complicated. The politics just pretend it is.

LOCATION IS THE REAL VARIABLE

To the extent there is a legitimate water debate here, it is not about data centers as a category. It is about data centers in the wrong places. A water-cooled facility sited in the Sonoran Desert raises genuine questions about regional hydrology. The same facility in the Mid-Atlantic, where the Chesapeake watershed receives over forty inches of rainfall annually and where municipal infrastructure is engineered for robust industrial demand, raises a very different set of questions. The critics almost never make this distinction. Every data center, to hear them tell it, is equally dangerous whether it is dropping into a Nevada gulch or onto a former Montgomery County industrial parcel.

This is not policy analysis. It is theater. Smart infrastructure governance looks at the specific site, the specific water source, the specific cooling technology, and the specific regional capacity. Modern data center operators increasingly use recycled water, closed-loop cooling systems, and non-potable industrial water supplies. The industry is not standing still. The critics are.

ELECTRICITY IS THE LARGER STORY, AND CRITICS IGNORE IT

Here is what the water argument also obscures: electricity demand is almost certainly the more consequential long-term challenge, and it is one that has a clear, achievable solution. Green himself made this point directly. Power grids are straining not because data centers are uniquely sinful consumers but because America spent twenty years retiring baseload generation, tangling new construction in regulatory vines, and pretending that the intermittent output of wind and solar could substitute one-for-one for dispatchable coal and gas. The demand side is not the problem. The supply side is the problem.

Microsoft, Google, and Amazon are collectively committing over one hundred billion dollars annually to new AI infrastructure. That investment requires reliable power. The answer to that requirement is not to stop the investment. The answer is to start permitting new generation capacity at the speed the moment demands. Faster approvals, expanded transmission, modernized grid interconnection standards, and serious engagement with nuclear baseload options are the policy responses appropriate to the scale of what is being built. Blocking data centers does not solve the energy challenge. It simply relocates the investment to countries with fewer objections and fewer democratic accountability mechanisms.

The answer to America’s energy challenge is not to stop building. It is to start permitting.

COMMUNITIES ARE GETTING A REAL DEAL

The argument that communities receive nothing in return for hosting data centers is simply false, and in places like Northern Virginia it is demonstrably, empirically, verifiably false. Data center tax revenues in Loudoun County alone have grown to represent a substantial portion of the county’s operating budget, generating funding for schools, emergency services, road improvements, and public safety. The construction phase alone delivers thousands of jobs in electrical work, HVAC installation, structural engineering, fiber routing, and general contracting. The permanent operational workforce is small, as critics love to note, but by that standard the interstate highway system was also a terrible deal for America because it does not employ many people per lane-mile.

Infrastructure is not measured primarily by headcount. It is measured by the productivity it enables, the economic activity it supports, and the strategic position it secures. By all three measures, data centers are exceptional performers. They are the railroads of the AI age. The men who opposed railroad construction in the 1870s because trains were loud and disrupted the pastoral landscape were not wrong about the disruption. They were simply wrong about the tradeoff.

THE CONSERVATIVE CASE IS NOT COMPLICATED

For Republicans who have absorbed the right’s current populist skepticism of large technology companies, the temptation to join the anti-data-center coalition is understandable. The instinct, however, is being misapplied. Cultural antipathy toward Silicon Valley’s politics is a legitimate grievance. Restricting the construction of American economic and defense infrastructure because of that antipathy is a strategic own goal of historic proportions.

China is not having this debate. China is building. The People’s Liberation Army is not pausing its AI development program because a provincial official raised concerns about cooling tower condensation. American data center capacity is American AI capacity, which is American national security capacity. The supply chain for artificial intelligence runs through the physical infrastructure of compute, and that infrastructure must be built here, on American soil, under American law, answerable to American courts and American democratic institutions. Surrendering that ground in a fit of anti-tech populism would be one of the more consequential errors this generation of conservatives could make.

The principled center-right position on data centers is not complicated. Build the infrastructure. Site it intelligently. Negotiate fair tax arrangements that benefit host communities. Require modern water and energy efficiency standards. Expand power generation to meet the demand. Hold operators accountable through transparent, rules-based permitting rather than arbitrary veto. That is good governance. The alternative, amplifying activist talking points that dissolve under the most basic factual scrutiny, is not conservatism. It is noise dressed up as principle.

CONCLUSION

The activists will keep citing big numbers. They will keep presenting millions of gallons in isolation, stripped of every comparison that would render them comprehensible. They will continue conflating water withdrawn with water consumed, confusing temporary construction disruption with permanent environmental damage, and treating every data center permit as a crisis requiring mobilization. Let them. The answer is to keep returning to the facts.

Corn uses eighty times more water. Almonds use four to seven times more. Virginia and Maryland are not Arizona. Modern facilities use recycled water and closed-loop systems. Data centers generate billions in local tax revenue. American AI capacity is American strategic capacity. The math is not on the side of the critics. The only thing they have going for them is the certainty that most people will never check.

Check anyway.

---

NOTES AND SOURCES

1. Hank Green, “How Much Water Does AI Use?” (SciShow / YouTube, 2024) https://www.youtube.com/watch?v=b0C56yqIkbk

2. U.S. Geological Survey, “Estimated Use of Water in the United States” https://www.usgs.gov/mission-areas/water-resources/science/water-use-united-states

3. Bruce Mehlman, “Six-Chart Sunday: You Cannot Be Serious?!?” Age of Disruption Substack, June 7, 2026 https://brucemehlman.substack.com/p/six-chart-sunday-you-cannot-be-serious

4. U.S. Department of Energy, “Advancing Water Security at Data Centers” https://www.energy.gov/oe/clean-energy-resources-meet-data-center-electricity-demand

5. California Department of Food and Agriculture, Almond Acreage and Water Use Reports https://www.cdfa.ca.gov/Statistics/

6. Loudoun County, Virginia, FY2025 Budget Overview: Data Center Tax Revenue https://www.loudoun.gov/budget7. National Security Commission on Artificial Intelligence, Final Report (2021) https://www.dwt.com/-/media/files/blogs/artificial-intelligence-law-advisor/2021/03/nscai-final-report–2021.pdf

Dwight Patel is a Senior Contributor to Direct Line News, covering national security, cybersecurity policy, and federal governance, and is 1st Vice Chairman of the Maryland Republican Party

Direct Line News is a free-access conservative publication committed to good governance, fiscal responsibility, and honest commentary. Inspired by Thomas Paine’s Common Sense. No paywalls. No apologies.

About The Author