The issue is not whether data centers should exist. The issue is whether private companies should be allowed to privatize the benefits while socializing the costs.

By Davidws10

The public debate over artificial intelligence has become too abstract. One side speaks as if AI is an inevitable technological salvation. The other warns as if every server rack is a step toward civilizational ruin. Both framings miss the more immediate problem.

AI is not merely software. It is physical infrastructure. It requires land, electricity, water, transmission capacity, backup generation, roads, cooling systems, tax treatment, and local permission. At hyperscale, AI data centers are not just another commercial building. They are industrial facilities, and they should be regulated as such.

That does not mean banning them. It means ending the fiction that they are weightless.

Lawrence Berkeley National Laboratory has reported that U.S. data-center electricity use rose from 58 terawatt-hours in 2014 to 176 terawatt-hours in 2023. Data centers accounted for about 4.4 percent of total U.S. electricity consumption in 2023, and that share is projected to rise to between 6.7 and 12 percent by 2028. The same report notes that data-center power demand more than doubled from 2017 to 2023, largely because of AI servers.¹

Globally, the International Energy Agency estimates that data centers consumed about 415 terawatt-hours in 2024, or roughly 1.5 percent of global electricity use. Its base-case projection has that figure roughly doubling to about 945 terawatt-hours by 2030. At a planetary scale that may still sound manageable. At a local scale, it can become a serious problem, because data-center demand is concentrated in particular regions and on particular electric systems rather than evenly spread across society.²

That concentration is where the public interest begins.

A community may be told that a new data center means jobs, investment, and technological prestige. Sometimes that is true. But the local balance sheet is not always clean. Data centers can require new substations, transmission upgrades, backup generation, water capacity, and road improvements. They can increase electricity demand faster than utilities can responsibly plan. They can impose noise, land-use burdens, and water stress. And when utilities recover infrastructure costs through broad rate increases, households and small businesses can end up paying for a buildout they did not ask for and may not meaningfully benefit from.

This is the heart of the matter. The question is not whether AI companies should be allowed to build. The question is whether they should be allowed to build on public systems without paying the full cost of what they consume.

The Harvard Electricity Law Initiative has warned that utility rate structures and special contracts can force ratepayers to help fund discounted power for large technology customers. Whether one agrees with every part of that critique or not, the underlying problem is obvious: utilities exist under public authority, and the costs of serving one very large customer can be shifted unless regulators require transparency and fair allocation.³

That is not anti-technology. That is basic fairness.

A farmer who needs irrigation, a factory that needs permits, a port that needs dredging, or a refinery that needs environmental review is not allowed to pretend its physical footprint does not exist. AI data centers should receive no special exemption simply because their product sounds futuristic.

The water issue deserves the same practical treatment. Not every data center uses the same amount of water. Cooling methods vary. Some newer systems reduce direct water use significantly. But in the wrong location, at the wrong scale, water demand can become a public concern. The Environmental and Energy Study Institute has noted that large data centers can consume up to 5 million gallons of water per day, comparable to the water use of a town of 10,000 to 50,000 people.⁴

That does not mean every data center is a water crisis. It does mean communities have a right to know, before approval, how much water will be withdrawn, how much will be consumed, what happens during drought, and whether public water systems will be expanded for a private customer.

The same principle applies to clean-energy claims. A company should not be allowed to announce that a facility is powered by clean energy if the practical effect is to draw heavily from a local grid still dependent on fossil generation while relying on paper accounting elsewhere. Clean-energy commitments should be real, local, time-aligned, and enforceable. If the load runs 24 hours a day, the power plan should address 24 hours a day.

There is a responsible path forward. It begins by treating AI data centers as major-load industrial users.

They should be required to disclose projected electricity demand, water demand, backup generation, emissions, and cooling methods before approval. They should pay directly for the grid upgrades they require. They should be subject to tariffs that protect residential and small-business ratepayers from cost shifting. They should not receive tax abatements unless the public benefit is specific, measurable, and greater than the public cost. They should face stricter scrutiny in water-stressed regions. They should participate in demand-response programs where feasible. Local communities should retain real zoning authority, public hearings, noise standards, and setback rules.

None of this is radical. It is normal public governance.

The strongest argument for AI is that it may produce immense public value: medical research, logistics efficiency, scientific discovery, language access, education, defense analysis, engineering, productivity, and tools not yet imagined. That argument deserves to be taken seriously. But a technology that claims to serve the future must not be permitted to quietly degrade the present.

The public should reject both extremes. AI data centers are not inherently evil. Nor are they harmless engines of progress. They are large industrial consumers of power, water, land, and public infrastructure. They must be judged accordingly.

The proper policy is not panic. It is discipline.

Let the companies build where the grid can support them. Let them innovate in cooling, efficiency, and energy management. Let them buy or build clean power that actually serves the load they create. But do not let them shift the bill to households. Do not let them drain local water without scrutiny. Do not let them hide behind nondisclosure agreements while public utilities reconfigure themselves around private demand. Do not let tax incentives become tribute paid by small communities to some of the wealthiest corporations on earth.

AI may be digital. Its costs are not.

If the future requires more data centers, then the future also requires stronger rules. Regulate them as industrial infrastructure. Make them transparent. Make them pay their way. Make them prove the public benefit.

That is not anti-AI. It is pro-accountability.

Source Notes

1. Lawrence Berkeley National Laboratory, “Berkeley Lab Report Evaluates Increase in Electricity Demand From Data Centers,” reporting U.S. data-center electricity use, share of total electricity use, and 2028 projections. Source link

2. International Energy Agency, “Energy and AI: Energy Demand from AI,” reporting global data-center electricity use and 2030 projections. Source link

3. Harvard Electricity Law Initiative, “Extracting Profits from the Public: How Utility Ratepayers Are Paying for Big Tech’s Power,” discussing utility rates, special contracts, and ratepayer risk. Source link

4. Environmental and Energy Study Institute, “Data Centers and Water Consumption,” discussing water use and community impacts. Source link