Environmental Arguments Against AI Data Centers: Electricity, Water, Carbon, and Land Use

Quick Answer

AI data centers can add very large electricity loads faster than clean generation and grid infrastructure can be built.
Cooling can consume significant water, making location especially important in water-stressed regions.
Fossil-fuel electricity, backup engines, and on-site generation can add greenhouse-gas and local air-pollution impacts.
Large campuses and the infrastructure serving them can expand industrial land use, impervious surfaces, transmission, and development pressure.
The environmental case depends heavily on the individual project's power source, cooling design, location, and mitigation commitments.

AI data centers have become one of the fastest-growing new sources of electricity demand. That does not make every data center environmentally damaging, but it does make proposals worth examining carefully. The environmental footprint can change dramatically depending on where a facility is built, how it is cooled, and what actually supplies its electricity.

The strongest arguments against a proposed AI data center are therefore not vague claims about technology consuming resources. They focus on measurable local questions: How many megawatts will it require? Where will the electricity come from? How much water will cooling consume? What new generation, transmission, roads, and substations will be needed?

Those questions reveal where environmental opposition has a solid factual foundation and where newer designs can reduce the concern.

1. AI Data Centers Can Create Enormous New Electricity Demand

The scale and speed of electricity demand is one of the strongest environmental concerns. U.S. data centers consumed about 4.4% of national electricity in 2023, and Berkeley Lab projects that share could rise to approximately 6.7% to 12% by 2028.

Berkeley Lab estimated that U.S. data centers used about 176 terawatt-hours of electricity in 2023. Its 2028 scenarios range from roughly 325 to 580 terawatt-hours. The expansion of AI servers is an important reason for that increase because accelerated computing requires high-power processors and increasingly dense server installations.

The global trend points in the same direction. The International Energy Agency's updated outlook projects data-center electricity consumption rising from about 485 TWh in 2025 to around 950 TWh by 2030. Electricity consumed by AI-focused data centers is projected to grow substantially faster than overall data-center consumption.

From an environmental perspective, the concern is partly about timing. A data center can be developed faster than major transmission projects or many new power plants. If additional clean generation is not available when the facility starts operating, existing fossil generation or new dispatchable generation may help fill the gap.

The environmental argument is therefore stronger than simply saying that AI uses electricity. The real issue is whether the electricity system can satisfy a large new load without increasing fossil-fuel generation or slowing broader grid decarbonization.

2. Cooling Can Put Pressure on Local Water Supplies

Water is a legitimate concern, but the size of the impact depends heavily on the cooling system. The strongest environmental objections apply to water-intensive facilities proposed in places where freshwater is already scarce or contested.

Servers turn most of the electricity they consume into heat, and that heat must be removed continuously. Some data centers use cooling towers that reject heat through evaporation. Water lost through evaporation must be replaced, creating an ongoing demand for makeup water.

Berkeley Lab estimated direct water consumption by U.S. data centers at approximately 66 billion liters in 2023. Hyperscale and large colocation facilities represented most of that total. Water is also consumed indirectly when electricity generation itself requires water.

That national number does not mean every facility has the same impact. Cooling technologies differ substantially. Dry cooling, air-side economizers, closed-loop systems, reclaimed water, and other designs can reduce freshwater consumption. Climate also matters because cooling requirements vary by location.

There is also a tradeoff humans have somehow engineered into the problem: saving water can sometimes require more electricity. Berkeley Lab notes that air-cooled chillers can avoid on-site water consumption but generally use more energy than water-cooled systems. A serious environmental review therefore needs to examine both water and energy rather than celebrating improvement in one metric while quietly making the other worse.

3. More Computing Can Mean More Carbon Emissions and Local Air Pollution

A data center's operational emissions depend largely on the electricity supplying it. Facilities connected to fossil-heavy grids or supported by combustion-based on-site generation have a stronger environmental case against them than facilities supplied by additional low-carbon generation.

Berkeley Lab estimated that the regional grid electricity associated with U.S. data-center consumption in 2023 corresponded to about 61 billion kilograms of carbon-dioxide-equivalent emissions. The researchers also emphasized that emissions vary significantly by location because electricity mixes differ across the country.

The IEA expects renewable energy to supply a large portion of additional global data-center demand. But renewables are not expected to supply all of it. Natural gas and, in some regions, coal are also part of the projected generation mix used to serve expanding data centers.

Local air pollution can also enter the picture. Data centers need highly reliable electricity and commonly use stationary engines or turbines for backup or supplemental generation. The EPA maintains specific Clean Air Act resources for data centers because these combustion sources can be subject to air-pollution standards and permitting requirements.

That creates a practical environmental question for any project: What will supply power during normal operation, peak demand, and grid emergencies? A proposal backed by new clean generation and storage presents a very different environmental profile from one relying heavily on fossil generation or large fleets of combustion engines.

4. The Environmental Footprint Extends Beyond the Server Building

A large AI campus is not merely a warehouse containing computers. It may require roads, substations, transmission infrastructure, cooling equipment, power generation, parking, and extensive impervious surfaces.

Land development has environmental consequences even when the finished building itself emits little pollution. EPA notes that development and impervious surfaces can increase stormwater runoff, reduce soil infiltration, affect groundwater recharge, increase erosion, and alter water quality.

The supporting electricity infrastructure matters too. Very large loads may require new substations, transmission lines, generation facilities, or other upgrades. Those facilities have their own construction and land-use footprints.

For a project on previously developed industrial land, these concerns may be relatively limited. A greenfield campus that clears forest, affects wetlands, fragments habitat, or triggers substantial new infrastructure presents a different environmental calculation.

That makes siting one of the more defensible environmental arguments against particular projects. The question is not whether a data center occupies land. Nearly every human activity manages that feat. The question is whether the chosen site unnecessarily sacrifices environmentally valuable land when lower-impact alternatives exist.

5. The Strongest Environmental Case Is About Conditions, Not a Blanket Ban

Not all AI data centers have the same environmental footprint. Opposition is most evidence-based when it identifies specific problems with a proposed site's electricity supply, water source, cooling design, land conversion, or pollution controls.

A useful environmental review starts by asking whether new electricity demand will be matched with new clean generation rather than relying primarily on the existing grid. It should also examine whether the project can shift computing loads away from grid emergencies and whether storage or other flexibility can reduce peak demand.

For water, the relevant questions include the cooling technology, annual freshwater consumption, seasonal demand, local drought risk, and whether reclaimed or non-potable water can be used. A large national water-use statistic tells residents surprisingly little about what a specific facility will do to their watershed.

The same approach applies to land and air quality. A project on an existing industrial site with low-water cooling and new clean generation can address several major objections. A greenfield project dependent on scarce freshwater and fossil generation cannot make those concerns disappear merely by putting the letters “AI” on the economic-development brochure.

For environmental protection, that distinction matters. The goal is to identify projects whose environmental costs are unnecessarily high and push development toward lower-impact designs and locations.

Key Takeaways at a Glance

Electricity demand is the largest systemic issue. Rapid AI expansion can require substantial new generation and grid infrastructure.
Water impacts are highly local. Cooling technology and regional water scarcity matter more than a national average alone.
The power source determines much of the carbon impact. Fossil-heavy grids and combustion generation create a different footprint from additional low-carbon power.
Siting matters. Greenfield development, habitat disruption, and major infrastructure expansion can add environmental costs beyond the servers themselves.
Project-specific evidence is stronger than blanket opposition. Power, water, land, and pollution data provide the clearest basis for evaluating a proposal.

Environmental Issue	Main Concern	What to Examine
Electricity	Rapid growth in large power loads	New generation, grid capacity, peak demand
Water	Cooling can consume freshwater	Cooling design, water source, drought risk
Carbon	Additional electricity can increase fossil generation	Actual local generation mix and new clean power
Air quality	Combustion engines and turbines produce emissions	Backup generation, permits, operating hours
Land	Development can affect runoff and habitat	Site history, habitat, transmission, impervious area

What Makes an Environmental Objection Credible?

The best environmental argument against an AI data center is built around the actual project rather than the technology's reputation. Electricity demand, water consumption, generation sources, air permits, land conversion, and supporting infrastructure can all be examined before deciding whether the environmental cost is reasonable.

Some facilities can substantially reduce these impacts through low-water cooling, reclaimed water, cleaner electricity, energy storage, demand flexibility, efficient hardware, and careful site selection. Those improvements are important because they also reveal which impacts are avoidable.

For environmental protection, the most persuasive standard is therefore straightforward: large computing projects should demonstrate that their benefits do not depend on unnecessarily consuming scarce water, extending fossil-fuel dependence, or imposing avoidable ecological damage on the communities where they are built.