What Is a Data Center and Why Are They Bad for Power, Water, and Land

A data center is a purpose-built facility that houses servers, storage, and networking gear and supplies them with conditioned power, cooling, and physical security so digital services can run continuously. The "bad" label comes from resource intensity: large campuses can draw grid-scale electricity, consume millions of gallons of water for cooling, and lock up land, while backup generators and diesel fuel add air-quality risks. The facility is not the problem by itself; the concentration and growth of demand are.

What a data center physically does

Start with the workload. Racks of servers convert electricity into computation and, unavoidably, into heat. A modern server rack can dissipate tens of kilowatts, and high-density artificial-intelligence racks can exceed 100 kW per cabinet. That heat must be removed continuously, because semiconductor reliability degrades with temperature and thermal runaway can destroy equipment.

Power delivery is a chain of transformations. Utility electricity enters at medium voltage, passes through switchgear, transformers, and uninterruptible power supplies (UPS), then reaches server power supplies. Each conversion loses energy as heat, and those losses stack. A facility's power usage effectiveness (PUE) is total facility power divided by IT power; a PUE of 1.2 means 20 percent overhead on top of computing load.

Cooling is the second major system. Air-cooled halls push cold air through raised floors or containment aisles. Liquid cooling, including direct-to-chip and immersion, moves heat more efficiently and is increasingly used for dense AI hardware. Cooling equipment, pumps, and fans consume power and often evaporate water.

Redundancy defines reliability tiers. Uptime Institute's tier system ranges from Tier I (basic capacity, no redundancy) to Tier IV (fault-tolerant with concurrent maintainability). Higher tiers add duplicate power paths, generators, and cooling loops, which raise capital cost and standby fuel storage.

Plain language: a data center is a factory that turns electricity into computing and heat, and it needs machinery to move both.

How much power and water the scale implies

The numbers are large but uneven. Global data centers consumed roughly 1 to 2 percent of the world's electricity in recent estimates, and that share is rising with AI demand. Individual hyperscale campuses can draw 100 MW or more, comparable to a small city, and some planned AI campuses target 1 GW.

Water use varies enormously by design. Evaporative cooling can consume millions of gallons per day at large sites, while closed-loop or air-cooled designs use far less. A commonly cited industry metric is water usage effectiveness (WUE), liters of water per kilowatt-hour of IT load; low values indicate efficient water use, and high values signal heavy evaporation. Reported WUE values span a wide range, so site-level figures matter more than averages.

Here is a comparison of common cooling approaches:

The table shows the core trade-off: water and energy trade against each other. A site that saves water by running chillers spends more electricity, and a site that saves energy with evaporation spends more water.

Edge cases that change the verdict

Location dominates the impact. A facility in a region with a low-carbon grid and abundant water may be far less harmful than one powered by coal in a drought-prone area. Grid carbon intensity, water stress, and local air quality determine most of the real-world cost.

Utilization matters as much as capacity. A half-empty facility wastes embodied carbon and standby infrastructure. Hyperscalers typically run high utilization; speculative builds can sit underused.

Waste heat recovery is an emerging counterpoint. Some operators pipe server heat into district heating networks, offsetting fossil heating. The practice is real but still limited in scale.

Counterintuitive caveat: a data center can be greener than the alternative it replaces. Streaming a file may use less energy than driving to a store, and cloud consolidation can beat inefficient on-premises servers. The comparison, not the facility alone, determines whether it is "bad."

Red flags to watch in local proposals:

Water permits that exceed sustainable aquifer recharge in stressed basins. Diesel generator fleets sized for days of runtime with no emissions controls. Grid interconnection requests that defer needed transmission upgrades onto ratepayers. Noise and traffic impacts from continuous cooling and generator testing. Zoning that converts farmland or habitat without mitigation.

Misconceptions worth correcting

Misconception one: data centers are all equally wasteful. Efficiency varies widely. Well-designed hyperscale sites report PUE near 1.1 to 1.2, while older enterprise rooms can exceed 2.0.

Misconception two: water consumption is always huge. It depends on cooling design and climate. Closed-loop systems can cut evaporative losses dramatically.

Misconception three: the cloud removed the hardware. The cloud is data centers, just owned by someone else.

Misconception four: local opposition is only about water. Grid capacity, land use, tax incentives, and generator emissions frequently drive disputes.

Professional caveat: published PUE and WUE figures are often self-reported and measured differently across operators. Treat them as directional unless independently verified.

Timeline of a contested project

The timeline shows why projects surprise communities: the utility and permitting phases can outlast political attention spans, and decisions get made before public scrutiny peaks.

Practical actions for residents, buyers, and operators

For residents facing a proposed facility, request the water permit volume and source, the grid interconnection agreement, and the generator air permit. Ask for measured, not modeled, PUE and WUE from comparable operating sites. Push for closed-loop cooling in water-stressed areas and for waste-heat reuse commitments.

For buyers of cloud or colocation services, ask providers for site-level PUE, WUE, and carbon intensity, and prefer regions with low-carbon grids. For operators, prioritize high utilization, liquid cooling for dense racks, and transparent reporting.

Reconcile the thresholds explicitly: a PUE below 1.3 is generally efficient, WUE below 2 liters per kWh is low for evaporative designs, and grid carbon intensity below roughly 200 grams of CO2 per kWh indicates a relatively clean grid. These are rules of thumb, not universal standards, and should be verified against current operator disclosures.

Limits of what can be concluded

Data center impacts are real but highly site-specific, and current global statistics carry uncertainty because reporting is inconsistent. Energy and water figures change with technology, climate, and utilization, so older measurements may not reflect current operations. This article explains stable mechanisms and general ranges; it does not assess any specific project, and local decisions should rely on permits, utility data, and independent review.

This article is general educational information, not a legal, environmental, or financial opinion. It does not promise outcomes for any project or investment. For decisions about a specific facility, permit, or contract, consult a qualified local attorney, engineer, or environmental professional.