Data centres: The Water Paradox of the Digital Age

Introduction

Data centres are the physical backbone of the global digital economy, enabling cloud computing, AI, and streaming services. While their energy consumption is widely discussed, their water footprint—often indirect and highly localised—is less scrutinised. This analysis examines the impact of data centre water usage on local water supply, ecology, and human health, concluding that without rigorous siting and technology choices, data centres can severely degrade living conditions in water-stressed regions.


1. How Data Centres Use Water: Direct and Indirect Consumption

Data centres consume water primarily for cooling (to dissipate heat from servers) and, to a lesser extent, for humidification and on-site power generation (e.g., evaporative cooling in gas turbines).

  • Direct water use (withdrawal & consumption): Most centres use evaporative cooling (cooling towers) or adiabatic systems. Water evaporates, removing heat but not returning to the local basin. Typical water usage effectiveness (WUE) ranges from 0.2 to 1.8 L/kWh. A 20-megawatt data centre can consume 1–4 million gallons (3.8–15 million litres) of water per day—equivalent to the daily water use of a town of 10,000–40,000 people.
  • Indirect water use: Water consumed at power plants (thermoelectric cooling) to generate electricity for the centre. In regions with once-through or evaporative cooling at power plants, this doubles the water footprint.

Key nuance: Unlike agriculture (consumptive but productive) or municipal use (returned as treated wastewater), data centre cooling water often undergoes chemical treatment (biocides, anti-scaling agents) and is evaporated, leaving concentrated brine or contaminants in blowdown discharge.


2. Impact on Local Water Supply: Competition and Scarcity

The severity depends entirely on baseline water stress and source of water.

2.1 Competition with municipal and agricultural users

In water-scarce regions (e.g., semi-arid US West, parts of India, Spain, Chile), data centres often draw from the same groundwater aquifers or surface water rights as farms and cities.

  • Example (Dalles, Oregon): Google’s data centres consumed nearly 30% of the city’s total water supply in 2021, drawing from the Columbia River but also from municipal wells during drought, prompting legal conflicts with local irrigators and residents.
  • Example (Meath County, Ireland): Several proposed data centres faced resistance because the region already had “regular water supply deficits” according to Irish Water, threatening residential connections.

2.2 Aquifer depletion and saltwater intrusion

Heavy groundwater pumping for data centre cooling can lower the water table, increasing pumping costs for households and causing wells to dry up. In coastal areas, this can induce saltwater intrusion into freshwater aquifers, rendering them unusable for drinking or irrigation.

2.3 Timing of use vs. natural recharge

Data centres operate 24/7, withdrawing water even during droughts or low-flow periods when rivers and aquifers are most stressed. This baseload water demand exacerbates seasonal scarcity far more than intermittent agricultural use.


3. Ecological Implications: From Streamflow to Biodiversity

Water consumption is not simply “lost”—it is removed from a local hydrological system, with cascading ecological effects.

3.1 Reduced streamflow and aquatic habitat

Evaporative cooling removes water from the watershed. Lower streamflows increase water temperatures, reduce dissolved oxygen, and concentrate pollutants.

  • Impact: Fish species (e.g., salmon in Pacific Northwest) rely on minimum flows for migration and spawning. During summer low flows, data centre consumption can push rivers below critical ecological thresholds, causing fish kills or population collapse.

3.2 Thermal pollution from discharge

Even data centres that return water (e.g., once-through cooling) discharge heated water, raising ambient river temperatures. Many aquatic species are ectothermic; a 2–3°C rise can shift reproductive cycles, favour invasive species, and induce algal blooms.

3.3 Chemical contamination of surface and groundwater

Cooling water contains biocides (e.g., chlorine, bromine, or isothiazolones), corrosion inhibitors (e.g., benzotriazole), and anti-scaling agents (e.g., phosphonates).

  • Blowdown discharge (wastewater) is often released into municipal sewers or directly into water bodies. Wastewater treatment plants are not designed to remove many of these chemicals, leading to their accumulation in rivers and groundwater.
  • Ecotoxicity: Benzotriazole is persistent and toxic to aquatic organisms (algae, daphnia, fish) at low µg/L concentrations. Biocides can disrupt microbial communities, which underpin nutrient cycling.

3.4 Habitat fragmentation from infrastructure

Pipelines, pump stations, and intake structures can alter river morphology, block fish passage, and disturb riparian zones.


4. Health Implications for Nearby Populations

Health effects arise via water quantity (lack of access) and water quality (chemical contamination).

4.1 Direct water scarcity and hygiene

When data centres compete with municipal supply during droughts, residents face:

  • Water rationing (reduced pressure, scheduled cuts)
  • Increased water bills due to higher marginal costs or infrastructure surcharges
  • Inability to maintain basic hygiene – linked to skin infections, diarrheal diseases (if alternative water sources are unsafe), and mental stress.

4.2 Drinking water contamination

  • Nitrates and pathogens: Lowered water tables can draw in nitrates from septic systems or agricultural runoff, leading to methemoglobinemia (“blue baby syndrome”) in infants.
  • Heavy metals: Changes in redox conditions from aquifer depletion can mobilise arsenic, lead, or uranium from geological formations into drinking water wells.
  • Disinfection by-products (DBPs): If municipal water systems respond to scarcity by using more recycled or surface water with higher organic content, chlorination can form trihalomethanes (THMs) and haloacetic acids (HAAs), linked to bladder cancer and reproductive harm.

4.3 Chemical contaminants specific to data centre discharge

Few studies have examined human exposure to cooling water additives via drinking water. However:

  • Benzotriazole is not regulated in drinking water in most countries but is classified as an emerging contaminant; chronic exposure is linked to liver and thyroid effects in animal models.
  • Phosphonates can chelate metals, increasing their solubility and transport into drinking water sources.

4.4 Air quality and respiratory health

Evaporative cooling plumes can release aerosolised chemicals (biocides, corrosion inhibitors) and legionella bacteria if cooling towers are poorly maintained.

  • Legionnaires’ disease outbreaks have been linked to data centre cooling towers (e.g., 2018 outbreak in Virginia, US, traced to an industrial cooling tower). Nearby residents, especially elderly or immunocompromised, are at risk.

4.5 Psychosocial stress and economic health

Water conflicts generate community stress, legal battles, and perceived injustice. Loss of agricultural livelihoods due to water diversion leads to economic decline and associated mental health burdens.


5. Mitigation and Trade-offs: Is There a Viable Path?

Not all data centres harm water supplies. The outcome depends on technology and location.

StrategyWater impactTrade-offs
Air cooling (direct expansion)Zero on-site water useHigher energy use (+30–50%), thus higher indirect water at power plant, and higher carbon emissions if grid is fossil-fuel-based
Liquid immersion coolingNear-zero water loss (heat rejected via dry coolers)Higher capital cost, limited to high-density compute (AI, HPC)
Recycled/grey water useReduces freshwater withdrawal, but still evaporates waterRequires dual plumbing; blowdown still contains contaminants; availability limited
Siting in water-rich, cool climates (e.g., Nordic countries)Low overall impact, as water is abundant and often hydro-poweredLatency issues for some applications; political/subsidy constraints
Closed-loop chilled water with cooling towers and high cycles of concentrationReduces blowdown volume but increases chemical concentrationsStill consumes water; risk of scaling and corrosion

Key conclusion: The least harmful data centres are those sited in water-abundant, cool climates using air or immersion cooling and powered by non-thermal renewables (solar, wind, hydro without evaporative loss). The most harmful are in arid/semi-arid regions using evaporative cooling with groundwater and fossil-fuel power.


6. Case Study Synthesis: Northern Chile (Data Centre Growth in Atacama)

The Atacama Desert—one of the driest places on Earth—has become a data centre hub due to cheap solar energy and low land costs.

  • Water source: Fossil aquifers (non-renewable).
  • Competition: Local indigenous communities rely on the same aquifers for drinking water and llama/camelid grazing (small-scale agriculture).
  • Observed impacts: Falling water tables have forced some communities to truck in water at 5–10× previous costs; increased arsenic mobilization due to water level drawdown; no regulatory limits on cooling water additives.
  • Health outcomes: Elevated rates of chronic kidney disease of unknown origin (CKDu) in agricultural workers, possibly linked to dehydration and trace metal exposure. No direct causal link to data centres yet, but aquifer chemistry changes are documented.

7. Objective Conclusion and Policy Implications

Balanced statement: Data centres are critical infrastructure, but their water use is not benign. In water-scarce regions, their consumptive cooling can directly reduce residential water access, degrade aquatic ecosystems, and introduce persistent chemicals into water sources, with potential long-term health risks ranging from infectious disease to chronic toxicity.

Strong case for action:

  • Transparency: Mandatory public reporting of water source, consumption (by season), blowdown volume, and chemical additives.
  • Siting regulation: Prohibit evaporative cooling in water-stressed basins (as defined by the Aqueduct Water Risk Atlas or equivalent).
  • Health monitoring: Epidemiological studies near large data centres to track Legionella, emerging contaminants, and waterborne disease rates.
  • Technology mandates: Require air cooling or liquid immersion cooling in all new data centres located in moderate-to-high water stress regions, even if energy costs rise.

Final objective truth: The marginal convenience of evaporative cooling does not justify externalising water scarcity and contamination onto vulnerable populations. The digital economy must decouple from freshwater consumption—just as it is decoupling from fossil fuels. Otherwise, data centres will become yet another infrastructure class that solves global problems for the wealthy while creating local health and ecological crises for the poor.