Thirsty Technology

How the Global Data Center Boom Is Intensifying Water Shortages

---

Executive Summary

The global expansion of artificial intelligence, cloud computing, streaming services, and digital commerce is creating an unprecedented demand for data centers. These facilities depend on thousands of servers that generate heat continuously and must be cooled to prevent equipment failure. Many data centers use large quantities of water directly for cooling and indirectly through the power plants that generate their electricity. This demand is emerging while approximately four billion people experience severe water scarcity during at least one month of the year. The central policy question is therefore not whether societies should use data centers, but how governments can prevent digital growth from undermining water security.

Data centers are not the principal cause of the global water crisis. Agriculture accounts for approximately 72 percent of worldwide water withdrawals, while municipalities and industries account for most of the remainder. Climate change, population growth, pollution, deteriorating infrastructure, and groundwater depletion also contribute substantially to water insecurity. Nevertheless, data centers can intensify shortages because their demand is concentrated in particular cities, watersheds, and utility districts. A facility representing a small share of national water consumption may still become one of the largest industrial users in its local community.

Governments should require facility-level disclosure, independent water-impact assessments, drought contingency plans, and meaningful public participation before approving large data centers. Operators should be encouraged or required to use recycled wastewater, closed-loop cooling, captured rainwater, or other non-potable sources whenever technically practical. Public officials must also evaluate the combined demand of every proposed facility within the same watershed instead of reviewing each project in isolation. Companies receiving tax incentives should pay the infrastructure and environmental costs associated with their operations rather than transferring those costs to households. These protections can support technological development without treating limited freshwater supplies as an unrestricted corporate resource.

I. The Physical Infrastructure Behind the Digital Economy

The popular image of digital technology is misleadingly weightless. Terms such as cloud storage suggest that information exists somewhere above society, disconnected from physical land, energy, and water. In reality, the cloud consists of warehouses filled with servers, networking equipment, backup generators, cooling systems, and electrical infrastructure. Every artificial intelligence request, financial transaction, streamed movie, government database, and social media post depends on this physical system. The expansion of digital services therefore carries environmental consequences that are easily hidden from consumers.

Servers convert much of the electricity they consume into heat. Excessive heat can damage processors, interrupt services, shorten equipment life, and create fire or safety risks. Data centers consequently require cooling systems that operate throughout the day, often regardless of whether the surrounding community is experiencing drought. Some facilities rely heavily on air cooling, while others use cooling towers in which water absorbs heat and evaporates. Evaporative systems can reduce electricity consumption, but they may consume substantial quantities of water that are not immediately returned to the local supply.

The scale of this infrastructure is growing rapidly because artificial intelligence requires especially powerful computing systems. The International Energy Agency projects that global data center electricity consumption will rise from about 485 terawatt-hours in 2025 to approximately 950 terawatt-hours in 2030. Electricity consumption by AI-focused data centers is expected to triple during that period. Efficiency improvements will reduce the energy required for some individual computing tasks, but total consumption may continue rising as companies deploy more processors and consumers use more AI services. This rebound effect demonstrates why greater technical efficiency does not automatically produce lower overall resource consumption. International Energy Agency

II. Understanding the Data Center Water Footprint

The water footprint of a data center begins with the water used at the facility itself. Cooling towers commonly transfer heat from computer equipment into water and release some of that water into the atmosphere through evaporation. Additional water may be needed to control mineral accumulation, clean equipment, maintain humidity, or replace water discharged from the cooling system. Operators may withdraw this water from municipal systems, rivers, reservoirs, aquifers, or reclaimed-water networks. Facilities supplied by municipal utilities may receive treated drinking water even when cooling does not require water of drinking quality.

A proper assessment must distinguish between water withdrawal and water consumption. Withdrawal refers to the total volume taken from a river, aquifer, reservoir, or municipal system. Consumption refers to the portion that evaporates or is otherwise unavailable for immediate reuse within the original watershed. A facility can withdraw a large volume while returning much of it, or it can withdraw less while consuming a high percentage through evaporation. Communities need both measurements because withdrawals affect infrastructure capacity while consumption affects the amount of water remaining for other users.

Data centers also have an indirect water footprint through electricity generation. Coal, natural-gas, and nuclear power plants frequently use water to produce steam or cool generating equipment. A data center using little water on-site may therefore be responsible for significant water consumption elsewhere through its power supply. Semiconductor fabrication, construction, equipment manufacturing, and backup power production add further demands that are not always included in operational statistics. Policymakers should examine the entire water and energy system instead of judging a facility solely by the pipes entering its property.

III. A Growing Demand Within an Existing Global Crisis

Data center expansion is occurring within a much larger crisis of water scarcity and unequal access. Approximately four billion people experience severe water scarcity during at least one month each year, according to UN-Water. The 2026 United Nations World Water Development Report also states that 2.1 billion people still lack safely managed drinking water. Climate change is making rainfall less predictable while intensifying droughts, floods, heat waves, and the loss of mountain glaciers. These pressures make new industrial water demands more consequential than they might have been under historically stable conditions. UN-Water, United Nations World Water Development Report 2026

It would be inaccurate to blame data centers for worldwide water shortages as though they were the dominant global user. Agriculture accounts for approximately 72 percent of water withdrawals, compared with about 16 percent for municipalities and 12 percent for industries. Inefficient irrigation, aging water systems, pollution, population growth, and unsustainable groundwater extraction remain central causes of scarcity. However, global percentages can obscure the impact of a large facility on a small or already stressed watershed. Water policy must therefore consider both the worldwide distribution of demand and the local concentration of consumption.

The United States illustrates the difference between national and local effects. Data centers directly consumed an estimated 17.4 billion gallons of water in 2023, according to research summarized by the Missouri Science and Technology Policy Initiative. Annual direct consumption could increase to between 38 billion and 73 billion gallons by 2028, depending on computing growth and cooling technology. Those totals remain modest compared with national agricultural consumption, but the demand is not distributed evenly across the country. A cluster of facilities can become a major local burden when it depends on the same municipal system, aquifer, or drought-sensitive river. Missouri Science and Technology Policy Initiative

IV. Geography Determines the Severity of the Risk

The water impact of a data center depends heavily on where it is built. A facility using reclaimed water in a cool, water-secure region presents a different risk from an identical facility using potable water in an arid metropolitan area. Temperature, humidity, electricity sources, cooling technology, seasonal demand, and watershed conditions all affect the final footprint. Annual totals can also conceal the problem of peak demand during the hottest months, when cooling requirements and household water use may rise simultaneously. Location should therefore be treated as a central component of data center efficiency rather than a secondary business consideration.

The rapid development of facilities in the American Southwest demonstrates this concern. Arizona, Nevada, and parts of Texas combine growing populations, high temperatures, recurring drought, and expanding industrial demand. The World Resources Institute reported in 2026 that approximately two-thirds of United States data centers built or under development since 2022 were located in water-stressed areas. Communities in these regions must balance residential growth, agricultural production, tribal water rights, ecosystem preservation, and industrial investment. Approving additional facilities without cumulative analysis can commit water that may not remain reliably available throughout their operating lives. World Resources Institute

Similar conflicts are developing outside the United States. Community opposition in Chile pushed Google to reconsider a water-intensive cooling design while the Santiago region faced prolonged drought. European countries seeking investment must weigh their ambitions to become digital hubs against heat and water stress, particularly in southern Europe. India, China, and Southeast Asian countries face the combined pressures of urbanization, groundwater depletion, industrialization, and growing digital demand. Middle Eastern facilities may rely on desalination, which can provide water but requires energy and produces concentrated brine that must be managed responsibly.

V. Economic Benefits and Unequal Community Costs

Data centers provide real economic and social benefits. They support communications, banking, medical research, government services, education, entertainment, and the growing artificial intelligence industry. Construction can generate substantial temporary employment, while operating facilities create technical, maintenance, security, and administrative jobs. Local governments may also receive tax revenue or infrastructure investment from technology companies. A serious policy analysis must recognize these benefits rather than treating every data center as inherently harmful.

The distribution of those benefits, however, may not match the distribution of the costs. Data centers can receive tax abatements, discounted land, expedited permits, or specially negotiated utility arrangements in exchange for locating within a jurisdiction. Municipalities may then need to expand water-treatment plants, electrical substations, transmission lines, pipelines, wells, or wastewater systems. If contracts do not assign these expenses clearly, households and small businesses may help finance infrastructure built primarily for corporate customers. The public may also bear long-term environmental risks after short-term construction activity has ended.

The employment argument deserves particular scrutiny because modern data centers are highly automated. A multibillion-dollar campus may create thousands of construction jobs but employ a comparatively small permanent workforce once operations begin. Governments should therefore calculate the number, quality, duration, and local availability of jobs before awarding public subsidies. They should also compare data center development with alternative uses of the same land, power, and water. Economic development becomes difficult to justify when the public assumes the costs while corporate owners retain most of the gains.

VI. Transparency Is the First Policy Requirement

Effective water governance is impossible without reliable information. Some companies disclose company-wide water consumption while providing limited information about individual facilities. Global totals do not tell residents how much water is being withdrawn from their aquifer or municipal system. Companies may also use different definitions, reporting periods, and accounting boundaries, making direct comparisons difficult. Mandatory and standardized facility-level disclosure would allow governments and communities to evaluate the actual risks.

Every large data center should report annual and seasonal water withdrawals, consumption, discharge, sources, cooling methods, and Water Usage Effectiveness. Water Usage Effectiveness measures the amount of water used relative to the electricity consumed by computing equipment. This measurement is useful, but it should not be treated as a complete sustainability rating. A highly efficient facility can still consume enormous volumes if it operates at a very large scale. Reports should therefore provide both efficiency ratios and absolute volumes.

Corporate water-replenishment commitments also require closer examination. A company may fund wetlands, leak repairs, conservation programs, or watershed restoration in an effort to replace the water associated with its operations. These projects can produce genuine benefits, but replenishing water in one place does not necessarily repair the shortage created somewhere else. Timing also matters because water restored during a wet season may not compensate for heavy consumption during a summer drought. Governments should require replenishment within the affected watershed whenever possible and verify claims through independent audits.

VII. Governing the Water and Energy Tradeoff

Reducing direct water use can sometimes increase electricity consumption. Dry cooling and conventional air conditioning may conserve water but require more power, particularly during extreme heat. If that electricity comes from water-intensive or carbon-intensive generation, the facility may shift rather than eliminate its environmental burden. Conversely, evaporative cooling can reduce electricity use while consuming more local water. Policymakers must evaluate both resources together instead of maximizing one efficiency measurement at the expense of the other.

The appropriate solution will differ by climate and region. Facilities in cool environments may rely more heavily on outside air, while those in arid regions may use closed-loop liquid cooling or reclaimed wastewater. Immersion systems place computing equipment in specialized fluids that absorb heat efficiently and can reduce reliance on evaporative cooling. Advanced controls can also adjust cooling according to temperature, humidity, workload, and electricity conditions. No single technology will solve the problem in every location, but governments can require operators to demonstrate that they selected the least harmful practical system.

Electricity procurement is equally important. Wind and solar photovoltaic generation generally have lower operational water requirements than thermal power plants, although manufacturing and supply chains still carry environmental costs. Data centers can also schedule flexible computing tasks for periods when renewable electricity and water are more available. Waste heat may be captured for district heating or nearby industrial uses in suitable climates. Integrated planning can turn water, energy, and heat from separate problems into a coordinated infrastructure strategy.

VIII. A Policy Framework for Water-Secure Digital Development

The first policy priority should be water-sensitive siting. Governments should require independent assessments of present supply, future population growth, climate conditions, drought probability, agricultural demand, ecosystem requirements, and tribal or Indigenous water rights. Applications should disclose expected water consumption under normal operations, heat waves, equipment failures, and maximum computing capacity. Facilities using potable water should face heightened scrutiny in regions classified as highly or extremely water-stressed. Projects that cannot demonstrate a reliable and equitable supply should be redesigned, relocated, or rejected.

The second priority should be establishing a clear hierarchy of water use. Drinking, sanitation, public health, emergency response, and basic ecosystem protection should take precedence over discretionary industrial consumption. Data centers should use treated wastewater, captured stormwater, recycled cooling water, or other non-potable sources whenever those alternatives are safe and practical. Operators should also maintain drought plans that automatically reduce water use when reservoirs, rivers, or aquifers fall below established thresholds. These rules would provide certainty to companies while protecting communities from emergency competition.

The third priority should be assigning costs to the parties creating them. Data center operators should pay for the additional water, wastewater, and electrical infrastructure their facilities require. Tax incentives should be conditioned on independently verified employment, efficiency, transparency, and community benefits. Residential customers should not face higher rates because utilities underestimated corporate demand or negotiated inadequate service agreements. Public officials should publish contracts, impact studies, tax concessions, and water allocations before granting final approval.

IX. From Individual Permits to Watershed Governance

The traditional permitting process often evaluates one proposed facility at a time. This approach can underestimate the combined burden created when several companies build within the same region. Ten individually acceptable projects may collectively exceed the sustainable capacity of a municipal system or aquifer. Governments should therefore establish watershed-level budgets that account for every existing and proposed user. Those budgets must be updated as climate conditions, population, and industrial demand change.

Regional coordination is particularly important when water crosses political boundaries. Rivers and aquifers frequently serve multiple cities, states, provinces, tribal nations, or countries. One jurisdiction may approve a facility and receive its tax revenue while communities downstream experience reduced supply or environmental damage. Shared water commissions should have the authority to review large industrial projects with cross-border effects. Data center development should not become a method for one government to capture economic benefits while exporting resource costs to its neighbors.

Public participation must also occur before decisions become effectively irreversible. Residents should receive understandable information about anticipated water use, infrastructure costs, employment, tax incentives, and drought risks. Hearings should provide meaningful opportunities to change or reject proposals rather than merely satisfy procedural requirements. Indigenous communities and other groups with established water rights must be included as governing participants, not consulted only after plans have been developed. Democratic oversight is essential because water allocation determines which communities and economic activities will be protected during future shortages.

X. Conclusion: Deciding What Water Is For

The rise of data centers does not create the global water crisis by itself. Agricultural demand, climate change, population growth, pollution, weak infrastructure, and unsustainable extraction remain larger drivers in many regions. Data centers nevertheless represent one of the fastest-growing sources of concentrated industrial demand. Their expansion is occurring precisely when many governments are struggling to guarantee water for households, farms, and ecosystems. Ignoring this additional pressure would be both economically shortsighted and environmentally irresponsible.

The proper response is not to reject artificial intelligence or dismantle the digital economy. Governments should instead require technology companies to build in appropriate locations, disclose their water footprints, use the safest available sources, and finance the infrastructure they need. Companies must also evaluate absolute consumption rather than relying exclusively on efficiency improvements. A facility that uses less water per computation may still increase total demand if its workload expands rapidly. Sustainable digital development requires firm public standards rather than voluntary promises alone.

Water policy ultimately reflects political priorities. A community deciding whether to approve a data center is also deciding how much water should remain available for residents, food production, economic development, and the natural environment. Those choices should be made openly, with reliable evidence and enforceable protections. Artificial intelligence may operate in the cloud, but the infrastructure supporting it remains firmly rooted in physical communities. Every digital service has a material cost, and the water required to sustain it must come from somewhere on Earth.

References

International Energy Agency. (2026). Key questions on energy and AI. https://www.iea.org/reports/key-questions-on-energy-and-ai

Lakhanpal, V. (2026, April 8). Data center water use. MOST Policy Initiative. https://mostpolicyinitiative.org/science-note/data-center-water-use/

UN-Water. (n.d.). Water scarcity. Retrieved August 26, 2026, from https://www.unwater.org/water-facts/water-scarcity

United Nations Educational, Scientific and Cultural Organization. (2026). The United Nations world water development report 2026: Water for all people: Equal rights and opportunities. https://unesdoc.unesco.org/ark:/48223/pf0000397159

Walker, C. D., & Goldsmith, I. (2026, February 17). From energy use to air quality, the many ways data centers affect US communities. World Resources Institute. https://www.wri.org/insights/us-data-center-growth-impacts