AI consumes water and energy, generates heat and noise and leaves an environmental cost for society.
By: Ramesh Raja
We use Artificial Intelligence almost casually now. We ask AI to write, translate, calculate, research, design, analyze images, prepare reports and answer questions. The technology appears on our mobile phones and computers as if it exists in a weightless digital world. But behind every AI response is a very physical infrastructure of enormous computers, electricity, buildings, cooling systems, water and power networks. That infrastructure is the data center. As AI expands, data centers are expanding with it, creating an important question that deserves much greater public attention: how much water, energy and environmental space are we willing to consume for our digital future?
AI Needs Data Centers
A data center is not simply a room full of computers. It is a highly engineered industrial facility containing thousands of servers, high-performance GPUs and other processors, storage equipment, networking systems, batteries, power-distribution equipment, cooling machinery, fire-protection systems and backup generators. AI is particularly demanding because advanced AI models require enormous computing power, with powerful processors operating continuously and generating considerable heat. This creates a simple chain: AI needs computing power, computing power needs data centers, data centers need electricity, electricity used by computers ultimately becomes heat, heat has to be removed, and some cooling systems require substantial quantities of water. Cooling equipment, transformers and backup generators can also create noise and, depending on the power source and operating conditions, air emissions. The International Energy Agency estimates that data centers consumed about 415 terawatt-hours of electricity worldwide in 2024, equivalent to around 1.5 percent of global electricity consumption, and projects that their electricity demand could more than double to about 945 TWh by 2030, with AI as an important driver. The AI revolution, therefore, is not only a digital revolution. It is rapidly becoming an energy, water, land-use and environmental issue.
Why Does a Data Centre Need Water?
The computers themselves do not drink water. Water is required mainly because the enormous heat produced by computing equipment has to be removed continuously. In a conventional evaporative cooling arrangement, heat generated by IT equipment is transferred to a chilled-water system and then to a condenser-water loop, which carries the heat to a cooling tower. The cooling tower uses evaporation to reject that heat into the atmosphere. The U.S. Department of Energy explains that this process consumes water through evaporation and also loses water through blowdown, drift and other processes. It uses the metric Water Usage Effectiveness, or WUE, measured in liters of water consumed per kilowatt-hour of IT energy use, to assess data-center water efficiency. This is why the statement that “AI needs water” needs some qualification: not every data center consumes large quantities of freshwater, and newer facilities can use dry cooling, closed-loop systems and advanced liquid cooling to reduce water consumption. Nevertheless, where evaporative cooling is used, water consumption can become substantial, particularly when a large facility is located in a water-stressed community.
How Much Water Can They Consume?
The real-world figures demonstrate why this issue cannot be dismissed as an exaggerated environmental concern. One of the most important long-running case studies is The Dalles, Oregon, where Google’s data-center water use has been documented from 2012 onward. A 2026 study published in PLOS Water reports that Google’s water demand in The Dalles increased by approximately 342 percent between 2012 and 2024. In 2024 alone, Google’s local data-center water demand reached approximately 461.1 million gallons for the year, equivalent to about 1.26 million gallons, or approximately 4.77 million liters, every day. That represented around 30 percent of the city’s average daily water demand of 4.2 million gallons. The study also projects that The Dalles could face a potential water deficit of about 90 million gallons by 2034 under its cited demand projections. This does not mean that every AI data center consumes this amount of water, because cooling technology, climate, computing load and facility size vary greatly. But the case is important because it demonstrates that one major technology operation can become a significant component of a small city’s total water demand. A separate 2026 study estimated that if current water-use intensity continues, U.S. data centers could collectively require 697 to 1,451 million gallons per day of additional water capacity by 2030, with the impacts highly concentrated in communities that host data centers.
The Dalles: A Major Reference in the Water Debate
The Dalles case is particularly valuable because it is not based simply on a prediction or social-media allegation. The city’s water-use data became available through a public-disclosure agreement following a legal dispute over access to the information. The 2026 PLOS Water study used annual data covering 2012 to 2024 and showed how Google’s water demand had grown over time. The case has consequently become one of the most useful references in discussions about data-center water consumption because it connects a real facility, a real municipal water system and publicly documented water demand over more than a decade. It also raises a broader principle: communities cannot properly assess the environmental consequences of data centers if they do not know how much water those facilities are consuming. Transparency about water withdrawals, consumption and peak demand should therefore be regarded as an essential part of environmental governance.
The Heat Does Not Disappear
Water is only one side of the problem. The enormous quantity of electricity entering a data center ultimately becomes heat, and cooling systems remove that heat from the servers and reject it into the surrounding environment. This does not mean that every data centre automatically creates a measurable city-wide heat island, and such claims should be made cautiously. However, very large facilities concentrated in one area can become significant sources of local waste heat and energy demand. The International Energy Agency reports that cooling and environmental-control systems account for approximately 7 percent of electricity consumption in efficient hyperscale data centers and more than 30 percent in less-efficient facilities. A data center is therefore effectively a giant computing and heat-management machine operating continuously. It consumes enormous quantities of electricity to perform invisible digital work and then continuously transfers the resulting heat to the environment. The public sees an AI answer on a screen, but behind that answer are machines consuming electricity and systems working around the clock to remove the heat they generate.
Then Comes the Noise
There is another environmental impact that communities can experience directly: noise pollution. Large data centers contain powerful fans, cooling towers, chillers, pumps, transformers, ventilation equipment and sometimes gas turbines or diesel backup generators. Some of these systems operate continuously because the servers cannot simply be switched off at night. The result can be a persistent mechanical hum rather than an occasional industrial disturbance. One of the most visible current examples is Loudoun County, Virginia, widely known as “Data Center Alley”. In March 2026, NBC4 Washington reported residents complaining about a continuous droning sound from a CloudHQ data center in the Ashburn area; its reporters recorded approximately 90 decibels using an app, although such an app reading should not be treated as an official regulatory measurement. The complaints have continued. In July 2026, residents near a Vantage data-center facility in Sterling complained after backup generators operated for almost 24 hours during a power failure. One resident reported a reading of 60 decibels at his property line, compared with Loudoun County’s cited nighttime limit of 55 decibels, while the county’s environmental authorities were drawn into the dispute. These incidents demonstrate why noise should be considered an environmental and community-planning issue rather than merely an engineering detail.
Where can the Public Read the Complaints?
The Loudoun experience is useful because the complaints are not confined to informal social-media discussions. Residents can submit complaints through Loudoun County’s official complaint and service-request system, while county zoning officials conduct inspections and noise measurements where appropriate. The ongoing Sterling controversy has also been documented by local journalism, including reports describing continuous noise, generator operation and residents’ concerns. This makes Loudoun an important reference for other jurisdictions: the public needs not only noise limits on paper but also a transparent mechanism through which residents can report problems, obtain responses and see whether operators are complying with the regulations. A regulatory system that measures noise only when nobody is complaining is inadequate; continuous industrial noise requires continuous accountability.
What about Water Contamination?
The question of contamination also requires technical precision. A data center does not automatically contaminate groundwater or drinking water simply because it uses water for cooling. Cooling systems are generally designed as controlled systems, and wastewater can be treated. However, cooling-tower water can become concentrated with dissolved minerals and may contain treatment chemicals, while the blow down discharged from evaporative systems therefore requires appropriate management. The U.S. Department of Energy identifies corrosion, scaling, fouling and microbiological activity as important concerns in open-recirculating cooling systems. Consequently, an environmental assessment should not merely state how much water a data center will withdraw. It should explain where the water comes from, how much is actually consumed, how much is discharged, what the discharge contains, how it will be treated and where it will ultimately go. Water quantity and water quality are separate environmental questions, and both deserve public scrutiny.
The Public Is Beginning to Cry Out
The controversy is now spreading beyond the United States. In Visakhapatnam, Andhra Pradesh, India, Google’s proposed $15 billion data-center project has become the subject of environmental opposition over water and wildlife concerns. Reuters reported on August 6, 2026 that the project was facing protests and legal challenges, while the area was already experiencing a reported daily water shortfall of about 70 million liters. Activists have raised concerns about pressure on local water resources and the nearby Kambalakonda Wildlife Sanctuary. Google says it intends to use advanced air cooling and sound-dampening technologies, while the government and project supporters have disputed some of the allegations. The Andhra Pradesh High Court is scheduled to hear the matter on August 24. This is particularly relevant to Pakistan because it demonstrates that the AI infrastructure debate is no longer confined to wealthy Western countries. It is reaching South Asia, where water scarcity, urban pressure and environmental vulnerability make the choice of cooling technology and project location particularly important.
Pakistan Must Learn Before It Builds
Pakistan cannot afford to approach AI infrastructure merely as a race to attract investment. Water scarcity, unreliable electricity, urban congestion and environmental degradation are already serious challenges. If large AI data centers are established in Pakistan, their environmental footprint should be assessed before approval through comprehensive Environmental Impact Assessment. Environmental authorities should require separate assessment of water consumption, peak water demand, wastewater, noise, electricity requirements, backup-generator emissions, heat rejection and electronic waste. Operators should disclose their Water Usage Effectiveness and energy performance, while regulators should examine not only average water use but also peak demand during the hottest periods when both cooling requirements and public water demand can rise. The U.S. Department of Energy itself notes that data-center water demand is directly related to IT heat load and cooling-system efficiency and recommends strategies for reducing water consumption, including improved cooling management and recovery of cooling-tower blowdown.
We Can Reduce the Water Footprint
The encouraging fact is that technology can reduce much of this environmental burden. Data centers can use closed-loop liquid cooling, direct-to-chip cooling, air cooling, dry cooling and hybrid systems, while treated wastewater, recycled water and other non-potable sources can be considered where technically and environmentally appropriate. The choice involves trade-offs because some dry systems can require more electricity, while evaporative cooling can reduce energy consumption at the cost of water. The right solution therefore depends on the local climate, water availability, electricity system and technical design. But in a water-stressed country such as Pakistan, the principle should be clear: high-quality drinking water should not automatically become the default industrial cooling resource when technically suitable recycled or non-potable alternatives are available. Environmental assessment should require developers to demonstrate why their selected cooling technology is appropriate for the local water situation.
How Can Noise and Waste Heat Be Controlled?
Noise should be assessed before construction rather than after residents begin complaining. Data centers should incorporate acoustic barriers, low-noise cooling equipment, vibration control and strict boundary noise limits, while continuous monitoring should be undertaken after commissioning, particularly during nighttime operation. Emergency generators and gas turbines should also be included in the assessment because their noise and air emissions can be considerably greater than normal equipment operation. Waste heat deserves similar attention. Instead of simply rejecting heat into the environment, future facilities should examine opportunities for heat recovery and reuse where local conditions make this practical. Depending on location and application, recovered heat can potentially support district heating, industrial processes, agriculture or other uses. Even where direct reuse is not feasible, facilities should select heat-rejection systems that minimize local environmental impacts. Heat should therefore be treated either as a resource to be recovered or as an environmental liability that must be properly managed.
How Can We, the Public, Play Our Role?
The public also has an important role in shaping responsible AI infrastructure. We should not reject AI, nor should we accept every AI project simply because it carries the label of technological progress. Whenever a major data-center project is proposed, citizens can ask local authorities, municipal bodies and environmental agencies how much water the facility will consume, whether it will use drinking water or recycled water, what cooling technology will be installed, how much electricity it will require, where that electricity will come from, where wastewater will be discharged, what substances it will contain, how much noise will reach neighboring communities and what measures will control heat, emissions and generator noise. Environmental Impact Assessment documents should be made publicly accessible, and citizens should be able to submit objections and receive documented responses. Universities can independently research the water and energy footprint of AI infrastructure in Pakistan, professional bodies can develop technical guidelines, engineers can design water-efficient cooling and wastewater-reuse systems, environmental organizations can monitor compliance, journalists can investigate public data, and communities can participate in planning and environmental hearings. The lesson from The Dalles and Loudoun County is that public access to information and a functioning complaint mechanism are not obstacles to development; they are safeguards for responsible development.
We Should Not Stop AI. We Should Make AI Responsible
AI offers enormous opportunities for Pakistan. It can improve education, healthcare, engineering, agriculture, transport, disaster management, scientific research and government services. But technological progress should not mean transferring hidden environmental costs to the public. The world has already learned this lesson from factories, power plants, automobiles and other industrial technologies, where economic development without environmental responsibility eventually produced costs that society had to bear. AI should not repeat that mistake. The future should be based on water-efficient AI, energy-efficient data centers, renewable power, recycled water, closed-loop cooling, responsible wastewater management, strict noise controls, transparent environmental reporting, heat management and meaningful community participation. The public is ready to use AI, but governments, technology companies and environmental agencies must ensure that the water, air, peace and environmental quality of communities are not sacrificed to power it. The examples of The Dalles, Loudoun County and Visakhapatnam show that the public is already asking questions, filing complaints, demanding information and challenging projects. Pakistan should learn from these experiences before major AI infrastructure begins placing new pressure on our already limited water and environmental resources. The question is no longer whether AI will shape our future; the real question is what kind of future will we allow AI to build?
Read: Pakistan Railways: Put Back on Track
________________
The author of this article, Engr. Ramesh Raja, is a Civil Engineer, visionary planner, PMP certified and literary enthusiast with a passion for art and recreation. He can be reached at engineer.raja@gmail.com
