DOSSIER

TECH

The invisible cost of artificial intelligence for water resources

Artificial intelligence has become part of the daily routine for millions of people. It writes texts, creates images, answers questions, and helps companies automate tasks at an impressive speed. But behind the answers generated in seconds lies a gigantic infrastructure that consumes enormous amounts of energy and natural resources. Among them is a fundamental element for human life and increasingly scarce in various regions of the planet: water.

When a person asks a chatbot a question or requests the creation of a text, the answer seems to appear instantly. However, this process depends on data centers full of servers operating continuously.

These devices generate heat on a large scale. To avoid overheating that could compromise the systems, companies use cooling structures that often depend on large volumes of water.

Researchers at the University of California, Riverside, estimated that producing a simple text of about 100 words can consume approximately 519 milliliters of water when considering both direct cooling costs and indirect consumption associated with electricity generation.

Although the value seems small in isolation, the situation changes when billions of requests are processed daily worldwide.

The challenge lies precisely in the scale. Each individual interaction represents a tiny fraction of the total consumption, but the sum of millions of users simultaneously utilizing artificial intelligence systems creates a growing demand for water resources.

The processing centers that support modern artificial intelligence operate with thousands of specialized chips working without interruption.

These components can dissipate hundreds of watts of energy each. In large-scale advanced model training operations, tens of thousands of these processors work simultaneously for weeks or months.

All this heat needs to be removed. One of the most widely used methods is evaporative cooling, in which water absorbs the heat produced by the equipment and some of it evaporates into the atmosphere.

According to experts, a significant portion of the water used in these systems is permanently lost during the evaporation process.

The new generation of data centers specifically designed for artificial intelligence also presents characteristics that amplify the problem. These facilities are larger, concentrate more equipment, and operate with thermal densities far superior to those observed in traditional cloud computing structures.

In some cases, the daily water consumption of a single technology complex can rival that of small cities.

The numbers that are worrying researchers...Environmental reports released by large technology companies show a clear trend of increasing water consumption.

In recent years, companies such as Google, Microsoft, and Meta have recorded significant increases in water use in their operations.

Researchers estimate that the global water demand associated with artificial intelligence could reach between 4.2 and 6.6 billion cubic meters per year by 2027.

The numbers are impressive because they are equivalent to the annual water consumption of entire countries. In some projected scenarios, the global AI infrastructure could use a volume close to half of the entire annual water withdrawal of the United Kingdom.

In addition to direct consumption for cooling, there is an even larger component that is often ignored: the water used in generating the electricity that powers the data centers.

Studies indicate that this indirect consumption can exceed several times the amount used directly in cooling systems.

The impact is already appearing in drought-stricken regions...One of the biggest concerns of experts is the location of many of these technological centers.

Several projects are being built in areas that already face water scarcity or prolonged periods of drought. Recent cases involving planned or operational facilities in Chile, Mexico, Uruguay, Spain, and southwestern states of the United States have broadened the debate about the use of local resources.

In some of these regions, communities already live with restrictions on water supply while large technological projects seek to guarantee access to aquifers and municipal systems.

Researchers warn that the accelerated expansion of artificial intelligence is happening at a time when global water scarcity is also growing rapidly. International projections indicate that a significant portion of the world's population may face severe water stress by the end of the decade.

By 2030, AI's water use will match the needs of 1.3 billion people while its power use triples that of 650 million, UN University investigation warns...By 2030, the global data centres powering artificial intelligence are projected to consume 945 terawatt-hours of electricity. This is nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria—countries collectively home to more than 650 million people. Their associated water footprint will equal the basic annual domestic water needs of all 1.3 billion people in Sub-Saharan Africa, and their land footprint will exceed 14,500 square kilometers, roughly twice the Jakarta metropolitan area, home to more than 32 million people. 

These stark findings are detailed in the new report, Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints, by the United Nations University Institute for Water, Environment and Health (UNU-INWEH). Researchers have previously warned about the greenhouse gas emissions of data centers before. But the UN scientists now argue that the environmental costs of AI and data centers cannot be understood through carbon emissions alone. In their report, they quantify the carbon, water and land footprints of AI's electricity use across the globe and highlight the big differences between these footprints in the world’s 20 largest data center hubs.

"This report is not a case against artificial intelligence, a technological transformation that is improving the lives of billions of people around the world," said Professor Kaveh Madani, Director of UNU-INWEH who led the investigation team. "It is a call for using it responsibly and addressing its unintended impacts proactively to make it sustainable and equitable. We have a narrow window to ensure that the backbone of the technological revolution of our era develops within planetary limits, and that the communities who provide the critical minerals for advancing AI and the ones that host its infrastructure and e-waste are also among those who benefit from it." 

The true environmental cost of AI...An environmental footprint that is being measured incorrectly. The report finds that the environmental cost of AI is being systematically mismeasured. Most existing assessments focus on the carbon emissions associated with training large models. However, every kilowatt-hour of electricity used to train or run an AI system also carries a water footprint, stemming from cooling and power generation, and a land-use footprint, stemming from energy infrastructure and supply chains. These three footprints do not move in the same direction. The transition from coal to bioenergy, for example, can, on average, reduce the carbon footprint of electricity by 70%, while increasing its water footprint by more than thirty times and its land-use footprint by one hundred times. The report concludes that "low carbon" does not automatically mean "low water consumption" or "low land use" and warns that assessing the sustainability of AI through a single metric may obscure trade-offs and shift environmental burdens to regions already facing water or land-use stress. The numbers are rapidly adding up at the infrastructure level. Global data centers will consume approximately 448 terawatt-hours of electricity in 2025. If considered as a single country, they would be the 11th largest electricity consumer in the world, behind France and ahead of Saudi Arabia.

"What surprised us most was how often the choices that seem most environmentally friendly from a carbon emissions standpoint end up being worse for water or soil," said Dr. Miriam Aczel, a researcher at UNU-INWEH and lead author of the report. "If we continue to assess the sustainability of AI solely by carbon emissions, we might think that renewable energy makes AI infrastructure clean, but this solves one problem while creating others, often in locations that didn't ask for them."

Inference, efficiency, and the rebound effect...Public discussion has largely focused on the energy required to train massive models. Training GPT-3 was estimated to require 1.3 gigawatt-hours (GWh) of electricity, while estimates suggest GPT-4 consumed between 50 and 70 GWh. However, the report reveals this framing is outdated. Once a model is deployed, inference—the continuous running of models to answer everyday user prompts—becomes the dominant cost, accounting for 80 to 90 per cent of total AI energy use. ChatGPT alone is estimated to process around 2.5 billion prompts per day, translating to roughly 383 GWh of electricity per year for a single product. Offsetting associated carbon emissions would require 2.6 million tree seedlings grown for 10 years, enough trees to cover a land area the size of Manhattan. The water footprint is equivalent to the minimum annual domestic water needs of roughly 500,000 people in Sub-Saharan Africa, and the land footprint is equal to over 800 football fields.

Marking its 30th anniversary of operation in 2026, the United Nations University Institute for Water, Environment and Health (UNU-INWEH) is one of 13 institutions that make up the United Nations University (UNU), the academic arm of the UN. Known as 'The UN's Think Tank on Water', UNU-INWEH addresses critical water, environmental, and health challenges around the world. Through research, training, capacity development, and knowledge dissemination, the institute contributes to solving pressing global sustainability and human security issues of concern to the UN and its Member States. Headquartered in Richmond Hill, Ontario, UNU-INWEH has been hosted and supported by the Government of Canada since 1996. With a global mandate and extensive partnerships across UN entities, international organizations, and governments, UNU-INWEH operates through its UNU Hubs in Calgary, Hamburg, New York, Lund, and Pretoria, and an international network of affiliates.

mundophone