UN Report Warns AI Could Consume 3% of World's Electricity and Deplete Drinking Water by 2030

UN report warns AI could soon use 3% of world's electricity and more water than we need to drink

One argument often used to quell concerns about the rising energy and resource demand of data centres is that artificial intelligence (AI) models will need less in the future as they improve and become more efficient. But this seemingly logical thinking is a trap, according to a new United Nations report that quantifies the environmental costs of AI. The report estimates that by 2030, AI's energy use could double to consume 3% of the world's electricity, produce emissions to equal the UK and deplete more water for cooling than the annual drinking water need of the global population. It also anticipates the use of AI will follow an economic principle known as the "Jevons paradox", which predicts that when technological improvements increase the efficiency of a resource, it leads to a rise, rather than a fall, in the total consumption of that resource. The paradox is named after economist William Stanley Jevons who observed this effect with the use of coal in 19th-century England. Efficiency gains did not reduce overall consumption. Instead, the lower costs resulted in expanded use and higher overall demand. As AI models become cheaper and more attractive, the report expects this to encourage new uses and higher volumes of use, eroding and possibly erasing any savings from efficiency advances. To avoid falling into this trap, it lays out a roadmap for responsible AI use based on guiding principles of transparency, efficiency by design, equity and justice, lifecycle responsibility, global cooperation and sustainable use. The scale of the problem Last year, data centres already consumed as much electricity as Saudia Arabia, which ranks as the world's 11th largest electricity consumer. If electricity use doubles as projected by 2030, the associated carbon footprint would require 6.7 billion trees grown over ten years to offset this demand. Data centres would also require 9.3 trillion litres of water and land nearly ten times the size of Mexico City. Beyond resource use, the report also underscores the structural inequity at the heart of the AI boom, with only 32 nations hosting AI-specific cloud infrastructure and 90% of that capacity located in the US and China. It warns of a widening digital divide between nations that build and control AI systems and those that consume them, with the latter often bearing a disproportionate environmental burden caused by mineral extraction and e-waste. Responsible AI use Two main forces shape AI's operational footprint: how much we use it and how we use it. This involves all tasks AI models perform, from text and code generation to image and video. Each of these tasks requires different levels of computational effort. The model choice also matters as each AI system performs these task with distinct energy and environmental costs. The report argues responsible AI requires full value-chain governance, from mineral sourcing to recycling and safe disposal. It calls for a twinning of capability and environmental stewardship - thinking about both what AI can do for us and the protection of the natural environment. This would mean making environmental disclosures a routine part of AI development, at both the model and task level, and incorporating projected AI demand in climate and energy planning. Responsible AI is crucial as countries are promoting and adopting AI across government and the public sector. In Aotearoa New Zealand, the government has launched a national AI strategy and a public service AI framework. While the framework was informed by the OECD's values-based AI principles, including inclusive and sustainable development, there is no requirement for environmental disclosures and no regulator compiling energy use or emissions. Likewise in Australia, improving public services is part of the national AI plan. For example, the National Film and Sound Archive of Australia has created Bowerbird, a machine learning-enabled mass audio and video transcription engine, to document material. The Department of Veteran's Affairs has developed a proof-of-concept tool to see whether AI can help speed up the processing of claims. Both countries take a deliberate "light touch" and principles-based regulatory approach to AI. But this approach risks overlooking the growing environmental cost of AI that can't be solved by improving it. The natural environment is foundational to the economy, culture and wellbeing. It should be at the centre of our thinking. It's time to rethink the AI innovation playbook and shift focus toward a sustainable tech future.

AI to double data centre power and water consumption by 2030, UN researchers say

SINGAPORE, June 3 (Reuters) - Data centres are expected to consume twice as much power and water by 2030 as they expand to meet the surge in demand from artificial intelligence, U.N. researchers said on Wednesday. Unless governments heed the rising environmental costs of AI, the rapid rollout could also strain scarce land resources and create mountains of electronic waste, the United Nations University Institute for Water, Environment and Health warned in a report. Here are a few takeaways: Reporting by David Stanway; Editing by Sherry Jacob-Phillips Our Standards: The Thomson Reuters Trust Principles., opens new tab

AI could consume up 3% of world's electricity the UN warns

AI could soon use more water than we need to drink, UN report finds. One argument often used to quell concerns about the rising energy and resource demand of data centers is that artificial intelligence (AI) models will need less in the future as they improve and become more efficient. But this seemingly logical thinking is a trap, according to a new United Nations report that quantifies the environmental costs of AI. The report estimates that by 2030, AI's energy use could double to consume 3% of the world's electricity, produce emissions to equal the UK and deplete more water for cooling than the annual drinking water need of the global population. It also anticipates the use of AI will follow an economic principle known as the "Jevons paradox", which predicts that when technological improvements increase the efficiency of a resource, it leads to a rise, rather than a fall, in the total consumption of that resource. The paradox is named after economist William Stanley Jevons who observed this effect with the use of coal in 19th-century England. Efficiency gains did not reduce overall consumption. Instead, the lower costs resulted in expanded use and higher overall demand. As AI models become cheaper and more attractive, the report expects this to encourage new uses and higher volumes of use, eroding and possibly erasing any savings from efficiency advances. To avoid falling into this trap, it lays out a roadmap for responsible AI use based on guiding principles of transparency, efficiency by design, equity and justice, lifecycle responsibility, global cooperation and sustainable use. The scale of the problem Last year, data centers already consumed as much electricity as Saudi Arabia, which ranks as the world's 11th largest electricity consumer. If electricity use doubles as projected by 2030, the associated carbon footprint would require 6.7 billion trees grown over ten years to offset this demand. Data centers would also require 9.3 trillion liters of water and land nearly ten times the size of Mexico City. Beyond resource use, the report also underscores the structural inequity at the heart of the AI boom, with only 32 nations hosting AI-specific cloud infrastructure and 90% of that capacity located in the US and China. It warns of a widening digital divide between nations that build and control AI systems and those that consume them, with the latter often bearing a disproportionate environmental burden caused by mineral extraction and e-waste. Responsible AI use Two main forces shape AI's operational footprint: how much we use it and how we use it. This involves all tasks AI models perform, from text and code generation to image and video. Each of these tasks requires different levels of computational effort. The model choice also matters as each AI system performs these task with distinct energy and environmental costs. The report argues responsible AI requires full value-chain governance, from mineral sourcing to recycling and safe disposal. It calls for a twinning of capability and environmental stewardship -- thinking about both what AI can do for us and the protection of the natural environment. This would mean making environmental disclosures a routine part of AI development, at both the model and task level, and incorporating projected AI demand in climate and energy planning. Responsible AI is crucial as countries are promoting and adopting AI across government and the public sector. In Aotearoa New Zealand, the government has launched a national AI strategy and a public service AI framework. While the framework was informed by the OECD's values-based AI principles, including inclusive and sustainable development, there is no requirement for environmental disclosures and no regulator compiling energy use or emissions. Likewise in Australia, improving public services is part of the national AI plan. For example, the National Film and Sound Archive of Australia has created Bowerbird, a machine learning-enabled mass audio and video transcription engine, to document material. The Department of Veteran's Affairs has developed a proof-of-concept tool to see whether AI can help speed up the processing of claims. Both countries take a deliberate "light touch" and principles-based regulatory approach to AI. But this approach risks overlooking the growing environmental cost of AI that can't be solved by improving it. The natural environment is foundational to the economy, culture and wellbeing. It should be at the center of our thinking. It's time to rethink the AI innovation playbook and shift focus toward a sustainable tech future. This edited article is republished from The Conversation under a Creative Commons license. Read the original article.

UN calculates nation-sized environmental footprints for AI and data centers

WASHINGTON (AP) -- The environmental footprint of data centers already rivals some of the world's largest countries, according to a United Nations University report, which also predicts their water and energy use and pollution will double in just four years as use of artificial intelligence grows. Last year, global data centers used 448 trillion watt-hours of electricity, more than all but 10 countries of the world, said the report issued Wednesday. That electricity use produced about 208 million tons (189 million metric tons) of carbon dioxide, about the same amount as Argentina, and producing that much energy consumed about 1.2 trillion gallons (4.5 trillion liters) of water, according to the report on the environmental consequences of AI's energy use. By 2030, data centers will account for nearly 3% of the world's projected electricity use, with 935 trillion watt-hours. If data centers were a country, the country would be projected to rank sixth-highest in power use in 2030. That would produce nearly 440 million tons (399 million metric tons) of carbon dioxide, the report said. The study focused on energy use and didn't examine the massive amount of water used to cool data centers. "If you look at these numbers, we're seeing scales comparable to nations," said study co-author Kaveh Madani, a water scientist and director of the United Nations University Institute for Water, Environment and Health in Canada. "The demand is enormous." Much of the growth of data centers is being driven by AI. About 20% of data centers' energy is currently due to AI, but that should grow to 40% by 2030, the report said. First global look at ecological impact The report is significant because of the credibility and authority of the U.N., not just because of any one set of eye-popping numbers, said Fengqi You, a Cornell University energy engineering professor who directs the college's AI sustainability issues. "Its value is that a U.N. institution is putting carbon, water, land, life-cycle impacts and environmental justice into one frame" for an issue that is often shrouded in secrecy and partial disclosures, said You, who was not part of the report. "The general public should be concerned, but not panicked," he added. Jean Su, director of the Energy Justice Program at the Center for Biological Diversity, said the report is important because it is the first U.N., or even global, report "that shines a light on the environmental harms of AI." National Artificial Intelligence Association President Caleb Max emphasized how his industry is becoming more efficient and how it benefits the public: "AI is rapidly becoming part of our everyday lives and adding benefits that improve safety, live longer, work more efficiently, enhance food production, and reduce poverty. The evidence is growing daily that the energy return on investment of AI development is transformative for our world and therefore more than worth it." Josh Levi, president the Data Center Coalition, said the industry takes its environmental impact seriously. "We remain committed to working with policymakers, local communities, and industry partners to ensure that as data centers grow, they do so responsibly, transparently, and in ways that reflect the best available practices," he said in a statement. How much energy your query uses and how to trim it Madani, also the winner of the most recent of the Stockholm Water Prize, said the numbers show the environmental cost of AI, which may seem cleaner at first glance than other mechanical devices, such as cars and furnaces, that have visible pollution. "AI is not just a virtual thing. We're talking about something that has physics, something that has real impacts. There is infrastructure there. There is energy that is being used," Madani said. "A lot of hardware is behind all these operations that to us seem very, very clean because we don't see smoke out of our devices. On our cellphone, there is no visible smoke or out of our computer or something. But somewhere else someone is suffering." People can reduce AI's massive energy appetite by being less polite and more concise in their queries, Madani said. The report found that cutting word use in requests by 30% can reduce energy used by AI by 25%. That would save about the same amount of electricity as what about 700,000 people in Africa use in a year, the report said. "If you're too polite, then that extra 'please' you put there can make a huge difference," Madani said. "You've got to be very precise and be short." A typical ChatGPT-style query is about 200 times more energy-intensive than the type of basic text classification used in an email spam filter, for example. AI-generated images or video require much more energy. And the more complicated the AI, the more energy it takes to train or learn. The report said GPT-3 used about 1.3 billion watt-hours to train, but the next version used 50 to 70 billion watt-hours. But it's not training that really feasts on power, said study co-author Miriam Aczel, a United National University environmental policy researcher. About 90% of the power use of AI comes from operational requests, she said. GPT alone accounts for 2.5 billion prompts a day, she said. Efficiency still means more power use Even though tech advocates can argue that their machines are becoming more efficient, there's a common paradox that finds when things get more efficient, they are used more often and total energy use soars even if individual uses are more efficient, Madani said. While some companies tout the use of renewable energy for data centers, Madani said that means the supply of clean electricity is depleted and thus dirtier energy is used elsewhere. One of the problems in conducting this study is that many companies and places are not transparent about what data centers and AI are consuming or even where and how big they are, Aczel and Madani said. "We cannot manage what companies do not disclose," Cornell's You said. ___ The Associated Press' climate and environmental coverage receives financial support from multiple private foundations. AP is solely responsible for all content. Find AP's standards for working with philanthropies, a list of supporters and funded coverage areas at AP.org.

'Some places carry the costs and other places capture the benefits': UN warns that AI is stripping natural resources in exchange for making the rich richer

* UN warns AI's environmental footprint is much, much more than just energy * AI data centers could consume the water equivalent of 1.3b people by 2030 * Report calls for more diverse reporting and robust governance to protect people A new UN report is arguing artificial intelligence's impacts are far from parity - instead, that its environmental impact is being underestimated because most discussions only focus on carbon emissions. Instead, the United Nations is urging companies, investors and governments to also include water consumption and land use in their evaluations. This comes as AI data centers alone are expect to consume 945TWh of electricity by 2030 - the equivalent of 1.95b homes, or three times the population of Pakistan, Bangladesh and Nigeria. UN is worried about AI's environmental impacts Electricity aside, the UN is also warning that their water consumption by the end of the decade would be equal to 1.3b people in Sub-Saharan Africa (9.3 trillion liters), and the land use could equate to 14,500 square kilometers (twice as big as Jakarta, home to 32m people). But it's far more than the environment alone that the AI industry is putting under pressure - unlike conventional software, artificial intelligence leans heavily on physical data center campuses, grid connections, cooling systems and semiconductors, expanding its impacts heavily across both Scope 2 and Scope 3. Professor Kaveh Madani, Director of the United Nations University Institute for Water, Environment and Health, stressed that the report should serve as a blocker to AI. Instead, Madani calls for responsibility and sustainability. "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." Interestingly, while much of the debate often centers around model training, researchers now believe inference (the everyday use after deployment) accounts for around 80-90% of AI's energy demand. ChatGPT alone is said to process around 2.5b prompts per day, and energy demand is only increasing as response quality improves. Looking ahead, the UN is called for the mandatory reporting of carbon, land and water footprints as well as 'efficiency by design' approaches. The paper also encourages stronger governance to prevent the environmental costs from being shifted onto the most vulnerable communities. Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

New reports give sobering numbers on data centers' resource needs

Why it matters: They show how many different kinds of impact data centers can have -- including their water needs, but others too. Driving the news: The UN's academic arm warns that too often a "carbon-only lens" dominates the topic. * "'[L]ow-carbon' is not automatically 'low-water' or 'low-land,'" states the report from United Nations University (UNU). * "Evaluating sustainability through a single metric can hide trade-offs and shift burdens onto places already facing water stress or land pressure." 🧮 Stunning stats: On "current trajectories," data center power demand could be "nearly triple" the combined annual electricity use of Pakistan, Bangladesh, and Nigeria by 2030, the report finds. * ⚡ It would be equivalent to the 6th-largest power consuming country. * 💧 The "associated water footprint" would equal the "annual domestic water needs of all 1.3 billion residents of Sub-Saharan Africa." * 🧑🌾 "The land footprint associated with this electricity would exceed 14,500 km², nearly 10 times the size of Mexico City," it states. * 🥐 "AI infrastructure could generate up to 2.5 million metric tons of e-waste annually by 2030, equivalent to discarding nearly 250 Eiffel Towers every year." The other report, from Bank of America, finds that "most US water utilities have yet to fully account for the water implications of AI‑driven infrastructure growth." * About 75% of water use is from "off site" needs -- power generation and hardware manufacturing. * For looking within data centers, it's also a good resource for exploring the water vs. electricity tradeoffs of different chip cooling methods that Amy touched on earlier this week. Zoom out: AI can help improve climate friendly tech, such as helping create better battery chemistries. And tech giants are staking new clean energy projects. * But there's growing backlash over data centers' effect on power bills -- real or perceived -- as well as massive energy demands, and emissions from coal- and gas-fired power helping to meet it. Yes, but: One emerging school of thought is that the AI buildout offers a unique chance to spur U.S. grid modernization that's desperately needed anyway. What we're watching: The UN report offers principles for building a "responsible AI ecosystem" globally.

AI could consume more electricity than entire nations by 2030

By 2030, the data centers powering global artificial intelligence are projected to consume 945 terawatt-hours of electricity per year. To put that in human terms: it's nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria - countries home to more than 650 million people. The water those data centers will require equals the basic annual domestic water needs of every person in Sub-Saharan Africa. The land they'll occupy exceeds twice the area of the Jakarta metropolitan area, home to 32 million people. These numbers come from a new report by the United Nations University Institute for Water, Environment and Health (UNU-INWEH). The research suggests that the conversation around AI's environmental impact has been missing most of the picture. Carbon alone is the wrong measure Most assessments of AI's environmental cost focus on carbon emissions, particularly the emissions from training large models. That framing has always been incomplete, and the report argues it's becoming dangerously misleading. Every kilowatt-hour of electricity that runs an AI system carries not just a carbon footprint but a water footprint from cooling and power generation, and a land footprint from energy infrastructure and supply chains. And crucially, these three don't move in the same direction. Switching from coal to bioenergy can cut the carbon footprint of electricity by 70%, while increasing its water footprint more than thirtyfold and its land footprint a hundredfold. "What surprised us most is how often the choices that look greenest from a carbon perspective end up worse for water or for land," said lead author Miriam Aczel. "If we keep judging AI sustainability by carbon alone, we might think that renewables make AI infrastructure clean - but that is solving one problem while creating other problems, often in places that didn't ask for it." "Low-carbon" is not automatically "low-impact." Evaluating AI's sustainability through a single metric can hide trade-offs. It can also shift environmental burdens onto regions already facing water or land stress. In many cases, these regions are not the ones benefiting from the AI being run there. The AI training story is already outdated Public discussion has largely focused on the energy cost of training massive AI models. Training GPT-3 required an estimated 1.3 gigawatt-hours of electricity, while GPT-4 is estimated to have consumed between 50 and 70 gigawatt-hours. But once a model is deployed, training becomes a minor part of the picture. Inference - the continuous running of deployed models to answer everyday queries - accounts for an estimated 80 to 90% of total AI energy use. ChatGPT alone processes around 2.5 billion prompts per day, translating to roughly 383 gigawatt-hours of electricity per year for a single product. Offsetting that carbon footprint would require 2.6 million tree seedlings grown for a decade - enough trees to cover an area the size of Manhattan. Not all queries are equal One of the more striking parts of the report concerns the enormous variation in energy use across different types of AI tasks. A typical conversational chat query uses around 200 times more energy than basic text classification. Generating a single AI image can require around 1,450 times the baseline, while a single short AI-generated video can consume as much electricity as 200,000 spam classifications. These differences are largely invisible to users. Model choice, prompt length, output format, and resolution all materially shape the footprint - but most of these are determined by product defaults the user never sees or consciously chooses. More efficient AI means more consumption The report also invokes what economists call the Jevons Paradox - the well-documented tendency for efficiency improvements to be swallowed by increased consumption rather than resulting in net savings. As AI becomes more efficient and cheaper to run, it gets used more, across more tasks, at higher volumes. Without explicit limits on tokens, output resolution, or default response length, per-query improvements are likely to be absorbed by sheer growth in volume. "A lot of people think that the environmental footprint of AI reduces as technology improves and processes become more efficient," said lead investigator Kaveh Madani, the director of UNU-INWEH. "But more efficient and affordable AI and energy mean more consumption of AI, making the overall footprint far bigger than what we save through efficiency gains." Local costs, distant benefits The report is at its most pointed when it maps where the infrastructure is being built against who bears the consequences. In Ireland, data centers accounted for 21% of total metered electricity in 2023 - more than all urban households combined. The national grid operator has paused new approvals around Dublin until 2028. In Querétaro, Mexico, expanding computer infrastructure is drawing on water supplies during prolonged droughts. In Uruguay, plans for a water-intensive data center coincided with a 2023 drought that depleted Montevideo's freshwater reserves and made tap water temporarily unsafe to drink. In each case, the communities experiencing the strain are not necessarily the ones using the AI being run in those facilities. "If you map where data centers are getting built against where water stress is worst, you tend to see the same regions in some instances," said study co-author Mir Matin. "And the communities living near these sites are not necessarily the ones using the AI being run there." "That asymmetry is the issue. Without fixing it, we'll just be repeating older patterns, where some places carry the costs and other places capture the benefits." Inequality beyond local impacts The inequality runs deeper than local environmental impacts. Only 32 countries in the world host AI-specialised data centers. More than 90% of that capacity is concentrated in just two countries - the United States and China. More than 150 countries currently have little or no access to sovereign AI computing infrastructure. The same countries excluded from AI's benefits are, in many cases, the ones bearing the costs of critical mineral extraction and electronic waste processing. AI infrastructure is projected to generate up to 2.5 million tons of electronic waste per year by 2030. This is the equivalent of discarding nearly 250 Eiffel Towers annually, much of it processed in low income economies with limited environmental safeguards. The report isn't arguing against AI It's worth being clear about what this report is and isn't. It's not making the case that AI is bad or that its development should stop. "This report is not a case against artificial intelligence, a technological transformation that is improving the lives of billions of people around the world," Madani said. "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." What it is arguing is that the current governance framework is missing most of what matters. Carbon-only metrics are insufficient. Disclosure standards need to require water and land footprints alongside carbon, across both training and inference, across jurisdictions. Siting decisions need to be treated as environmental decisions. Efficiency improvements need to be accompanied by demand-side guardrails to prevent volume growth from swallowing the gains. A question of governance The report also makes a point that tends to get lost in technical discussions about AI sustainability: this is ultimately a governance question, not a technical one. The capability to build AI within planetary limits exists. Whether that happens depends on measurement, transparency, and shared responsibility across the entire system, from the companies building the models to the governments permitting the infrastructure to the investors funding both. "The global system building artificial intelligence must also govern it sustainably and fairly," said Tshilidzi Marwala, the Rector of the United Nations University. "AI can certainly advance prosperity and human well-being. Whether it does so equitably is now a governance question, not a technical one," he concluded. The full report can be found here. -- - Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

AI Could Use as Much Water as 1.3 Billion People by 2030

The water used by artificial intelligence is expected to equal the needs of 1.3 billion people by 2030 -- threatening natural resources for billions around the world. That's according to a new report from the United Nations University Institute for Water, Environment and Health (UNU-INWEH) which quantifies the carbon, water, and land footprints of AI's electricity use around the globe. The report finds that AI's environmental cost is often mismeasured -- focusing solely on carbon emissions. However, cooling and generating power for data centers comes with a "water footprint," while the energy infrastructure and supply chains to build the data centers have a "land footprint." These are important factors to consider, the report says, when analyzing the stressors a region might be facing due to data centers. By 2030, the report finds, global data centers 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 that together are home to more than 650 million people. The water footprint of data centers is projected to equal the basic domestic water needs of all 1.3 billion people in Sub-Saharan Africa for a year, while their land footprint could exceed 5,590 square miles, roughly twice the Jakarta metropolitan area that's currently home to more than 32 million people. But switching to cleaner sources of energy isn't as simple as it sounds. Minimizing one footprint could come at the expense of magnifying another, researchers say. For example, switching from coal to bioenergy cuts electricity's carbon footprint by 70% -- but increases its water footprint more than 30-fold and its land footprint 100-fold. "What surprised us most is how often the choices that look greenest from a carbon perspective end up worse for water or for land," Miriam Aczel, UNU-INWEH researcher and the lead author of the report, said in a press release. "If we keep judging AI sustainability by carbon alone, we might think that renewables make AI infrastructure clean but that is solving one problem while creating other problems, often in places that didn't ask for it." For a number of communities around the globe, AI is already using up significant energy resources. In 2025 alone, data centers consumed an estimated 448 terawatt-hours of electricity, the report found -- more than the country of Saudi Arabia. In many cases, this excessive energy use comes at a cost to those who reside near them. In Ireland, data centers accounted for 21% of total metered electricity in 2023, exceeding electricity use by urban households. (The country's national grid operator has since paused new approvals around Dublin until 2028.) Large data centers can consume up to 5 million gallons per day to keep servers cool. In communities already facing water scarcity, this risks putting a strain on already sparse resources. In Querétaro, Mexico, plans for fast-tracked data centers stand to jeopardize water supplies amid prolonged droughts. Uruguay saw a similar battle after plans to build a water-intensive data center were announced during a 2023 drought that depleted freshwater reserves in the country's largest city, making tap water unsafe to drink -- sparking protests over the prioritization of industrial demands over human needs. In addition to the strain on resources and impact to local environments, there is a separate inequality at play, the report notes. As data centers continue to explode around the world, the researchers warn of a widening "digital divide," in which wealthier countries are able to invest in AI infrastructure while lower-income nations struggle to access and participate in the AI economy. In some ways, this divide is already apparent. As of 2025, only 32 countries -- 16% of nations -- host AI-specialized data centers, and 90% of that capacity is concentrated in two countries: the U.S. and China. Moreover, AI infrastructure could generate up to 2.5 million metric tons of electronic waste each year by 2030, which could expose frontline communities -- predominantly in low-income countries where many countries export their waste -- to toxic substances. "The concentrated development of AI infrastructure in the privileged areas of the world is creating a large digital divide that poses profound challenges in the equitable development of AI," Tshilidzi Marwala, rector of the United Nations University and Under-Secretary-General of the United Nations, said in a press release. "AI can certainly advance prosperity and human well-being. Whether it does so equitably is now a governance question, not a technical one." To ensure that data center development doesn't come at a cost to communities, the report calls for a "responsible AI ecosystem," and notes that permitting, environmental impact assessment, and community consultation should reflect the reality of water and land use along with carbon. Governments, investors, and financial institutions must implement the guardrails that will minimize environmental consequences, said Kaveh Madani, director of UNU-INWEH. "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."

Energy use of AI and data centres rivals most countries: UN report

"That extra 'please' you put there can make a huge difference," says one of the report's authors. The environmental footprint of data centres already rivals some of the world's largest countries, according to a United Nations University report released on 3 June. Their water use, energy use and pollution is predicted to double in just four years as use of artificial intelligence grows. Much of the growth of data centres is being driven by AI. About 20 per cent of data centres' energy is currently due to AI, but that should grow to 40 per cent by 2030, the report said. AI users can reduce the climate impact of their queries by less polite and more concise in their queries, one of the report's authors advises. The majority of people - 70 per cent - are polite to AI when interacting with it, according to a survey carried out by British publisher Future in 2024. Of the respondents, 55 per cent said they do this because "it's just the nice thing to do", while 12 per cent said it was because "when the robot uprising happens, I don't want to come first". Electricity use equal to that of Argentina Last year, global data centres used 448 trillion watt-hours of electricity, more than all but 10 countries of the world, said the report. That electricity use produced about 189 million tonnes of carbon dioxide, about the same amount as Argentina, and producing that much energy consumed about 4.5 trillion litres of water, according to the report on the environmental consequences of AI's energy use. By 2030, data centres will account for nearly three per cent of the world's projected electricity use, with 935 trillion watt-hours. If data centres were a country, the country would be projected to rank sixth-highest in power use in 2030. That would produce nearly 399 million tonnes of carbon dioxide, the report said. The study focused on energy use and didn't examine the massive amount of water used to cool data centres. "If you look at these numbers, we're seeing scales comparable to nations," says study co-author Kaveh Madani, a water scientist and director of the United Nations University Institute for Water, Environment and Health in Canada. "The demand is enormous." First global look at ecological impact of data centres The report is significant because of the credibility and authority of the UN, not just because of any one set of eye-popping numbers, says Fengqi You, a Cornell University energy engineering professor who directs the college's AI sustainability issues. "Its value is that a UN institution is putting carbon, water, land, life-cycle impacts and environmental justice into one frame" for an issue that is often shrouded in secrecy and partial disclosures, says You, who was not part of the report. "The general public should be concerned, but not panicked," he adds. Jean Su, director of the Energy Justice Program at the Center for Biological Diversity, said the report is important because it is the first UN, or even global, report "that shines a light on the environmental harms of AI". National Artificial Intelligence Association President Caleb Max emphasises how his industry is becoming more efficient and how it benefits the public: "AI is rapidly becoming part of our everyday lives and adding benefits that improve safety, [help people] live longer, work more efficiently, enhance food production, and reduce poverty. The evidence is growing daily that the energy return on investment of AI development is transformative for our world and therefore more than worth it." Josh Levi, president the Data Center Coalition, says the industry takes its environmental impact seriously. "We remain committed to working with policymakers, local communities, and industry partners to ensure that as data centres grow, they do so responsibly, transparently, and in ways that reflect the best available practices," he said in a statement. The report came just after Californian city Monterey Park became the first in the US to vote for a permanent ban on data centres on Tuesday (2 June). How much energy your query uses and how to trim it Madani, also the winner of the most recent of the Stockholm Water Prize, says the numbers show the environmental cost of AI, which may seem cleaner at first glance than other mechanical devices, such as cars and furnaces, that have visible pollution. "AI is not just a virtual thing. We're talking about something that has physics, something that has real impacts. There is infrastructure there. There is energy that is being used," Madani says. "A lot of hardware is behind all these operations that to us seem very, very clean because we don't see smoke out of our devices. On our cellphone, there is no visible smoke or out of our computer or something. But somewhere else someone is suffering." People can reduce AI's massive energy appetite by being less polite and more concise in their queries, Madani says. The report found that cutting word use in requests by 30 per cent can reduce energy used by AI by 25 per cent. That would save about the same amount of electricity as what about 700,000 people in Africa use in a year, the report said. "If you're too polite, then that extra 'please' you put there can make a huge difference," Madani says. "You've got to be very precise and be short." A typical ChatGPT-style query is about 200 times more energy-intensive than the type of basic text classification used in an email spam filter, for example. AI-generated images or video require much more energy. And the more complicated the AI, the more energy it takes to train or learn. The report said GPT-3 used about 1.3 billion watt-hours to train, but the next version used 50 to 70 billion watt-hours. But it's not training that really feasts on power, says study co-author Miriam Aczel, a United Nations University environmental policy researcher. About 90 per cent of the power use of AI comes from operational requests, she says. GPT alone accounts for 2.5 billion prompts a day, she says. Efficiency still means more power use Even though tech advocates can argue that their machines are becoming more efficient, there's a common paradox that finds when things get more efficient, they are used more often and total energy use soars even if individual uses are more efficient, Madani says. While some companies tout the use of renewable energy for data centres, Madani says that means the supply of clean electricity is depleted and thus dirtier energy is used elsewhere. One of the problems in conducting this study is that many companies and places are not transparent about what data centres and AI are consuming or even where and how big they are, Aczel and Madani say. "We cannot manage what companies do not disclose," Cornell's You says.

UN urges AI firms to reveal environmental footprint

Paris (France) (AFP) - A UN report on Wednesday urged artificial intelligence firms to disclose their environmental footprint, warning that the AI boom is putting growing pressure on power grids, water supplies and land resources. The study also urged governments to require standardised environmental reporting from AI providers, and called on users to choose less energy-intensive tools that can accomplish the same task. "What we are showing here is probably just the tip of the iceberg," Kaveh Madani, director of the United Nations University Institute for Water, Environment and Health (UNU-INWEH), told AFP. "We need to require more transparency. We need the providers to provide that information," Madani said. The authors of the report, "Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints", used primary data from a range of sources to make their estimates, Madani said. 'Not an anti-AI report' The global AI market is expected to grow from $189 billion in 2023 to $4.8 trillion by 2033, the UNU-INWEH report said. Data centres, the warehouses of servers that power AI and other digital services, consumed 448 terawatt-hours (TWh) of electricity in 2025. If data centres were a country, their consumption would have ranked in 11th place -- just under France with 468 TWh, the study said. AI workloads accounted for a fifth of the total electricity use at data centres last year, and they are expected to rise to 40 percent by 2030. Consumption by data centres is projected to exceed 945 TWh by 2030, ranking sixth among countries and emitting 399 million tonnes of CO2 equivalent. By comparison, the UK's net emissions reached 367 million tonnes last year. The report cautioned that reducing carbon emissions did not automatically reduce water or land impacts. Data centres could guzzle 9.32 trillion litres of water by 2030, enough to meet the annual basic water needs of the entire population of sub-Saharan Africa, the report said. The land they occupy would be 18 times bigger than New York City. ChatGPT alone is estimated to process around 2.5 billion prompts per day, translating into roughly 383 GWh of electricity a year -- enough to meet the annual demand of nearly three million people in sub-Saharan Africa, the report said. AI videos are the the most energy-hungry product. A single short AI-generated clip can draw as much electricity as hundreds of AI-generated images. The report also warned of a growing digital divide, with most AI-specialised data centres located in the United States, China and the European Union while many developing countries bear environmental costs linked to mineral extraction and waste disposal. "This is not an anti-AI report," Madani said. "We are simply saying that we have to proactively monitor their impacts to be able to curb them, to be able to control them before it's too late." Chatbot or cookbook? The report said AI developers and service providers should "make the invisible visible" by publishing clear, standardised accounts of energy and environmental footprints for training models and generating responses for users. AI firms should also improve efficiency of their systems. "Governments and regulators should treat environmental disclosure for AI as routine," it said. Government climate and energy plans should incorporate growing AI demand while efforts should be made to keep data centres away from water-stressed regions. But individual users should also avoid using AI for tasks that could be done with conventional tools, the authors said. "All of us can make a huge difference," Madani said. The report estimates that a single AI‑enhanced internet search may use 10 times more energy than a conventional search. "Do you need ChatGPT to give you a recipe" or "(do) you have a cookbook that's sitting on your kitchen counter that you could just open?" UNU-INWEH co-author Miriam Aczel told AFP. "There are a lot of simple behavioural tweaks that people can make that can help reduce their footprint," she said. "But I think all of this starts with knowledge, information and disclosures."

UN Calculates Nation-Sized Environmental Footprints for AI and Data Centers

WASHINGTON (AP) -- The environmental footprint of data centers already rivals some of the world's largest countries, according to a United Nations University report, which also predicts their water and energy use and pollution will double in just four years as use of artificial intelligence grows. Last year, global data centers used 448 trillion watt-hours of electricity, more than all but 10 countries of the world, said the report issued Wednesday. That electricity use produced about 208 million tons (189 million metric tons) of carbon dioxide, about the same amount as Argentina, and producing that much energy consumed about 1.2 trillion gallons (4.5 trillion liters) of water, according to the report on the environmental consequences of AI's energy use. By 2030, data centers will account for nearly 3% of the world's projected electricity use, with 935 trillion watt-hours. If data centers were a country, the country would be projected to rank sixth-highest in power use in 2030. That would produce nearly 440 million tons (399 million metric tons) of carbon dioxide, the report said. The study focused on energy use and didn't examine the massive amount of water used to cool data centers. "If you look at these numbers, we're seeing scales comparable to nations," said study co-author Kaveh Madani, a water scientist and director of the United Nations University Institute for Water, Environment and Health in Canada. "The demand is enormous." Much of the growth of data centers is being driven by AI. About 20% of data centers' energy is currently due to AI, but that should grow to 40% by 2030, the report said. First global look at ecological impact The report is significant because of the credibility and authority of the U.N., not just because of any one set of eye-popping numbers, said Fengqi You, a Cornell University energy engineering professor who directs the college's AI sustainability issues. "Its value is that a U.N. institution is putting carbon, water, land, life-cycle impacts and environmental justice into one frame" for an issue that is often shrouded in secrecy and partial disclosures, said You, who was not part of the report. "The general public should be concerned, but not panicked," he added. Jean Su, director of the Energy Justice Program at the Center for Biological Diversity, said the report is important because it is the first U.N., or even global, report "that shines a light on the environmental harms of AI." National Artificial Intelligence Association President Caleb Max emphasized how his industry is becoming more efficient and how it benefits the public: "AI is rapidly becoming part of our everyday lives and adding benefits that improve safety, live longer, work more efficiently, enhance food production, and reduce poverty. The evidence is growing daily that the energy return on investment of AI development is transformative for our world and therefore more than worth it." Josh Levi, president the Data Center Coalition, said the industry takes its environmental impact seriously. "We remain committed to working with policymakers, local communities, and industry partners to ensure that as data centers grow, they do so responsibly, transparently, and in ways that reflect the best available practices," he said in a statement. How much energy your query uses and how to trim it Madani, also the winner of the most recent of the Stockholm Water Prize, said the numbers show the environmental cost of AI, which may seem cleaner at first glance than other mechanical devices, such as cars and furnaces, that have visible pollution. "AI is not just a virtual thing. We're talking about something that has physics, something that has real impacts. There is infrastructure there. There is energy that is being used," Madani said. "A lot of hardware is behind all these operations that to us seem very, very clean because we don't see smoke out of our devices. On our cellphone, there is no visible smoke or out of our computer or something. But somewhere else someone is suffering." People can reduce AI's massive energy appetite by being less polite and more concise in their queries, Madani said. The report found that cutting word use in requests by 30% can reduce energy used by AI by 25%. That would save about the same amount of electricity as what about 700,000 people in Africa use in a year, the report said. "If you're too polite, then that extra 'please' you put there can make a huge difference," Madani said. "You've got to be very precise and be short." A typical ChatGPT-style query is about 200 times more energy-intensive than the type of basic text classification used in an email spam filter, for example. AI-generated images or video require much more energy. And the more complicated the AI, the more energy it takes to train or learn. The report said GPT-3 used about 1.3 billion watt-hours to train, but the next version used 50 to 70 billion watt-hours. But it's not training that really feasts on power, said study co-author Miriam Aczel, a United National University environmental policy researcher. About 90% of the power use of AI comes from operational requests, she said. GPT alone accounts for 2.5 billion prompts a day, she said. Efficiency still means more power use Even though tech advocates can argue that their machines are becoming more efficient, there's a common paradox that finds when things get more efficient, they are used more often and total energy use soars even if individual uses are more efficient, Madani said. While some companies tout the use of renewable energy for data centers, Madani said that means the supply of clean electricity is depleted and thus dirtier energy is used elsewhere. One of the problems in conducting this study is that many companies and places are not transparent about what data centers and AI are consuming or even where and how big they are, Aczel and Madani said. "We cannot manage what companies do not disclose," Cornell's You said. ___ The Associated Press' climate and environmental coverage receives financial support from multiple private foundations. AP is solely responsible for all content. Find AP's standards for working with philanthropies, a list of supporters and funded coverage areas at AP.org.

Energy, water use and pollution of AI and data centres rival most countries

WASHINGTON -- The environmental footprint of data centres already rivals some of the world's largest countries, according to a United Nations University report, which also predicts their water and energy use and pollution will double in just four years as use of artificial intelligence grows. Last year, global data centres used 448 trillion watt-hours of electricity, more than all but 10 countries of the world, said the report issued Wednesday. That electricity use produced about 208 million tons (189 million metric tons) of carbon dioxide, about the same amount as Argentina, and producing that much energy consumed about 1.2 trillion gallons (4.5 trillion liters) of water, according to the report on the environmental consequences of AI's energy use. By 2030, data centres will account for nearly three per cent of the world's projected electricity use, with 935 trillion watt-hours. If data centres were a country, the country would be projected to rank sixth-highest in power use in 2030. That would produce nearly 440 million tons (399 million metric tons) of carbon dioxide, the report said. The study focused on energy use and didn't examine the massive amount of water used to cool data centers. "If you look at these numbers, we're seeing scales comparable to nations," said study co-author Kaveh Madani, a water scientist and director of the United Nations University Institute for Water, Environment and Health in Canada. "The demand is enormous." Much of the growth of data centres is being driven by AI. About 20% of data centres' energy is currently due to AI, but that should grow to 40% by 2030, the report said. First global look at ecological impact The report is significant because of the credibility and authority of the U.N., not just because of any one set of eye-popping numbers, said Fengqi You, a Cornell University energy engineering professor who directs the college's AI sustainability issues. "Its value is that a U.N. institution is putting carbon, water, land, life-cycle impacts and environmental justice into one frame" for an issue that is often shrouded in secrecy and partial disclosures, said You, who was not part of the report. "The general public should be concerned, but not panicked," he added. Jean Su, director of the Energy Justice Program at the Center for Biological Diversity, said the report is important because it is the first U.N., or even global, report "that shines a light on the environmental harms of AI." National Artificial Intelligence Association President Caleb Max emphasized how his industry is becoming more efficient and how it benefits the public: "AI is rapidly becoming part of our everyday lives and adding benefits that improve safety, live longer, work more efficiently, enhance food production, and reduce poverty. The evidence is growing daily that the energy return on investment of AI development is transformative for our world and therefore more than worth it." Josh Levi, president the Data Center Coalition, said the industry takes its environmental impact seriously. "We remain committed to working with policymakers, local communities, and industry partners to ensure that as data centres grow, they do so responsibly, transparently, and in ways that reflect the best available practices," he said in a statement. How much energy your query uses and how to trim it Madani, also the winner of the most recent of the Stockholm Water Prize, said the numbers show the environmental cost of AI, which may seem cleaner at first glance than other mechanical devices, such as cars and furnaces, that have visible pollution. "AI is not just a virtual thing. We're talking about something that has physics, something that has real impacts. There is infrastructure there. There is energy that is being used," Madani said. "A lot of hardware is behind all these operations that to us seem very, very clean because we don't see smoke out of our devices. On our cellphone, there is no visible smoke or out of our computer or something. But somewhere else someone is suffering." People can reduce AI's massive energy appetite by being less polite and more concise in their queries, Madani said. The report found that cutting word use in requests by 30% can reduce energy used by AI by 25%. That would save about the same amount of electricity as what about 700,000 people in Africa use in a year, the report said. "If you're too polite, then that extra 'please' you put there can make a huge difference," Madani said. "You've got to be very precise and be short." A typical ChatGPT-style query is about 200 times more energy-intensive than the type of basic text classification used in an email spam filter, for example. AI-generated images or video require much more energy. And the more complicated the AI, the more energy it takes to train or learn. The report said GPT-3 used about 1.3 billion watt-hours to train, but the next version used 50 to 70 billion watt-hours. But it's not training that really feasts on power, said study co-author Miriam Aczel, a United National University environmental policy researcher. About 90% of the power use of AI comes from operational requests, she said. GPT alone accounts for 2.5 billion prompts a day, she said. Efficiency still means more power use Even though tech advocates can argue that their machines are becoming more efficient, there's a common paradox that finds when things get more efficient, they are used more often and total energy use soars even if individual uses are more efficient, Madani said. While some companies tout the use of renewable energy for data centres, Madani said that means the supply of clean electricity is depleted and thus dirtier energy is used elsewhere. One of the problems in conducting this study is that many companies and places are not transparent about what data centres and AI are consuming or even where and how big they are, Aczel and Madani said. "We cannot manage what companies do not disclose," Cornell's You said.