AI’s Environmental Toll Is ‘Tip of the Iceberg,’ UN Warns as It Calls for Mandatory Disclosures
The rapid ascent of artificial intelligence is imposing a hidden and growing burden on the planet’s resources, the United Nations warned Wednesday, urging AI developers to come clean about their environmental footprint before the damage becomes irreversible.
In a report released by the United Nations University Institute for Water, Environment and Health (UNU-INWEH), researchers called on governments to mandate standardized environmental reporting from AI providers and pressed individual users to opt for less energy-intensive tools when possible.
“What we are showing here is probably just the tip of the iceberg,” said Kaveh Madani, director of UNU-INWEH, in an interview. “We need to require more transparency. We need the providers to provide that information.”
The report, titled Environmental Cost of AI’s Energy Use: Carbon, Water and Land Footprints, draws on primary data from a range of sources to estimate the industry’s resource consumption—and warns that the full picture remains obscured by a lack of corporate disclosure.
Data Centers Now Rival Entire Countries
The global AI market is projected to balloon from $189 billion in 2023 to $4.8 trillion by 2033, according to the study. Underpinning that growth are data centers—the vast server warehouses that power AI and other digital services—which already consumed 448 terawatt-hours (TWh) of electricity in 2025. If data centers were a nation, their power use would rank 11th globally, just behind France’s 468 TWh.
AI workloads accounted for about one-fifth of data center electricity last year, but that share is expected to climb to 40 percent by 2030. By then, total data center consumption could exceed 945 TWh, placing the sector sixth among countries and generating an estimated 399 million tonnes of CO₂ equivalent—more than the United Kingdom’s net emissions of 367 million tonnes in 2024.
Water and Land Footprints Mount
The report cautioned that cutting carbon emissions does not automatically reduce water or land impacts. By 2030, data centers could consume 9.32 trillion litres of water annually—enough to meet the basic drinking, sanitation and hygiene needs of every person in sub-Saharan Africa. The physical land they occupy could expand to 18 times the size of New York City.
Even a single AI tool carries significant weight. ChatGPT alone processes an estimated 2.5 billion prompts per day, translating to roughly 383 gigawatt-hours (GWh) of electricity a year—enough to power nearly 3 million people in sub-Saharan Africa. AI-generated video clips are the most energy-hungry product, with a single short clip drawing as much electricity as hundreds of AI-generated images.
A Widening Digital Divide
The study also highlighted a growing global inequity: most AI-specialized data centers are concentrated in the United States, China and the European Union, while developing countries disproportionately bear the environmental costs of mineral extraction and electronic waste disposal.
“This is not an anti-AI report,” Madani stressed. “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.”
What Can Be Done
The report urges AI developers and service providers to “make the invisible visible” by publishing clear, standardized accounts of energy and environmental footprints for both training models and generating responses. It also calls for efficiency improvements, integration of AI demand into national climate and energy plans, and siting data centers away from water-stressed regions.
Individual users can also make a difference. A single AI-enhanced internet search may use 10 times more energy than a conventional search, the report notes. “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?” asked Miriam Aczel, a co-author from UNU-INWEH, adding: “There are a lot of simple behavioral tweaks that people can make. But I think all of this starts with knowledge, information and disclosures.”
The authors hope the report will push regulators and the public to treat environmental disclosure for AI as routine—not as an afterthought.