A new UN study warns that the rapid expansion of artificial intelligence could exact a severe environmental toll by the end of the decade, with water use, electricity demand, land consumption and e-waste all rising sharply. The researchers say AI could use as much water as the basic household needs of nearly 1.5 billion people, produce millions of tons of electronic waste, and widen the gap between countries that control computing power and those left behind.
The report says most of AI’s energy use does not come from training models, but from the billions of people using the technology every day. By 2030, data centers supporting AI services could consume about 945 terawatt-hours of electricity a year, almost triple the combined annual electricity use of Pakistan, Bangladesh and Nigeria, home to more than 650 million people. It also says AI’s land footprint could exceed 14,500 square kilometers, nearly 12 times New York City’s area and about two thirds of Israel’s.
The authors argue that the main issue is the consumers, not just the developers. They say 80 to 90 percent of total AI energy use comes from daily use, and cite one unnamed popular AI service that handles about 2.5 billion requests a day and uses hundreds of gigawatt-hours annually. They also warn that efficiency gains may backfire through a rebound effect, because cheaper and better systems attract more use, driving resource demand even higher.
The study says the environmental cost is not limited to carbon emissions. Renewable energy may cut greenhouse gases, but it can still increase water use and land use, especially in resource-stressed regions. It warns that data centers already account for a large share of national electricity use in some countries, while new facilities can consume huge amounts of water during droughts.
The report also highlights growing e-waste, saying AI infrastructure could generate about 2.5 million tons a year by 2030, much of it likely ending up in poorer countries that cannot safely process it. It adds that rising mining for minerals used in chips and AI accelerators raises environmental and human-rights concerns. More than 90 percent of dedicated AI computing power is concentrated in the United States and China, while over 150 countries still lack significant AI infrastructure. The authors say they are not calling to stop AI development, but to make it more responsible through transparency, better system design, international cooperation and lifecycle accountability, with governments factoring in power, water and land needs and companies building more efficient systems.
