AI’s environmental cost threatens water, land resources for billions by 2030: UN report

Artificial intelligence’s rapid expansion is putting growing pressure on global electricity, water, and land resources, with data centers powering the technology projected to consume 945 terawatt-hours of electricity by 2030, a UN University report warned on Friday.

The report by the UN University Institute for Water, Environment and Health said AI’s environmental impact is being “systematically mismeasured,” because most assessments focus primarily on carbon emissions while overlooking its water and land footprints.

By 2030, AI-related data centers’ water footprint could reach 9.3 trillion liters — equivalent to the basic annual domestic water needs of 1.3 billion people in Sub-Saharan Africa — while their land footprint could exceed 14,500 square kilometers, roughly twice the size of the Jakarta metropolitan area.

“This report is not a case against artificial intelligence,” said Kaveh Madani, director of UNU-INWEH, calling instead for responsible use of AI and proactive action to address its unintended impacts.

The report said global data centers consumed an estimated 448 TWh of electricity in 2025. If considered as a country, they would rank as the world’s 11th largest electricity consumer.

The report warned that “low-carbon” does not necessarily mean “low-water” or “low-land,” noting that some energy transitions can reduce emissions while increasing pressure on water and land resources.

The report also said public debate has focused too heavily on the energy required to train large AI models, even though most AI energy use comes from inference — the process of running deployed models to respond to user prompts — which accounts for 80% to 90% of total AI energy consumption.

ChatGPT alone is estimated to process around 2.5 billion prompts per day, requiring roughly 383 GWh of electricity a year, according to the report.

The environmental footprint varies sharply by task, with a typical AI-generated image requiring about 1,450 times more energy than basic text classification, while a short AI-generated video can consume as much electricity as 200,000 spam classifications.

The report called for a responsible AI ecosystem based on transparency, efficiency by design, environmental justice, lifecycle responsibility, global cooperation, and sustainable use.