Key facts
- AI data centers could consume water equal to 1.3 billion people's needs by 2030.
- Global data centers powering AI are projected to consume 945 TWh of electricity by 2030.
- AI infrastructure could generate 2.5 million metric tons of e-waste annually by 2030.
- Focusing solely on carbon emissions misrepresents AI's environmental cost.
- Large data centers can consume up to 5 million gallons of water daily for cooling.
A new report from the United Nations University Institute for Water, Environment and Health (UNU-INWEH) warns that by 2030, the water used by artificial intelligence could equal the needs of 1.3 billion people, potentially threatening natural resources globally. The report quantifies the carbon, water, and land footprints associated with AI's electricity consumption. Researchers found that focusing solely on carbon emissions misrepresents AI's environmental cost, as cooling and generating power for data centers have significant water footprints, and the infrastructure itself has a land footprint. By 2030, global data centers powering AI are projected to consume 945 terawatt-hours of electricity, nearly triple the combined annual use of Pakistan, Bangladesh, and Nigeria. The water footprint is expected to match the basic domestic needs of all 1.3 billion people in Sub-Saharan Africa for a year, while the land footprint could exceed 5,590 square miles. The report also highlights that choices appearing green from a carbon perspective, such as switching to bioenergy, can significantly increase water and land footprints. In 2025, data centers consumed an estimated 448 terawatt-hours of electricity, more than Saudi Arabia. In Ireland, data centers accounted for 21% of metered electricity in 2023, leading to a pause on new approvals around Dublin. Large data centers can consume up to 5 million gallons of water daily for cooling, straining resources in water-scarce regions like Querétaro, Mexico, and impacting communities in Uruguay. The report also points to a widening digital divide, with 90% of AI data center capacity concentrated in the U.S. and China, and warns that AI infrastructure could generate 2.5 million metric tons of electronic waste annually by 2030. The UNU-INWEH calls for a 'responsible AI ecosystem' that considers water and land use in permitting and environmental assessments, urging governments and financial institutions to implement guardrails to minimize environmental consequences and ensure equitable development.