A new report from the United Nations University finds that AI data centers could consume 945 terawatt-hours of electricity per year by 2030, nearly triple the combined annual use of Pakistan, Bangladesh, and Nigeria. The report warns that focusing only on carbon emissions misses the bigger picture. Switching to bioenergy can cut carbon by 70% but increase water use more than thirtyfold and land use a hundredfold. Most AI energy goes to inference, not training, and efficiency gains often lead to more consumption, not less. The report maps where data centers are built against local water stress and grid capacity. In Ireland, data centers used 21% of metered electricity in 2023. In Mexico and Uruguay, new facilities compete with drought-stricken communities for water. More than 90% of AI data center capacity sits in the US and China, while over 150 countries bear costs from mineral extraction and e-waste. The authors say the issue is governance, not technology, and call for mandatory water and land footprint disclosures alongside carbon reporting.
