Overview
- The U.N. University report, published Wednesday, found global data centres used about 448 terawatt-hours of electricity and 4.5 trillion litres of water in 2025 and projects roughly 945 TWh and 9.3 trillion litres by 2030 as AI workloads grow to about 40 percent of demand.
- The authors call for standardised, facility-level disclosure of energy, water and land footprints so governments can fold AI demand into energy, water and land-use planning and permit decisions.
- Major tech firms have offered public fixes: Google announced water‑replenishment commitments and local projects, and Microsoft described a new low‑water 'Fairwater' design it says greatly reduces ongoing cooling withdrawals.
- The report highlights trade-offs in mitigation: some low‑carbon options can greatly raise water or land footprints, so judging sustainability by carbon alone can shift burdens to water-stressed or land-limited places.
- Local communities and regulators are pressing for tighter reviews or moratoria on new builds, and the report warns that unchecked growth could deepen digital and environmental injustices in places hosting most AI infrastructure.