Rising Emissions, Depleting Water and Vanishing Land—UN Scientists: AI Is Threatening Natural Resources for Billions
https://unu.edu/inweh/news/environmental-cost-of-AIs-Enrgy-use-carbon-water-and-land-footprints
Publish Date: 2026-06-03 10:04:00
Source Domain: unu.edu
-
Mismeasurement of AI Environmental Costs: Current assessments focus mainly on carbon emissions from AI training, overlooking water and land footprints from cooling, power generation, and infrastructure.
-
Data Centre Energy Consumption: Global data centres consume a significant amount of electricity, water, and land, projected to reach 945 TWh, 9.3 trillion litres of water, and 14,500 km² by 2030.
-
Dominance of Inference: Once deployed, running models to answer user prompts consume 80-90% of total AI energy use, rather than training models.
-
Model Efficiency & Rebound Effect: Efficiency gains in AI model usage may increase overall energy consumption unless usage volume is controlled.
-
Geographical Inequality: AI infrastructure is largely concentrated in two countries, creating environmental and governance divides, impacting countries without access to AI compute.
-
Global Digital Divide: Over 90% of specialized AI data centers are in just two countries; less than half the world has sovereign AI compute infrastructure, leading to uneven benefits and burdens.
-
Sustainability Challenges: AI infrastructure is projected to create significant electronic waste and relies on minerals from regions with weaker environmental oversight.
-
Call for a Sustainable AI Ecosystem: The report proposes six principles for a responsible AI ecosystem to ensure transparency, efficiency, equity, global cooperation, and sustainability.
