AI's Environmental Cost: 945 TWh Power, 1.3B People's Water by 2030

The United Nations University Institute for Water, Environment and Health (UNU-INWEH) has released a new report, Environmental Cost of AI's Energy Use: Carbon, Water and Land Footprints, quantifying the full environmental impact of AI data centers. By 2030, global data centers powering AI are projected to consume 945 terawatt-hours (TWh) of electricity annually — nearly triple the combined electricity use of Pakistan, Bangladesh, and Nigeria (home to 650+ million people). Their water footprint will equal the basic domestic water needs of all 1.3 billion people in Sub-Saharan Africa, and land use will exceed 14,500 km², roughly twice the Jakarta metropolitan area (32 million people).

Carbon Is Not the Whole Story

The report argues that existing assessments focusing solely on carbon emissions from model training are systematically mismeasuring AI's environmental cost. Every kilowatt-hour of electricity used to train or run an AI system carries associated water and land footprints that are often overlooked. The UN scientists quantified these footprints across the world's 20 largest data center hubs, highlighting significant regional differences.

Why This Matters for Developers

For developers building AI applications, these numbers translate into real operational costs that will likely be passed down. As data centers face water and land constraints, expect higher prices for compute, stricter siting regulations, and potential pushback against data center expansion. This is not a case against AI, says Professor Kaveh Madani, Director of UNU-INWEH, but a call for responsible use and proactive mitigation of unintended impacts.

Key Findings

• Electricity: 945 TWh by 2030 (triple the usage of 650 million people's countries).
• Water: equal to the basic domestic water needs of 1.3 billion people in Sub-Saharan Africa.
• Land: over 14,500 km², more than twice the Jakarta metro area.
• Report quantifies carbon, water, and land footprints separately, not just carbon.

For developers, this underscores the importance of optimizing model efficiency — smaller models, better hardware utilization, and renewable energy sourcing — to reduce the environmental footprint of AI workloads. The report highlights a narrow window to ensure AI develops within planetary limits and that communities supplying critical minerals and hosting infrastructure also benefit.