The United Nations University Institute for Water, Environment and Health (UNU-INWEH) argues that the environmental impact of artificial intelligence is far greater than many people realize. While public discussions often focus on AI’s carbon emissions, the report highlights that AI also has significant water and land footprints due to the enormous infrastructure required to power data centers, cooling systems, electricity generation, and hardware supply chains. Researchers say evaluating AI through carbon emissions alone provides an incomplete picture of its environmental cost.
According to the study, global data centers consumed about 448 terawatt-hours of electricity in 2025—more than the annual electricity use of most countries—and generated roughly 208 million tons of carbon dioxide emissions. The infrastructure also required approximately 1.2 trillion gallons of water, both for cooling equipment and producing electricity. If current trends continue, the report projects that data centers could account for nearly 3% of global electricity consumption by 2030, with environmental impacts rising sharply as AI adoption expands.
The researchers emphasize that AI's environmental burden varies depending on how electricity is generated. A reduction in carbon emissions through cleaner energy sources does not automatically reduce water or land use. In some cases, energy sources with lower carbon footprints may require significantly more land or water resources. The report argues that policymakers and companies need to consider these trade-offs rather than focusing on a single sustainability metric.
Another notable finding is that AI’s environmental impact is increasingly driven by everyday use rather than model training alone. The report estimates that inference—the process of generating responses to user requests—now accounts for the majority of AI-related energy consumption. Activities such as generating images and videos can consume dramatically more energy than basic AI tasks, reflecting the growing resource demands of advanced generative AI systems.
UNU-INWEH researchers stress that the report is not an argument against AI. Instead, they call for greater transparency, standardized environmental reporting, improved efficiency, and responsible deployment practices. Their message is that AI can deliver significant societal benefits, but its long-term sustainability will depend on recognizing and managing its carbon, water, and land footprints before they grow beyond manageable levels.
