The United Nations University Institute for Water, Environment and Health (UNU-INWEH) published a study on June 3 calling on major AI companies to disclose the full environmental footprint of their operations, including energy consumption, carbon output, water usage, and land demands. The report also urges governments to enforce standardized reporting and push the industry toward renewable energy adoption.

Global data center electricity consumption was estimated at 448 TWh in 2025. By 2030, the report projects that figure will climb to 945 TWh. That’s more than double, and it’s roughly triple the combined annual energy consumption of Pakistan, Bangladesh, and Nigeria.

Water tells an equally sobering story. The projected water footprint of AI-related activities could hit 9.32 trillion liters by 2030. In English: that’s enough to cover the annual domestic water needs of 1.3 billion people in Sub-Saharan Africa.

Then there’s the land. The physical footprint of AI data centers is projected to exceed 14,500 square kilometers by the end of the decade. For context, that’s larger than the entire country of Montenegro.

One of the report’s more nuanced findings is that reducing carbon emissions alone won’t necessarily lower water or land usage. The relationship between these impacts depends heavily on local energy sources. A data center powered by hydroelectric might look clean on carbon metrics but could still be guzzling water at an unsustainable rate.

The UN report doesn’t just diagnose the problem. It prescribes a framework built around six principles for responsible AI development: transparency, efficiency by design, equity and environmental justice, lifecycle responsibility, global cooperation, and sustainable use.

The equity dimension is particularly pointed. The report emphasizes that AI’s environmental burden falls disproportionately on communities that benefit least from the technology. The UN frames this as an environmental justice issue, not merely a technical one.

Efficiency by design means baking resource consciousness into AI models from the start, not bolting it on after deployment. Lifecycle responsibility extends that thinking across the full supply chain, from chip manufacturing to server decommissioning. And global cooperation acknowledges that data centers don’t respect national borders; a company headquartered in San Francisco might run servers that strain water tables in rural Ireland.

The European Union has already been moving toward mandatory sustainability reporting for tech companies. This report will likely strengthen that push and inspire similar efforts elsewhere.