UN Report: AI's Water Footprint Will Match the Needs of 1.3 Billion People by 2030 as Environmental Costs Extend Far Beyond Carbon
A new report from the United Nations University has put numbers on something the AI industry rarely discusses: what the data centers powering AI actually cost the planet in water, land, and carbon - not just in electricity. The findings reframe the environmental debate around AI infrastructure in ways that will matter to businesses building sustainability commitments around their AI usage.
By 2030, the global data centers powering artificial intelligence are projected to consume 945 terawatt-hours of electricity - nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, countries collectively home to more than 650 million people. Their associated water footprint will equal the basic annual domestic water needs of all 1.3 billion people in Sub-Saharan Africa, and their land footprint will exceed 14,500 square kilometres - roughly twice the size of the Jakarta metropolitan area. Snaptrude
These are not distant projections. The infrastructure driving these numbers is being built right now, at the scale described in recent weeks by Alphabet's $84.75 billion equity raise and SpaceX's $30 billion compute lease with Google.
Why Carbon Is Not the Full Picture
Researchers have previously warned about the greenhouse gas emissions of data centers. But the UN scientists now argue that the environmental costs of AI and data centers cannot be understood through carbon emissions alone. The report quantifies the carbon, water, and land footprints of AI's electricity use across the globe and highlights the big differences between these footprints in the world's 20 largest data center hubs. Snaptrude
The counterintuitive finding is that low-carbon electricity is not automatically low-impact electricity. Some energy sources that may seem environmentally friendly from a carbon perspective can have significantly larger water and land footprints. Hydroelectric power, for example, has a very low carbon footprint but can have a very high water footprint depending on reservoir evaporation rates. businesswire
This means companies claiming carbon-neutral AI operations may still be imposing significant water and land costs - just in places where those costs are harder to see or measure.
The Scale of Individual Interactions
The water usage of a typical text prompt to a ChatGPT-type model is around 29 millilitres - similar to two tablespoons of water. A standard-resolution AI-generated image carries an estimated footprint of 2.9 watt-hours of electricity, 1.22 grams of CO₂ equivalent, 28.6 millilitres of water, and 0.45 square centimetres of land. businesswire
These per-interaction numbers seem small. They become significant at the scale at which generative AI is now being used - hundreds of millions of queries per day across platforms like ChatGPT, Claude, Gemini, and Perplexity.
By 2030, the water footprint linked to electricity generation and cooling for AI infrastructure could reach an estimated 9.3 trillion litres - enough to meet the drinking water requirements of the global population for approximately 1.6 years. Wikipedia
The Justice Dimension
The report frames this not just as an environmental challenge but as a governance and equity problem. The benefits of AI flow primarily to the countries and companies building and deploying it. The environmental costs are frequently concentrated in the regions where data centers are sited and where power is generated.
The report frames AI's environmental footprint as a governance and justice challenge, not only a technical problem. The benefits of AI often flow across borders and sectors while the environmental burdens of data center siting, electricity demand, water withdrawals, land use, mineral extraction, and e-waste can be concentrated in specific communities and regions. Snaptrude
The report calls for mandatory reporting of carbon, water, and land impacts for all AI systems and data centers, and recommends integrating forecasted AI demand into climate, energy, water, and land-use planning at the national level.
What This Means for Business Leaders
From four years advising executives on AI for business strategy, I have watched environmental considerations move from afterthoughts to board-level questions. This report accelerates that trend.
Companies building AI products or deploying AI at scale need to get ahead of disclosure requirements that are coming. The EU AI Act, Canada's AI for All strategy, and multiple national regulatory frameworks are moving toward mandatory environmental reporting for AI systems. The companies that have already built measurement infrastructure for carbon, water, and land impact will be positioned better than those doing it under regulatory deadline pressure.
The practical first step is model selection. Smaller, more efficient models running the same tasks as larger ones can produce dramatically lower footprints. AI coding tools, image generation platforms, and customer service tools all have significant variation in their energy and water intensity per task. Choosing deliberately - rather than defaulting to the most capable model for every use case - is where the most accessible environmental gains exist.
Cut Through the Noise
What are AI's projected environmental costs by 2030 according to the UN report? A June 2026 report by the UN University Institute for Water, Environment and Health projects that by 2030, global AI data centers will consume 945 terawatt-hours of electricity annually - nearly triple the combined electricity use of Pakistan, Bangladesh, and Nigeria. Their water footprint will equal the basic domestic water needs of all 1.3 billion people in Sub-Saharan Africa, and their land footprint will exceed 14,500 square kilometres.
How much water does a single AI prompt use? A typical text prompt to a ChatGPT-type AI model uses approximately 29 millilitres of water - roughly two tablespoons. A single standard-resolution AI-generated image carries a footprint of 2.9 watt-hours of electricity, 1.22 grams of CO₂ equivalent, 28.6 millilitres of water, and 0.45 square centimetres of land. These figures become significant at the scale of hundreds of millions of daily AI interactions globally.
Why does the UN say carbon emissions alone don't capture AI's environmental impact? Low-carbon electricity sources like hydroelectric power can have significantly higher water and land footprints than fossil fuel sources depending on reservoir evaporation and land occupation. This means companies claiming carbon-neutral AI operations may still impose significant water and land costs in the regions where their data centers operate. The UN report calls for mandatory three-dimensional reporting - carbon, water, and land - rather than carbon-only metrics.
What does the UN recommend to reduce AI's environmental footprint? The UN University report calls for mandatory reporting of carbon, water, and land impacts for all AI systems and data centers, integration of AI infrastructure demand into national energy, water, and land-use planning, and a "responsible AI" framework incorporating transparency, environmental justice, lifecycle responsibility, and sustainable use. It also recommends AI system design that prioritizes efficiency alongside capability, and international cooperation on environmental standards for AI infrastructure.
