UN Scientists Warn AI Data Centers Will Drink More Water Than 1.3 Billion People by 2030

News Summary

A landmark report published on June 3, 2026 by the United Nations University Institute for Water, Environment and Health (UNU-INWEH) reveals that the rapid expansion of artificial intelligence infrastructure is placing unprecedented pressure on the planet's most critical natural resources — water, land, and the atmosphere.

The Water Footprint of AI: A Global-Scale Reckoning

By 2030, data centers powering AI workloads worldwide are projected to consume 9.3 trillion liters of water annually — more than double the roughly 4.5 trillion liters used in 2024. To put that figure in perspective, UNU-INWEH researchers note that 9.3 trillion liters would satisfy the total drinking-water needs of every person on Earth (8.1 billion people) for approximately 1.6 years. The report also compares the projected figure to the basic domestic water needs of the 1.3 billion people living in Sub-Saharan Africa for an entire year.

Water is used in data centers primarily for cooling: chilled water circulates through server halls, and evaporative cooling towers release vast quantities into the atmosphere. As AI model training and inference workloads grow exponentially, so does the cooling demand. The report notes that in many high-growth deployment regions, this water is drawn from the same aquifers and river systems that support agriculture and local drinking supplies.

Electricity Demand and Carbon Emissions

The same analysis projects that global data center electricity consumption will nearly double by 2030, reaching 945 terawatt-hours (TWh) per year. By that point, AI workloads alone are expected to account for roughly 40% of all data center power demand — up sharply from today's levels. Generating that volume of electricity is estimated to produce approximately 399 million tonnes of carbon dioxide equivalent annually, comparable to the total annual emissions of a major industrialized nation.

Land Use and Electronic Waste

Beyond energy and water, the UNU-INWEH report highlights two additional resource pressures that receive less public attention. First, the physical footprint of data center campuses — including the facilities themselves, surrounding buffer zones, and associated energy infrastructure — is projected to exceed 14,500 square kilometers by 2030. Second, the rapid hardware refresh cycles driven by AI competition could generate up to 2.5 million tonnes of electronic waste per year by 2030, much of which is currently processed in low-income economies with limited environmental safeguards.

Geographic Concentration and Community Impact

One of the report's most pointed findings concerns where data centers are being built. Dr. Mir Matin, Manager of UNU-INWEH's Geospatial, Climate and Infrastructure Analytics Programme, emphasized that facilities are frequently sited in regions already facing significant water stress. The communities living near these installations often bear the environmental costs — competing for scarce water resources — while deriving little direct benefit from the AI services the centers support.

Efficiency Gains and the Rebound Effect

Researchers acknowledge that hardware efficiency has improved substantially over the past decade. Modern AI accelerators and liquid-cooling technologies consume meaningfully less energy and water per computation than their predecessors. However, the report cautions that efficiency gains have been consistently outpaced by the sheer growth in AI deployment — a dynamic sometimes called the "rebound effect." As individual operations become cheaper, total demand scales faster than savings accumulate.

Report Background and Scope

The UNU-INWEH report, released at 9:00 AM Eastern Time on June 3, 2026, synthesizes data from academic literature, industry disclosures, and the institute's own geospatial modeling. It covers the operational footprints of data centers globally, with particular attention to regions in Asia, the Middle East, and parts of Africa where AI infrastructure expansion is accelerating. The institute is part of the broader United Nations University system and focuses on water security, environmental health, and sustainable resource management.

Outlook and Recommendations

The report stops short of calling for restrictions on AI development, instead urging policymakers, technology operators, and investors to integrate resource-impact assessments into data center planning from the outset. Key recommendations include mandatory water-stress disclosures for new facilities, stronger incentives for closed-loop cooling systems that minimize consumptive water use, and greater transparency in corporate environmental reporting for AI infrastructure.

As AI capabilities continue to advance and adoption spreads across industries, the physical resource requirements underpinning that growth are coming into sharper focus. The UNU-INWEH findings add authoritative scientific weight to a conversation that has largely centered on algorithmic progress — redirecting attention to the rivers, aquifers, and power grids that make it all possible.