Artificial intelligence has transformed how we work, communicate, and solve complex problems, but its environmental cost is emerging as one of the most pressing concerns of our time. A groundbreaking United Nation report released on June 4, 2026, has sounded the alarm on AI’s staggering water consumption, energy demands, and land use requirements. The report, conducted by the United Nations University Institute for Water, Environment and Health, represents the most comprehensive assessment yet of AI’s environmental costs and describes the findings as alarming. The scale of AI’s resource consumption is difficult to comprehend. By 2025, data centers consumed 9.3 trillion liters of water—enough to provide drinking water for 8.1 billion people for more than 1.5 years. This figure is not merely a statistic; it represents a fundamental challenge to global water security as AI adoption accelerates worldwide. This Analysis reveals that the environmental footprint of artificial intelligence extends far beyond the commonly discussed carbon emissions, encompassing a water crisis that threatens to compete directly with human consumption needs in already water-scarce regions.
Energy Consumption on an Unprecedented Scale
The electricity requirements of AI infrastructure are growing at an exponential rate that threatens to reshape global energy systems. In 2025, data centers consumed 448 TWh of electricity—equivalent to France’s entire annual consumption. By 2030, this figure is projected to reach 945 TWh, which equals sub-Saharan Africa’s electricity needs for more than five years. AI’s share of global electricity consumption tells an even more striking story. Currently at 20% of data center electricity use, AI is projected to account for 40% by 2030, translating to 374 TWh for AI alone. This represents nearly 3% of global electricity consumption, a massive share for a technology that has existed in its current form for only a few years. The physical footprint required to generate this electricity is equally staggering. By 2030, the land needed to produce electricity for AI data centers will span 14,000 km², roughly the size of Northern Ireland. This land requirement creates competition with agriculture, conservation, and human settlement, adding another dimension to AI’s environmental conflict.
The Carbon Emissions Crisis
AI’s environmental impact extends far beyond water and electricity. The carbon emissions from AI infrastructure are projected to match the UK’s entire 2025 emissions by 2030. This timeline is particularly concerning because it represents just a few years of projected growth in AI adoption. The training of a single large AI model exemplifies the magnitude of this problem. Training one large model requires 100 GWh of electricity, equivalent to the annual residential power needs of 770,000 people in sub-Saharan Africa. The land footprint for this single model training operation is approximately 215 football fields, illustrating the spatial intensity of AI development. The hardware lifecycle of AI systems creates another environmental crisis. By 2030, AI is projected to generate electronic waste equivalent to throwing out 250 Eiffel Towers every year. This e-waste contains toxic materials, valuable rare earth metals, and represents a massive failure of resource efficiency in the technology sector.
Water Consumption: The Hidden Environmental Cost
Understanding AI’s water consumption requires examining three distinct categories. Scope 1 cooling accounts for 1–9 liters per kWh, representing the direct water used to cool data center equipment. Scope 2 electricity generation involves 7.6 liters evaporated per kWh, capturing the water consumed during electricity production. Scope 3 supply chain requires 8–10 liters per microchip, encompassing the water embedded in manufacturing AI hardware. This multi-layered approach to water usage demonstrates why simple solutions cannot address the problem. Each scope represents a different stage in the AI lifecycle, requiring different interventions and regulatory frameworks. The water cost of individual AI interactions is surprisingly significant. Each AI query uses approximately 1/15 teaspoon (0.067 ml) of water. As Sam Altman (OpenAI CEO) admitted, a single ChatGPT interaction uses “1/15 of a teaspoon” of water. While this seems negligible, the scale of usage transforms it into a massive global demand. Just 10–50 queries consume approximately 500ml, equivalent to one water bottle. The training of advanced AI models consumes water at an industrial scale. Training a single ChatGPT-5 model requires 1 billion liters of water, enough to supply a small city for months. This figure alone demonstrates why AI’s environmental footprint cannot be dismissed as trivial. The United Nation report emphasizes that these costs accumulate rapidly as AI adoption spreads across industries and geographic regions. The trajectory of AI water consumption points toward a potential crisis. By 2027, global AI water use is projected to reach 4.2–6.6 billion m³, which is 4–6 times Denmark’s annual water consumption. Research from the University of California, Riverside calculated that AI’s total water demand by 2027 could exceed half the UK’s total annual water withdrawal. This Analysis of water demand projections reveals a troubling trend: AI’s water consumption is growing faster than most experts predicted just five years ago, suggesting that current infrastructure planning may be fundamentally inadequate.
Industry Responses and Corporate Promises
Major technology companies have responded to growing concerns with ambitious promises. Google, Meta, and Microsoft all pledged water neutrality by 2030, according to BBC World Service reporting. However, a critical gap exists between these promises and transparency. None disclose how much water use is specifically due to AI, creating a significant accountability problem. The measurement gap is even more alarming. Up to 50% of data centers don’t measure water usage, meaning companies cannot accurately report their environmental impact even if they wanted to. This lack of data infrastructure undermines the credibility of corporate sustainability claims and makes regulatory oversight nearly impossible. Microsoft provides a revealing case study in the tension between sustainability commitments and AI growth. Brad Smith (Microsoft President), in 2020, presented an ambitious initiative aimed at reducing water consumption across the company’s expanding network of data centers. He committed to minimizing usage, aiding wetland restoration efforts, and implementing innovative water-saving technologies. Smith stated unequivocally that Water is crucial for life. However, the company’s updated projections contradict this vision. Microsoft now expects water use to reach 18 billion liters by 2030, a 150% increase from 2020’s 7 billion liters. This projection suggests that even the most water-conscious tech giant cannot reconcile AI growth with reduced consumption, raising questions about whether water neutrality is achievable at all.
Geographic Distribution and Water Stress
The location of new data centers exacerbates the water crisis. A Bloomberg News analysis found that roughly two thirds of new data centers built or in development in the US since 2022 are in places with high levels of water stress. This geographic concentration means AI infrastructure is competing directly with agriculture and human consumption in already water-scarce regions. The pollution risk is equally concerning. 55% of global data centers are in river basins with high water pollution risk, according to the UK Government’s Government Digital Sustainability Alliance. This dual threat of consumption and pollution creates compounding environmental damage that affects entire ecosystems and communities. Data centers are increasingly located in the global south, where water resources are often more vulnerable. This geographic shift means that regions with less infrastructure and fewer resources to manage water scarcity bear the brunt of AI’s environmental impact. The climate justice implications are significant, as the global north benefits from AI while the global south faces the environmental consequences.
Systemic Water Scarcity Context
AI’s water consumption occurs against a backdrop of worsening global water scarcity. The World Economic Forum projects that the world faces a 56% freshwater deficit by 2030, impacting businesses and individuals alike. This deficit means that AI’s growing water demands will compete with fundamental human needs. The UK Government Digital Sustainability Alliance provides additional context. Freshwater demand is expected to exceed supply by 40% by 2030, creating a structural mismatch between supply and demand that AI will exacerbate. The alliance calculated that AI is predicted to increase global water usage from 1.1bn to 6.6bn m³ by 2027, equivalent to more than half the UK’s total water usage. This Analysis of global water scarcity demonstrates that AI’s environmental impact cannot be understood in isolation. It must be viewed within the broader context of climate change, population growth, and changing consumption patterns that are already straining freshwater resources worldwide. The United Nation report makes clear that without intervention, AI will accelerate existing water crises rather than create new ones.
Proposed Solutions and Path Forward
The first step toward addressing AI’s environmental impact is comprehensive measurement and reporting. Researchers are calling for increased transparency in spatial and temporal water usage reporting, which would allow stakeholders to understand where and when water is consumed. Without this data, meaningful regulation and corporate accountability remain impossible. The current voluntary framework has failed to produce adequate results. Tech industry silence on AI water use sparks global transparency demands, suggesting that voluntary commitments alone cannot address the scale of the problem. Governments must establish binding standards that require comprehensive measurement and public reporting. The technology sector is exploring several technical solutions. Waterless cooling technologies are currently in trial, which could eliminate Scope 1 water consumption entirely. Heat reuse for warming homes represents a creative approach to capturing waste energy, though this addresses energy rather than water directly. Some researchers propose relocating data centers to areas with abundant water resources. Undersea and Arctic data center locations have been suggested, though these solutions introduce new engineering challenges and environmental risks. The fundamental question remains whether relocation solves the problem or merely shifts the environmental burden to different ecosystems. The University of Illinois College of Engineering emphasizes that comprehensive sustainable AI approaches must balance carbon and water efficiency. This holistic perspective recognizes that solving one environmental problem while exacerbating another merely relocates the crisis. True sustainability requires addressing all three scopes of water usage while also reducing carbon emissions and electronic waste.
Regulatory and Policy Implications
The United Nation report’s findings demand urgent policy responses. Governments must establish binding water usage standards for data centers, requiring comprehensive measurement and public reporting. International cooperation is essential because AI infrastructure operates across borders while water resources are local. The report warns that AI could drain more water than entire nations, making this a global governance challenge that transcends national boundaries. Without coordinated international action, countries will compete to attract data centers while externalizing environmental costs to neighboring regions.
The environmental cost of AI is no longer theoretical—it is measurable, quantifiable, and accelerating. With water consumption reaching 9.3 trillion liters annually, electricity demands approaching 945 TWh by 2030, and carbon emissions matching entire nations, AI’s growth trajectory is fundamentally incompatible with current environmental limits. This Analysis of the United Nation report reveals a technology at a crossroads. The choice is not whether to develop AI, but how to develop it sustainably. Without urgent intervention, the technology promised to solve humanity’s greatest challenges may instead exacerbate the very crises it was meant to address. The window for action is closing rapidly, and the next few years will determine whether AI becomes a tool for sustainability or a driver of environmental collapse.
