AI's environmental impact extends far beyond carbon emissions, threatening water supplies and land resources at a scale that rivals the basic needs of entire populations. A new study from UN University (UNU) warns that by 2030, AI-related water consumption could equal the annual domestic water needs of 1.3 billion people, while its land footprint may exceed 14,500 square kilometers, roughly twice the size of the Jakarta metropolitan area.

Why Are Water and Land Being Overlooked in AI's Environmental Debate?

Public conversation about AI's environmental toll has focused heavily on greenhouse gas emissions, particularly those tied to training large language models. However, this narrow focus masks a broader environmental crisis. The UNU report highlights a critical gap: solutions that appear "green" in one dimension often worsen pressures in others, particularly in regions already facing resource scarcity. For example, switching to certain renewable energy sources may reduce carbon emissions but can significantly increase water consumption and land use.

The problem is compounded by how we measure AI's impact. Data centers require electricity not just for computation but also for cooling systems and energy production. Each unit of electricity carries what researchers call a "water footprint" for cooling and energy generation, plus a "land footprint" associated with power generation and supply chains. This multi-dimensional environmental burden is rarely discussed in mainstream coverage.

What's Driving the Massive Energy Demand?

Most people assume that training advanced AI models consumes the bulk of energy. In reality, the opposite is true. Day-to-day usage of AI systems accounts for roughly 80 to 90 percent of total energy demand, according to the UNU study. One widely used AI service is estimated to process around 2.5 billion prompts per day, consuming hundreds of gigawatt-hours of electricity each year.

Energy consumption varies dramatically depending on the task. Generating a single AI image can require more than a thousand times the energy of simple text classification, while video generation demands even greater resources. This variation means that as AI applications become more sophisticated and widespread, energy demands will continue climbing.

Data centers globally could consume 945 terawatt-hours of electricity annually by 2030, nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria, countries collectively home to more than 650 million people. Efficiency improvements alone are unlikely to offset these rising demands. The report points to the "rebound effect," in which lower costs and improved performance drive higher usage, ultimately increasing total resource consumption rather than reducing it.

How Can Governments and Companies Build More Responsible AI?

• Transparency Requirements: Governments should mandate that companies disclose their full environmental footprint across water, land, and carbon metrics, not just greenhouse gas emissions alone.
• Infrastructure Planning: Integrate AI infrastructure into national energy, water, and land-use planning to prevent data centers from straining already-stressed resources in vulnerable regions.
• Efficiency by Design: Companies should prioritize systems that minimize resource consumption from the outset, rather than treating efficiency as an afterthought.
• Lifecycle Responsibility: Address the growing electronic waste challenge, with AI infrastructure projected to generate up to 2.5 million tonnes of e-waste annually by 2030, much of which will burden lower-income countries with limited disposal capacity.
• Global Cooperation: Establish international standards and governance frameworks to ensure that environmental costs are not simply exported to regions with weaker regulations.
• Sustainable Use Practices: Users should choose lower-impact AI applications where possible, avoiding unnecessary high-energy tasks like image or video generation when simpler alternatives exist.

The UNU researchers stress that their report is not an argument against AI itself. Rather, it calls for urgent action to ensure that the technology develops within planetary limits. The study outlines a framework for a "responsible AI ecosystem" built on principles including transparency, efficiency by design, equity, lifecycle responsibility, global cooperation, and sustainable use.

Who Bears the Environmental Cost While Others Reap the Benefits?

The environmental impacts of AI infrastructure are not evenly distributed globally. While the benefits of the technology are global, its costs are often concentrated in specific regions. In some countries, data centers already account for a significant share of national electricity consumption, placing pressure on energy systems. In others, expanding facilities are drawing heavily on water supplies, sometimes amid drought conditions.

The concentration of AI computing power amplifies this disparity. More than 90 percent of AI-specialized computing capacity is concentrated in just two countries, the United States and China. At the same time, over 150 nations lack significant domestic AI infrastructure. This imbalance not only limits economic opportunities but also raises questions of environmental justice, as some countries bear the environmental costs without sharing in the benefits of AI-driven growth.

The production of critical minerals needed for AI hardware also raises concerns about environmental degradation and social inequities in extraction regions. As demand for AI infrastructure grows, so too does the pressure on mining operations in developing nations, often with minimal environmental oversight.

The path forward requires more than technological innovation. It demands governance choices that prioritize planetary limits alongside AI advancement, ensuring that the technology's benefits are shared equitably while its environmental burdens are not concentrated on the world's most vulnerable populations.