The Environmental Cost of AI: Water, Land, and Climate at Risk

6 June 2026 Climate and environment

This is almost three times the total annual energy consumption in Pakistan, Bangladesh, and Nigeria combined – countries where more than 650 million people live in total.

However, this is just the tip of the iceberg. In addition to the carbon footprint, every kilowatt-hour consumed by data centers also leaves behind a “water footprint” (water needed for cooling and energy production), as well as a “land footprint” associated with electricity generation and supply chains.

Reconsideration of approaches to environmental assessment

According to a new UN University study, by the end of the decade, water consumption related to AI could be comparable to the annual household consumption of 1.3 billion people. At the same time, its “water footprint” could exceed 14,500 square kilometers — which is about twice the area of the Jakarta metropolitan area (the largest urban agglomeration in the world).

The report emphasizes that at the moment the impact of AI on the environment is assessed inaccurately. Priority is usually given to greenhouse gas emissions, especially those associated with training large models, but this approach ignores other ecological costs.

Decisions that seem “green” in one aspect can worsen the situation in others, especially in regions already facing a resource deficit. For example, switching some renewable energy sources can reduce carbon emissions but significantly increase water consumption and land use.

The main culprit is everyday use of II

Public discussions are mainly focused on energy, which is essential for training advanced AI models. However, research shows that the everyday use of this technology accounts for approximately 80 to 90 percent of the total energy consumption.

The scale of the problem is enormous: according to estimates, one of the widely used AI services processes about 2.5 billion requests per day, consuming hundreds of gigawatt-hours of electricity annually.

Energy consumption also heavily depends on the specific task. Generating just one image using AI can require a thousand times more energy than simple text analysis, and creating videos demands even more colossal resources.

Only an increase in efficiency is unlikely to compensate for this growing demand. The report refers to the so-called “rebound effect” (Jevons effect), when cost reduction and increased productivity stimulate more active use of technology, which ultimately increases the total volume of resource consumption.

Some countries suffer more than others

The environmental consequences of the development of IT infrastructure are uneven. While the advantages of the technology have a global character, the costs often “concentrate” in specific regions.

In some countries, a significant portion of national electricity consumption is already accounted for by data centers, creating a substantial load on the power system. In others, expanding facilities actively deplete water reserves, sometimes under drought conditions.

At the same time, the authors of the report warn about the growing problem of electronic waste (e-waste): it is predicted that by 2030, the AI infrastructure will generate up to 2.5 million tons of such waste annually. The consequences of this will be especially acutely felt by countries with low income levels, where capabilities for safe disposal are limited.

The production of minerals that are critically important and indispensable for the hardware support of AI can also lead to environmental degradation in the region and mining area.

Digital and ecological inequality

The deployment of AI infrastructure also creates new disparities in the level of access to this technology and its impact on the life of societies. According to the report, more than 90 percent of specialized AI computing power is concentrated in just two countries – the USA and China. At the same time, more than 150 countries lack significant internal infrastructure to support AI.

This imbalance not only limits the economic opportunities of the population but also raises issues of environmental justice, as some countries bear environmental costs without participating in the distribution of benefits from growth stimulated by foreign direct investment.

On the path to a responsible AI

Despite the discouraging conclusions, experts emphasize that the report is not an argument against the International Institute itself. It calls for actions that will help ensure the development of technology within planetary boundaries.

The study outlines the basics of the “environmentally responsible supply chain ecosystem,” built on principles such as transparency, effectiveness at the design stage, fairness, responsibility, global cooperation, and ecological use.

Governments are recommended to integrate AI infrastructure into plans for the use of energy, water, and land resources. Companies are encouraged to design systems that minimize resource consumption. Users can also contribute by choosing applications with a lower ecological impact whenever possible.

In the final analysis, as stated in the report, the future of AI will depend not only on technological innovations but also on management decisions being made today.