UN Report Exposes the Massive Water, Energy, and Land Footprint of AI Data Centers

Artificial intelligence is often discussed as a purely digital phenomenon—a cloud of algorithms and neural networks. But a new comprehensive report from the United Nations University (UNU) reframes artificial intelligence as a massive physical infrastructure system.

The 56-page report, released in June 2026, quantifies the carbon, water, and land footprints of the data centers that power AI. It marks a shift in how global institutions view the technology, moving beyond just carbon emissions to a multi-dimensional accounting of planetary resources.

In response to these findings, UN Secretary-General António Guterres launched the AI Environmental Transparency Initiative at London Climate Action Week. The initiative calls on major technology companies to publicly disclose their resource consumption and commit to powering all data centers with renewable energy by 2030.

To understand the scale of the challenge, it helps to look at the electricity required to train and operate large language models. In 2025, global data centers consumed an estimated 448 terawatt-hours (TWh) of electricity.[2]

If data centers were a country, that consumption would rank them 11th globally. By 2030, as AI adoption accelerates, that figure is projected to reach 945 TWh—nearly triple the combined annual electricity use of Pakistan, Bangladesh, and Nigeria.

AI's share of this energy pie is growing rapidly. While AI accounted for roughly 20% of data center electricity use last year, the UNU report projects that share will double to 40% by the end of the decade.[2]

This electricity demand translates directly into a significant carbon footprint. In 2025, data centers produced approximately 189 million tonnes of carbon dioxide equivalent.

But carbon is only one part of the equation. The report highlights a critical, often-overlooked metric: the water footprint. Data centers require vast amounts of water for cooling systems to prevent servers from overheating, as well as for the generation of the electricity that powers them.[1]

By 2030, AI-related water consumption could hit 9.3 trillion liters. To put that in perspective, this volume is equivalent to the basic annual domestic water needs of 1.3 billion people.

The physical space required for this infrastructure is equally vast. The land footprint associated with data centers and their supply chains is projected to exceed 14,500 square kilometers by 2030—an area roughly twice the size of the Jakarta metropolitan area.

The UNU analysis emphasizes that these environmental metrics are deeply interconnected, and optimizing for one can sometimes worsen another. This is the core challenge of sustainable computing design.

For example, transitioning a data center to run entirely on hydropower significantly reduces its carbon emissions. However, hydropower has a massive water and land footprint due to reservoir evaporation and dam construction, illustrating the complex trade-offs involved in siting these facilities.

The hardware lifecycle presents another physical limit. As companies race to deploy the latest, most efficient specialized chips, older servers are decommissioned. The UN report warns that AI infrastructure could generate up to 2.5 million tons of electronic waste annually by 2030.

The environmental burdens of this infrastructure are often localized, while the benefits of AI are distributed globally. Data centers can strain local power grids and draw heavily on municipal water supplies, sometimes in regions already facing drought conditions.

Recognizing these pressures, the tech industry is actively developing solutions. Hardware manufacturers are designing specialized AI accelerators that perform more calculations per watt of energy, while software engineers are optimizing models to require less computational power for everyday queries.[1]

Cooling technologies are also evolving. Many new facilities are shifting from traditional air conditioning to advanced liquid cooling systems, which are more efficient, or locating data centers in colder climates where outside air can naturally cool the servers.[1]

On the energy front, leading AI companies are moving beyond purchasing annual carbon offsets. Instead, they are pioneering "24/7 carbon-free energy" matching, investing directly in local wind, solar, and geothermal projects to ensure their facilities are powered by clean energy around the clock.

The UN's new transparency initiative aims to accelerate these industry efforts by establishing standardized reporting metrics. By tracking carbon, water, and land footprints together, policymakers and companies can make more informed decisions about where and how to build digital infrastructure.

Ultimately, the goal is not to halt the development of AI, which the UN acknowledges has immense potential to solve global challenges—including climate change itself through better weather prediction, grid optimization, and materials science.[1]

Instead, the focus is on building a sustainable ecosystem. By treating AI as a physical resource system and designing it within planetary limits, the industry can ensure that the technological revolution of this era does not come at the expense of the environment.[1]