AI could become one of the biggest water consumers on the planet.

A newly published report by the UNU Institute for Water, Environment and Health has sparked widespread discussion regarding the environmental impact of artificial intelligence (AI). The report outlines the significant energy, water, and land consumption associated with the infrastructure supporting AI technologies, primarily data centers. These facilities, which underpin much of our digital lives, are projected to become major consumers of natural resources globally.

According to the report, data centers worldwide are expected to consume approximately 448 terawatt-hours of electricity by 2025. Of this, AI workloads are anticipated to account for around 20 percent of their energy consumption. If considered as a country, data centers would rank as the eleventh-largest electricity consumer in the world. By 2030, this figure is predicted to double, with AI workloads making up 40 percent of the total energy demand. As a standalone entity, data centers would then be the sixth-largest electricity consumer globally, using over 945 terawatt-hours annually.

The report further estimates that the water footprint of data centers will reach 9.3 trillion liters by 2030. This amount of water is sufficient to meet the annual drinking needs of all 1.3 billion residents in Sub-Saharan Africa. Additionally, the land footprint associated with energy production and supply chains is forecasted to surpass 14,500 square kilometers—approximately twice the size of the Jakarta metropolitan area, where over 32 million people reside.

Carbon dioxide emissions from data centers are projected to reach 189 million tons in 2025 and are expected to rise to 399 million tons by 2030. Notably, the majority of this energy demand does not stem from training AI models such as ChatGPT, Claude, and DeepSeek, but rather from the daily operations involving billions of user queries.

Despite the significance of these findings, several experts have raised concerns about the methodology and data used in the report. They argue that the report lacks depth, is difficult to verify, and fails to provide adequate comparisons with other sectors. Some critics point out that the report relies on outdated data and does not sufficiently address the role of renewable energy sources such as photovoltaics, which have shown strong performance in recent years.

The report's projections regarding the doubling of carbon dioxide emissions by 2030 are based on data from the International Energy Agency, according to some experts. However, they caution against extrapolating results from academic case studies to estimate the energy consumption of AI-generated content on a massive scale. Furthermore, the lack of transparency from major tech companies regarding their energy usage complicates efforts to assess the true environmental impact of AI.

Hydrologists have also expressed skepticism about the water consumption calculations presented in the report. They note that the sources of the data used to estimate the water footprint are unclear, and there is no distinction made between consumed and used water. This ambiguity raises questions about the accuracy and reliability of the report's conclusions regarding water usage.

As AI becomes increasingly integrated into everyday life, the environmental consequences of its rapid growth are becoming more apparent. The report serves as a wake-up call, highlighting the urgent need for sustainable practices in the development and deployment of AI technologies. It underscores the importance of addressing the environmental challenges posed by AI infrastructure to ensure that technological progress does not come at the expense of ecological integrity.