There has been a fast growth of data centers across the globe due to artificial intelligence. Although most of the discussions surrounding AI have revolved around carbon dioxide emissions, a recent study conducted by the United Nations University Institute for Water, Environment and Health (UNU-INWEH) reveals that the consequences of the application of the technology extend beyond the carbon footprint issue.
The report identifies an emerging trend: a huge consumption of water, electricity, and land in order to facilitate AI applications. According to experts, if the current scenario continues as it is, then data centres equipped with AI will put natural resources under considerable strain within the next decade.
Specifically, in 2030, AI-enabled data centers will consume 945 terawatt-hours of electricity annually. This is almost three times as much as the annual electricity consumption of Pakistan, Bangladesh, and Nigeria. These three nations have a combined population of over 650 million individuals.
Moreover, the paper predicts an exponential rise in the use of water due to AI. Experts believe that the water footprint for AI data centers may be equivalent to the water needs of 1.3 billion people from Sub-Saharan Africa for an entire year. Furthermore, the land footprint of these data centers may exceed 5,590 square miles, which is almost two times larger than the metro region of Jakarta.
Beyond Carbon: The Hidden Environmental Costs of AI Infrastructure
As stated by the scientists, the debate on the topic of sustainable usage of AI is mostly centered around lowering carbon emissions.
Nonetheless, they believe that equal attention must be paid to such aspects as water consumption and use of land space. Cooling of server rooms demands significant amounts of water. Energy production and building of infrastructure entail usage of the land area.
It should be noted that addressing certain environmental issues may lead to others becoming more severe. Thus, according to experts, using bioenergy as an alternative to coal energy helps reduce carbon emissions by about seventy percent. Still, this change increases water consumption more than thirty times and land use up to one hundred times.
According to lead author Miriam Aczel, these results illustrate that sometimes the most environmentally sustainable choice may not be the optimal one when taking into account water and land utilization. Actions aimed at protecting the environment in some sphere may cause problems in others.
The negative impact of AI on the environment is already tangible in several countries. The consumption of electricity by data centers reached 448 terawatt-hours in 2025, which surpasses the annual electricity consumption in Saudi Arabia.
Energy, Water, and Global Inequality
One case in point in terms of the increased electricity demands is Ireland. In 2023, data centers used 21 percent of Ireland’s metered electricity consumption. Moreover, the energy consumption of data centers was higher than that of all urban households combined. Therefore, Ireland’s state-owned power company suspended construction of data centers near Dublin till 2028.
Water usage has become yet another problem faced by the industry in many cases. Data centers need as much as five million gallons of water per day to cool the equipment. Such high water demands may pose difficulties in arid regions.
In Querétaro, Mexico, rapid data center expansion poses problems for local water sources during droughts. In Uruguay in 2023, the same problem arose because the government planned to build an AI data center during one of its worst-ever water crises.
The drought depleted fresh water and negatively impacted the quality of drinking water in the capital, prompting mass demonstrations.
Another issue raised by the report is growing inequality in the development of the global AI economy. Rich countries continue to attract investment into building AI infrastructure, leaving poor countries out of that.
Only 32 countries had data centers specializing in artificial intelligence as of 2025. Two countries, namely the US and China, possessed around 90 percent of all such data centers in the world. Specialists warn that such uneven distribution will increase digital inequality.
Addressing the Environmental and Social Costs of AI Growth
Finally, another environmental problem identified in the paper is growing amounts of electronic waste. As of 2030, AI infrastructures will create at least 2.5 million metric tons of electronic waste annually. Most of this waste ends up in poor countries, putting local communities under the risk of being exposed to toxic waste.
UNU Rector Dr. Tshilidzi Marwala highlighted that the imbalanced availability of AI infrastructure creates many questions regarding equality and inclusivity. Although artificial intelligence offers various benefits, such as improving people’s quality of life and enhancing economic development, the issue of ensuring that these advantages are enjoyed equitably is crucial.
In light of the aforementioned problems, the researchers propose a more responsible approach to the growth of artificial intelligence. According to them, authorities should incorporate factors such as the consumption of water and soil, as well as issues relating to social implications, into licensing and environmental impact assessments. Moreover, it is essential to encourage investors and financial institutions to ensure that their projects minimize ecological risks.
Finally, the authors believe that the world has only limited time to steer AI toward a more sustainable path of development. Otherwise, people risk failing to reap the advantages of AI while overburdening both people and nature at the same time.
