The Growing Environmental Impact of Data Centers and AI

A recent report from the United Nations University reveals that data centers have an environmental footprint comparable to some of the world’s largest countries. Their use of electricity, water, and contribution to pollution is predicted to double within four years due to the increased prevalence of artificial intelligence (AI).

Last year, global data centers consumed 448 trillion watt-hours of electricity, surpassing the usage of all countries except ten. This energy consumption resulted in approximately 208 million tons of carbon dioxide emissions, on par with Argentina’s emissions. Producing this level of energy also required about 1.2 trillion gallons of water.

Projected Energy Use and Emissions

By 2030, data centers are expected to account for nearly 3% of global electricity use, reaching 935 trillion watt-hours. If data centers were a country, this would rank them as the sixth-highest in power consumption. This level of electricity usage could produce nearly 440 million tons of carbon dioxide emissions.

The study, co-authored by Kaveh Madani from the United Nations University Institute for Water, Environment, and Health, highlights the significant demand of data centers. AI currently accounts for about 20% of this energy use, projected to grow to 40% by 2030.

Importance of the Report

Fengqi You, an energy engineering professor at Cornell University, noted that the value of the report lies in the United Nations addressing carbon, water, land, life-cycle impacts, and environmental justice. Jean Su from the Center for Biological Diversity emphasized the report’s importance as it is a global attempt to expose the environmental challenges posed by AI.

Caleb Max, President of the National Artificial Intelligence Association, acknowledged the efficiency improvements within the AI industry and the benefits it provides. He highlighted AI’s role in enhancing safety, productivity, food production, and poverty reduction.

Josh Levi, President of the Data Center Coalition, stated that the industry remains committed to minimizing its environmental impact responsibly and transparently.

Reducing AI’s Energy Consumption

Kaveh Madani highlighted the substantial environmental costs associated with AI, often overlooked. He explained that being concise in queries can significantly reduce AI’s energy consumption. Cutting words by 30% in queries can lower energy use by 25%, equating to the electricity used by 700,000 people in Africa annually.

A typical AI query, such as those from ChatGPT, requires significantly more energy than basic tasks like email spam filtering. The more complex the AI, the greater the energy needed.

Challenges and Transparency

Miriam Aczel, a researcher with the United Nations University, stated that about 90% of AI’s power usage comes from operational requests, rather than training phases alone. Despite more efficient machines, increased consumption presents challenges.

Madani noted a paradox where increased efficiency leads to even higher total energy use. The report calls attention to the lack of transparency regarding data center and AI consumption details.

According to the Associated Press, conducting studies on this subject is hindered by inaccessible data on data centers’ size and resource use. The accurate management of environmental impacts requires more openness from companies.