TIP
Artificial Intelligence and the Environment
We often talk about artificial intelligence in connection with chatbots, image generation, and work automation. What impact does the operation of AI itself have on the environment? And conversely, how can AI help fight the climate change?
When AI takes
Artificial intelligence consumes energy and water both during model training and during regular user operation. How does it work? AI runs in data centers, which consume electricity, generate heat, and require water cooling. According to a report by the International Energy Agency (IEA, 2025), data centers consumed approximately 415 TWh of electricity in 2024, accounting for about 1.5% of global demand. An updated IEA report (2026) confirmed that consumption in 2025 rose by 17% year-over-year, with AI-focused data centers seeing an even more significant increase of 50%.
The environmental impacts, however, are not limited to water and electricity. De Vries-Gao (2026) estimated that AI systems alone could have produced 32.6 to 79.7 million tons of CO₂ emissions in 2025. A volume comparable to New York City’s carbon footprint. According to the same estimate, the water footprint of AI systems could reach 312.5 to 764.6 billion liters, roughly equivalent to the global annual consumption of bottled water.
If you'd like to get an estimate of your AI usage, you can calculate a rough figure.
When AI can provide
One innovative approach to mitigating these impacts is the use of waste heat from data centers. In their study, Diaz-Marin and Berquist (2025) proposed that waste heat from data centers could be used for water treatment and carbon capture, thereby helping the centers move toward carbon-negative and water-positive operations.
AI also offers significant potential for accelerating climate action—through climate modeling, emissions monitoring, improving energy efficiency, and optimizing renewable energy sources (Vinuesa et al., 2020). A study by Galmarini et al. (2025) identified five areas where AI can help build an effective response to climate threats and estimated the potential for emissions reductions in the energy, food, and transportation sectors, which together account for nearly half of global emissions.
Canada provided a concrete example of AI in practice when, in May 2026, it became the first country in the world to launch a hybrid weather forecasting system. The Global Environmental eMuLator (GEML) model, developed by researchers at Environment and Climate Change Canada based on Google DeepMind’s GraphCast technology, combines machine learning with traditional physical-numerical methods. The result is more accurate forecasts of extreme weather, giving residents more time to prepare (Environment and Climate Change Canada, 2026).
The key question, then, is not whether AI is good or bad for the environment. Rather, it is about how to ensure that the positive climate benefits of AI outweigh the environmental costs of its operation. This requires transparency from technology companies regarding energy and water consumption, investment in renewable energy sources and more efficient chips, and above all, as de Vries-Gao (2026) emphasizes, new regulatory policies that would require companies to publish detailed environmental metrics.
