Artificial intelligence (AI) is a hot topic these days. It is touted for its potential to help solve the world’s problems, but at the same time criticized for consuming valuable resources. Flavio Lehner, our chief climate scientist and an Assistant Professor in Earth and Atmospheric Science at Cornell University, looks at some of the questions we’ve received.
What is the environmental impact of AI and data centres?
AI requires a lot of computing power and thus electricity. Globally, data centres now consume around 1.5% of all electricity. How that electricity is produced is a key determinant of the environmental impact of AI.
If powered by fossil fuels, then the associated greenhouse gas emissions directly warm the planet, melt sea ice, and thus destroy polar bear habitat. This will also continue air pollution from burning fossil fuels.
If powered by renewable energy, this is less of an issue. However, even then, data centres consume renewable energy that could otherwise be used to displace fossil fuels; they thus slow the transition to clean energy, at least initially. There is currently an “arms race” in which most new renewable energy that comes online is used to satisfy society’s increasing demand, which includes growing demand from AI, rather than the historical demand. Only once over 100% of the increasing demand is met by renewable energy will fossil fuels be displaced systematically.
What about water consumption of AI data centres?
Water from lakes and other sources has been used to cool computers for decades, so that’s not new. In functioning democracies, environmental laws dictate how much water can be used and at what temperature it can be returned to nature. This is to ensure no negative effects on the environment. In other words, we know how to responsibly use water to cool computers, so it’s a question of putting in place and enforcing environmental laws that protect species dependent on clean and cool water. However, much like with renewable energy, data centres consume water that could otherwise be used in other ways by humans and ecosystems.
So far, polar bears live far from any major data centre, so they are not affected directly by the water consumption issue.
Are there any environmental benefits from AI?
AI is credited with making many processes more efficient, including in medical and environmental research. AI-based techniques have also helped create new knowledge from the wealth of environmental data that is pouring in from various observing systems such as satellites or Bear-dar, our polar bear-detecting radar. Some AI-based methods can lead to energy efficiency improvements, for example through “smart heating” or better taking into account weather forecasts when balancing the electricity grid.
However, the vast majority of AI use is not focused on solving environmental issues like climate change, so the relative benefits are currently small. We also know what we need to do about climate change: reduce emissions by replacing fossil fuels with renewable energy. The key challenges for this are often political and societal rather than technological. The notion that “AI will solve climate change” is an oversimplification and potentially distracts from focusing on the actual decision makers in government and industry who can make a difference if held accountable by constituents and consumers.
How should people who care about climate be using AI as individuals? At a company level?
The energy consumed by, say, a simple ChatGPT query is not actually that large and comparable to a Google search. The energy cost increases if AI is used for generation of images, videos, and professional-level data analysis workflows. Then it becomes comparable to running a microwave for 2-3 min per query (2025 numbers). Just like with air travel or energy consumption around the house, it is worth considering whether using AI is really needed in a given situation. Similar questions can be asked at a company level. There is a lot of hype around AI and a careful assessment of the costs (energy and financial) of deploying AI within a given company seems prudent. There are AI companies that claim to power their technology with renewable energy, which would be preferable.
AI and its energy efficiency are rapidly evolving topics, so we expect to update this Q&A over time or when new questions come in.
Learn more: