BEHIND THE CODE: THE ENVIRONMENTAL COST OF AI
- Poorvi Sanath
- Jun 8
- 3 min read
By POORVI SANATH
Artificial intelligence is no stranger to most of us. From healthcare to education to social media, its presence is ubiquitous. While this explosive advancement is exciting to witness, it warrants some concern.
A major source of strain on the environment is the data centers that house computing infrastructure like storage drives, servers and network equipment. These temperature-controlled buildings are used to train and run the deep-learning models that AI tools require. Although these centers have existed since the 1940s, the surge in generative AI has necessitated an exponential surge in data center construction.
Data centers are inherently energy intensive. In 2021, a research paper by scientists at the University of California at Berkeley posited that training AI models alone consumed 1,287 megawatt hours of electricity - this is enough to fuel around 120 U.S. homes for a year. Further, according to the International Energy Agency, a search through Chat-GPT, an AI tool, consumes 10 times the energy of a Google search. This demand for energy poses a risk to sustainability. “The pace at which companies are building new data centers means the bulk of the electricity to power them must come from fossil fuel-based power plants,” says Norman Bashir, a Computing and Climate Impact Fellow at MIT Climate and Sustainability Consortium. Since AI models have a short shelf-life (new models are released every week), energy consumed on older models is wasted. Newer models have more parameters, and thereby consume more energy.
Another problem with data centres is their requirement for water during construction, and to control the temperature of electrical components. Chilled water is used to absorb heat from equipment to keep it cool. Although AI is called ‘cloud computing’, its hardware has a real dependence on water that strains reserves globally. It is estimated that infrastructure for AI may soon consume six times as much water as Denmark, a country with a population of nearly six million. This is especially worrying given that a tenth of our global population live in countries with critical water stress and a quarter still lack access to clean water.
The environmental consequences of AI are differentiated around the world. This is a reflection of socioeconomic disparities and regulatory differences between communities. In Finland, Google ran its data center in 2022 on 97% of carbon-free fuel; in Asia, that percentage dropped to 4-18%. Environmental inequity is exacerbated in water stressed areas like Arizona and Chile that are disproportionately affected by the needs of data centers. The AI Now Institute also found, in its 2023 Landscape Report, parallels between this uneven environmental impact and historical practices of settler colonialism and racial capitalism.
That is not to say, however, that all AI is detrimental to the environment. It is invaluable to climate science in detecting anomalies in data and increasing efficiency. The United Nations Environment Programme, for instance, utilizes artificial intelligence to detect methane emissions from oil and gas installations. Additionally, optimization processes are being developed to reduce the dependency on energy. Data center operators are pursuing plans to reach net zero emissions through solar farms and renewable energy credits. They also aim to be “water positive” by 2030; this would mean a surplus of water.
Despite the remarkable growth of AI innovation, its impacts on the environment are majorly obscured. Whether the purported benefits of AI are worth the costs remains debatable, but given its inevitable expansion and potential, it is imperative to proceed with foresight and sensitivity.
Citations
AI has an environmental problem. Here’s what the world can do about that. United Nations Environment Programme.
Electricity 2024 - Analysis and forecast to 2026. International Energy Agency.
Li, P; Yang J; Islam, M; Ren, S (2025). Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv.org.
https://arxiv.org/pdf/2304.03271
Ren, S; Wierman, A (2024). The Uneven Distribution of AI’s Environmental Impacts. Harvard Business Review.
https://hbr.org/2024/07/the-uneven-distribution-of-ais-environmental-impacts
Water. United Nations.
https://www.un.org/en/global-issues/water
Zewe, A (2025). Explained: Generative AI’s environmental impact. MIT News.
https://news.mit.edu/2025/explained-generative-ai-environmental-impact-0117
