Artificial intelligence has the potential to destroy as well as preserve the environment. As of now, generative AI is stronger in its destructive side.<\/p>\n\n\n\n
While it is clear there is a strong potential for overall good in our future, as the Chief Digital Officer of the United Nations Environment Programme<\/a> Golestan Radwan says, \u201cWe need to make sure the net effect of AI on the planet is positive before we deploy the technology at scale.\u201d<\/p>\n\n\n\n As of now this is not case, but the future looks promising, especially with countries motivated by the Paris Climate Agreement to be net zero by 2050<\/a>. However, the task of becoming climate neutral will only increase in difficulty as more data centers pop up and use massive amounts of energy.<\/p>\n\n\n\n Perhaps the most widely know concern with rise of AI is the consumption of electricity and energy by the data centers required to run it. From 2024, it is projected that the energy use required for AI will increase 10x by 2026,<\/a> and exceed the equivalent of the entire country of Belgium\u2019s annual energy<\/a> consumption. The 2026 estimate for total U.S. energy usage is 6% for 2026<\/a>, a new projection doubles it to 12% by 2028<\/a> . Being mindful of energy consumption is important when most of the world still relies on fossil fuels, which \u201cresults in local air pollution, thermal pollution in water bodies, and the production of solid wastes, including even hazardous materials,\u201d according to the Harvard Business Review.<\/a><\/p>\n\n\n\n Optimistically, advancement efforts have helped lower power usage effectiveness (PUE) in major companies such as Google, whose data center\u2019s PUE came in at 1.08 for the first quarter<\/a> of 2025, compared to 20 years ago when the average PUE for data centers in the United States was 2.0<\/a>, and some climbing as high as 3.0.<\/a><\/p>\n\n\n\n Water usage may not seem obvious at first, but data center technologies produce heat, and large quantities of water<\/a> are used to cool the overheating machines. A common comparison is that one 100 word email written by ChatGPT uses a whole bottle of water<\/a>. It\u2019s important to note that the water used to cool data centers evaporates into steam and doesn\u2019t all go back to the water cycle, which is why there is an important distinguishment between water use and water consumption<\/a>. For more information on how exactly AI data centers consume water, you can read this study on AI\u2019s water footprint.<\/a><\/p>\n\n\n\n On the flip side, major companies which have integrated AI into their interfaces such as Google<\/a>, Meta<\/a>, and Microsoft<\/a> have pledged to be climate neutral\/negative by 2030, and Amazon<\/a> by 2040. Each of them have pledged to be water positive by 2030 as well. Some are doubtful<\/a> it is possible as usage increases exponentially, like University of Pennsylvania professor of electrical and systems engineering Benjamin Lee.<\/p>\n\n\n\n Rare minerals<\/a>, such as cobalt and gallium, are used to make the coveted chips that AI relies on. Often, mining for technology uses unethical labor<\/a>.<\/p>\n\n\n\nImpacts and Improvements<\/h3>\n\n\n\n
Energy<\/h5>\n\n\n\n
Water<\/h5>\n\n\n\n
Materials<\/h5>\n\n\n\n