AI's energy use is a big problem for climate change.
AI has created a surge in the demand for energy which has boosted climate-warming greenhouse gas emissions. Historically computing has been a fairly small climate polluter. As recently as 2020, data centers and data transmission networks produced just .6% of humanity's climate-warming greenhouse gasses. In the U.S. where AI has grown the fastest, it has changed the whole trajectory of the electric grid.
Vijay Gadepally, senior scientist at the MIT Lincoln Laboratory Supercomputing Center, states that the data center used to be about 1-2% percent of the U.S.'s electricity consumption. As of today, we're closer to 4-5%.
Energy needs in the U.S. are growing again. A U.S. Department of Energy study estimates that data centers' energy use will at least double and may more than triple by 2028. Growth similar to this is expected from China and Europe as well.
Every time you talk to generative AI models, you pass in some data that is flowing through hundreds of billions of parameters. The more parameters it has, the more accurate it is going to be. This means every time we use one of these models, and the more they are "trained" and improve, the more energy we use to get the right answers from them. AI has an appetite for computing power.
These AI data centers, and the AI servers associated, produce massive amounts of electronic waste (mercury and lead). They are large consumers of water (to cool electronic components). (Globally, AI related infrastructure may soon consume six times more water than Denmark. Considering a quarter of humanity already lacks access to clean water, that is a problem). AI relies on critical minerals and rare elements, and, again, uses massive amounts of electricity, spurring the emission of planet- warming greenhouse gases.
The Chief Digital Officer of the Unites Nations Environment Programme, Golestan (Sally) Radwan, states that there is still a lot to learn about the environmental impacts of AI but current data is concerning. We need to make sure the positives that come from using AI outweigh any negative environmental footprint it may leave.
We are not without hope. Yuan Yao, associate professor of industrial ecology and sustainable systems at the Yale School of the Environment, is part of an initiative led by the National Science Foundation to reduce the carbon footprint of computing by 45% within the next decade. Countries need to pay attention to the measure of environmental impact of AI and create regulations based on those measurements. Along with using other forms of renewable energy, countries need to encourage companies to "green" their data centers.
This data is seriously scary. I have never thought that simply using a computer could cause so much harm. It seems small, but with the amount of people that have computers and use them everyday... this number really adds up.
ReplyDeleteIt's even more scary because our phones contribute too! We're always using technology.
DeleteAI is a useful tool for research, as it's already being used to generate prediction models utilizing tens of thousands of different data points in a short amount of time - a feat that wasn't really possible before! As amazing as AI tools can be though, something absolutely needs to be done about their implementation as well as their raw power requirements.
ReplyDeleteDefinitely pros and cons of it's usage. I just don't think people really recognize how impactful it all can be.
DeleteThere is exciting new information about how a Chinese start up company has jumped ahead of all the huge American AI companies to create an excellent AI with 40x less energy demand!! Check this out: https://thediplomat.com/2025/02/chinas-overlooked-ai-energy-edge-over-the-us-cheaper-energy/
ReplyDeleteThank you for the resource!
DeleteI did not realize how energy intensity AI is and how it can impact climate change. Making sure that they get electric from green sources is key for this new technology.
ReplyDeleteCrazy to think AI helps us so much but quietly hurts the planet too. We need smart tech that’s smart for Earth too.
ReplyDeleteSo many individuals blindly use AI daily without being aware of the consequences it can have on our environment, I appreciate you taking the time to put this into a blog and further inform us of the harm AI is causing and the resources that are depleted by running AI platforms.
ReplyDeleteIt reminds me of our readings where carbon captures require an immense amount of energy to the point where they are doing more harm than good.
ReplyDeleteI knew that it took more power consumption, but this post really puts into perspective the real effects that it has on our planet.
ReplyDeleteWow, I didn't know AI had such a massive energy consumption. The fact that data centers' energy use in the US has already doubled to around 4-5% of total electricity consumption shows how quickly this issue is escalating.
ReplyDeleteMore and more development happens every day. People don't really realize how devastating it can be when we're not careful with these devices!
DeleteThis is a very interesting post! Especially with the increase of demand in every day users of AI such as students of all ages I can imagine this would create an uproar in energy consumption. I didn't know that these systems use water to cool down, and I am curious how that water is treated after use, or if there may be an alternative ti decrease water pollution.
ReplyDelete