In 2025, AI and weather change, two of the biggest political stalkers we’re facing, may meet.
The summer of 2024 was the hottest day on Earth since data collection began, leading to much media attention and people conversation. Additionally, this year marked the year that both Microsoft and Google, two of the biggest software companies investing strongly in AI research and development, failed to meet their climate goals. While this likewise made headlines and spurred anger, AI’s climate impacts are still far from being common understanding.
In fact, AI’s present “bigger is better” paradigm—epitomized by software companies ‘ pursuit of ever bigger, more powerful big language models that are presented as the solution to every problem—comes with extremely important costs to the environment. These range from producing enormous amounts of energy to power the data centers that house devices like ChatGPT and Midjourney to the thousands of gallons of water that are pumped through these information centers to prevent overheating and the tons of rare earth metals required to construct the equipment they contain.
Data centers currently use 2 percent of the world’s electricity. This figure exceeds one-fifth of the electricity produced in nations like Ireland, which is why the Irish government placed a strict moratorium on new data centers until 2028. While a lot of the energy used for powering data centers is officially” carbon-neutral”, this relies on mechanisms such as renewable energy credits, which do technically offset the emissions incurred by generating this electricity, but don’t change the way in which it’s generated.
Virginia’s Data Center Alley is primarily powered by nonrenewable energy sources like natural gas, and energy companies are putting off coal power plants to keep up with the rising demands of AI-driven technologies. In locations ranging from Arizona to Spain, data centers are slurping up significant amounts of freshwater from dwindling aquifers, pitting local communities against one another. In Taiwan, the government decided to give chip manufacturing facilities valuable water resources to stay ahead of the growing demands rather than letting local farmers use them to water their crops in the worst drought the nation has experienced in over a century.
My most recent research indicates that switching from older, standard AI models trained to handle a single task, such as question-answering, to the new generative models can consume up to 30 times more energy just for answering the same set of questions. We don’t know how much energy is used when generating an image with Google’s Gemini or when adding generative AI models to everything from search engines to text-processing software, but technology companies that are adding generative AI models are doing this without disclosing the carbon cost of these changes.
Either it’s not really an issue ( according to Bill Gates ), or an energy breakthrough will come along and magically fix things ( according to Sam Altman ), has been the focus of a lot of Big Tech discourse regarding AI’s environmental effects. By voluntary initiatives like the AI Energy Star project that I’m leading, which would help users compare the energy efficiency of AI models to make informed decisions, what we really need is more transparency about the impact of AI on the environment. I anticipate that voluntary initiatives like these will start to be implemented through legislation from national governments to intergovernmental organizations like the UN in 2025. In 2025, with more research, public awareness, and regulation, we will finally start to grasp AI’s environmental footprint and take the necessary actions to reduce it.
