ChatGPT Is Everywhere — Why Aren't We Talking About Its Environmental Costs?

Yahoo

07-05-2025

Daniel de la Hoz
Stay up-to-date with the politics team. Sign up for the Teen Vogue Take
We know that young people are concerned about sustainability and the environment — perhaps especially so under President Trump. Since taking office in January, his administration's 'assault on the country's climate ambitions,' a recent New York Times op-ed argues, 'is not just enraging but also perversely awe inspiring.' It makes sense that people — as they watch the government undermine environmental policy — are trying to figure out what they can do as individuals.
Teen Vogue recently observed that more readers are coming to our site via ChatGPT, specifically through searches for actions you can take on sustainability. It's no wonder, when we're seeing a growing reliance on AI chatbots for everything from drafting texts to therapy — and with sometimes scary results. (Nonetheless, Meta's Mark Zuckerberg recently suggested that AI chatbots could take the place of IRL friends.)
Those searching for how to live sustainably might not realize that using ChatGPT itself has consequences for the environment. 'Most people are not aware of the resource usage underlying ChatGPT,' Shaolei Ren, an associate professor at the University of California, Riverside, who studies AI's impact on climate, told the Associated Press in 2023. 'If you're not aware of the resource usage, then there's no way that we can help conserve the resources.'
So we're here to help break it down: ChatGPT is a large language model, or LLM, which is an AI-based machine-learning model that is trained on large amounts of data, enabling it to create writing that might look, in theory, as though a human made it. Training LLMs like ChatGPT require a huge amount of computing power, as well as for generating answers. The servers that provide this power need to be kept cool, which often requires significant amounts of water.
Then there are the energy needs. Substantial quantities of electricity are often required to train, fine-tune, and run LLMs, creating carbon emissions and potential energy strain, according to MIT News. Estimates vary on the amount of resources a ChatGPT search requires compared with a typical Google search (sans AI overviews), but a 2024 report from the International Energy Agency placed the chatbot's energy usage for a single query at nearly 10 times that of the search engine.
Amid the demand for expanding AI usage, companies like Google and Meta are rushing to expand their energy capacity, particularly by investing in nuclear energy, which doesn't emit greenhouse gases but, critics say, comes with its own set of potential problems for people and the environment. Microsoft is looking to resurrect a closed-down nuclear plant as one way to power its AI offerings.
A 2024 analysis of the energy output of using ChatGPT by the Washington Post and researchers at UC Riverside found that just one 100-word email drafted by ChatGPT-4 uses about a water bottle's worth of H2O, and enough electricity to power 14 LED lightbulbs for an hour. (If you want to see some visualizations of the resources used by AI to help make one example feel more tangible, the estimations in that WaPo piece are a great place to start.) According to the Post, data centers are also sapping the US power grid, which has historically been under-invested in and under-resourced.
An engineer described AI, and the data centers it necessitates, to Bloomberg as a 'big hammer' on the US energy grid: 'Take your house and increase that by 10,000. That is the difference between your house and a data center.'
On top of that, currently, these centers often rely on emissions-heavy forms of energy production, like coal. This has resulted in prolonging the existence of coal-based power plants in places like North Omaha, Nebraska. The 'low-income, largely minority' neighborhood has 'some of the region's worst air pollution and high rates of asthma,' according to the Washington Post. A local power company was set to stop burning coal at a '1950s-era' power plant in 2023; as of last fall, despite community concerns, the plant planned to keep burning coal until at least 2026, supplying power to data centers owned by Google and Meta.
We've seen what happens when power grids in places like Puerto Rico and Texas are unable to withstand the sort of extreme weather worsened by climate change. Critics argue that directing energy toward AI not only reappropriates resources that could otherwise be used to provide for people's basic needs, it also contributes to the worsening climate conditions that help make those grids so vulnerable. As over a billion people across the world live with high water vulnerability, Microsoft and Google have both reported drastically increased water consumption and greenhouse gas emissions over the last few years.
'There are tangible and measurable harms attached to data center expansion and the use of resources for powerful AI,' Tamara Kneese, the director of Data & Society's Climate, Technology, and Justice program, tells Teen Vogue via email. Data & Society is a nonprofit research organization that studies 'the social implications of data, automation, and AI,' per its website.
'Places with high concentrations of data centers — which are often clustered in areas where marginalized people who have historically experienced environmental racism live — are dealing with pollution and subsequent public health issues, increased utilities costs for ratepayers, and lost access to energy, water, and land,' Kneese continues. 'Transmission lines go through public parks and agricultural land. The communities around data centers and related energy infrastructures are being sacrificed for a speculative future AI, and in many cases communities are actually paying for data centers through subsidies granted to companies."
Says Kneese, 'We are told repeatedly that AI will eventually help us solve social issues, including climate change, even while AI infrastructures are right now burdening communities and undermining climate goals.'
These companies certainly seem to be aware of the consequences of pursuing more and more energy- and resource-consuming AI. Environmental reports for 2024 from Google and Microsoft detail ambitious sustainability goals that include reducing emissions and water usage, while making clear they know they haven't been meeting their targets. (Google, Open AI, and Microsoft have each responded to the arguments in this op-ed; you can find their responses at the bottom.)
In addition, following the January premiere of the latest version of China-based AI company DeepSeek's LLM, which uses fewer resources to create what a number of experts say are better search outcomes, there's evidence that it doesn't have to be like this.
'Wait a minute,' Karen Hao, a contributor at The Atlantic who covers AI's impact on society, wrote on Bluesky in a series of posts analyzing the news. 'You mean to say that we don't need to blanket the earth with data centers and coal & gas plants to maybe arrive at a future where we can wave a magical [artificial general intelligence] wand to make all of the consequences of that go away? Yes. This is a false trade off. Let that sink in.'
We get the appeal of a quick ChatGPT search, especially as it becomes more commonplace in our schools. A Pew Research Center survey last fall found that the percentage of teens ages 13-17 using ChatGPT for schoolwork had doubled since 2023. There are school districts and departments of education that are incorporating it into their classrooms too. This year in New Jersey, for example, 10 school districts were each awarded about $75,000 in grant money to 'help pay for programs focused on both teaching with AI and teaching about AI,' including AI literacy and ethics, according to Government Technology.
'We are told repeatedly that AI will eventually help us solve social issues, including climate change, even while AI infrastructures are right now burdening communities and undermining climate goals.'
Sociologist, professor, and cultural critic Tressie McMillan Cottom, in a recent column for the New York Times, observed that many academics, initially concerned by the onset of AI — how it could enable cheating, for instance — are now treating AI as something they must accept.
McMillan Cottom categorizes artificial intelligence as 'mid' tech — hardly the technological revolution worth the amount of waste and environmental damage it's meting out: '[Most] of us are using [AI] for far more mundane purposes. AI spits out meal plans with the right amount of macros, tells us when our calendars are overscheduled, and helps write emails that no one wants. That's a mid revolution of mid tasks.'
And that's without even mentioning that the LLMs themselves are imperfect and often provide false or inaccurate information. Sometimes these imperfections result in missing fingers in images of AI-generated 'people.' Other times the consequences are far more grave: civilians killed and others wrongfully detained by Israel's military due to AI-based information, according to the New York Times; reported plans to target students over AI-based analysis of their social media; reports of Elon Musk's Department of Government Efficiency attempting to incorporate AI into its government takeover.
McMillan Cottom argues that AI is in some ways perfect for the 'post-fact era.' I agree. It parasitically harvests information, removing it from its context, lacking the ability to analyze it with the same level of nuance that a person might. The Trump administration is bearing down even further on education, and all the while (and as a consequence) we're exposed to, as McMillan Cottom says, 'less research and more predicting what we want to hear.' There's real value in simply doing the reading yourself and showing your work, like your math teacher likely told you.
One small thing you can do if you want to reduce your AI usage: If you're using Google, adding
'-AI' to the end of all your search queries will remove the automated AI summary that the company has added to search outputs. If you'd rather avoid the tech behemoths entirely, you can switch to another search engine, like DuckDuckGo.
We know that tech corporations, fossil fuel companies, and governments bear the most responsibility for the accelerating climate crisis — not individuals. But you can choose to opt out of the AI hype.
Editor's note: In August 2024, Condé Nast, Teen Vogue's parent company, announced 'a multi-year partnership with OpenAI to expand the reach of Condé Nast's content.'
In response to requests for comment, the following companies shared:
Google: 'AI has the potential to help mitigate 5–10% of global greenhouse gas emissions by 2030 — for example, Google is using AI to reduce emissions by suggesting fuel-efficient routes on Maps and helping airplanes avoid contrails. To help minimize our environmental impact, we build efficient AI infrastructure and work hard to reduce and measure its water and carbon footprint.'
Open AI: 'Alongside others within the industry, we continue working hard to find new ways to ensure our technology is as efficient as possible, including when it comes to energy and water consumption. Even as we continue to see significant efficiency gains, promising research and innovation in this evolving space, we believe that being thoughtful about the best use of computing power remains critically important.'
Microsoft: 'Microsoft announced in 2020 that we are working toward become [sic] carbon negative, water positive, and zero waste by 2030, and remain focused on these goals.… Several actions we are taking are outlined in our Accelerating Sustainability with AI playbook.'
Originally Appeared on Teen Vogue
Want more Teen Vogue climate coverage?