This story is syndicated from The Black and White, the newspaper of Walt Whitman High School in Bethesda, MD. The original version ran here.
The international AI sector is making headlines now more than ever. Though American AI companies have seen fluctuations in their value — including shifts resulting from President Trump’s recent tariff orders — Trump committed to a $500 billion federal investment in AI infrastructure just days after taking office this January.
AI use also continues to rise, with generative artificial intelligence companies like OpenAI and DeepSeek making waves with their free, accessible platforms. According to a Precedence Research report, as of this month, the global AI market size is valued at $757.58 billion and is expected to grow exponentially in coming years.
In December 2024, Sam Altman, CEO of OpenAI — the company that developed ChatGPT — announced that the platform now boasts 300 million users per week. First released in November 2022, the generative AI system allows users to input a prompt and receive specific and immediate feedback, pulling from public information on the internet and training data to craft a unique response. Recent developments have also enabled the program to generate images and graphics.
As they evolve, these technologies continue to impact daily life through the convenience they offer for completing tasks and accessing information. Despite its contributions to efficiency, however, artificial intelligence’s growing presence has also taken a significant toll on the environment and is positioned to dramatically alter the world’s energy use — and its climate — in the years ahead.
The World Economic Forum estimates that training GPT-3, ChatGPT’s original model, required just under 1,300 megawatt hours of electricity, which is equivalent to the annual energy consumption of approximately 130 American homes. More advanced models, such as GPT-4, ChatGPT’s current model, are estimated to require about 50 times more energy than their prototypes. AI’s large user base significantly contributes to the program’s large energy consumption. One search on ChatGPT uses an estimated 2.9 watt-hours of energy, ten times the amount needed for a single Google search. Balkan Green Energy News reported that ChatGPT uses 226.8 gigawatt-hours yearly to generate responses — enough to power 21,602 American homes for a year.
Energy use of this immense scale has major environmental implications. Though other countries have moved away from relying on finite resources, the majority of energy production in the U.S. involves the combustion of fossil fuels, nonrenewable resources that produce greenhouse gases when burned. These gases accumulate in the atmosphere and trap heat, contributing to air pollution, water contamination and climate change, according to decades of research.
The independent U.S. Energy Information Administration reports that over 60% of Americans’ household energy comes from fossil fuels, with just over 20% coming from renewable sources.
2024 was Earth’s warmest year on record; temperatures reached 2.30 degrees Fahrenheit above the 20th-century baseline. The ten hottest years on record have all been in the past decade. Climatologists have repeatedly confirmed that carbon emissions play a major role in this problem, and according to a 2024 article from the World Meteorological Organization, carbon emission levels from 2024 are projected to have been one billion tonnes greater than the year prior. Notably, carbon emissions from the information and communications technology industry, which includes AI programs, are expected to make up 14% of global emissions by 2040, as predicted in a 2022 article from the Journal of Information, Communications and Ethics in Society. Currently, data centers — facilities that house computing systems and are the backbone of AI technology — comprise two to three percent of global greenhouse gas emissions, an amount comparable to that of the entire aviation industry.
Increased energy demand requires the production of more greenhouse gases, straining an already vulnerable environment. Climatologists warn that the Earth could reach irreversible levels of climate change within the coming decades, with rising sea levels giving way to extreme weather swings, coastal community erosion, mass migrations away from inhabitable areas and habitat destruction that threatens the planet’s food supply.
According to Dan Howells, Climate Campaigns Director at Green America, the AI industry’s energy requirements are halting energy and technology companies’ environmental efforts.
“A lot of them are going backward on community rights, stopping pollution in communities, climate change,” Howells said. “It’s three steps backward from what was two or three steps forward. We’re also at a point, especially with climate change and the effects that energy generation has had on our communities for over a century, that we don’t have a lot of time to take steps backward.”
AI technology also relies on water. In the face of climate change, water is a resource that, though renewable, will fluctuate through major disruptions like drought fires, floods and drinkable water scarcity. Like other servers, AIs run constantly, leading to a high risk of overheating and the need for water as a coolant.
To train GPT-3, Microsoft used 700,000 liters of water, enough to sustain 1,000 people for a year. Similarly, Meta required 22 million liters to train their open-source AI model, Llama-3. In a 2024 article, scientists from the University of California, Riverside, and the University of Texas at Arlington estimated that by 2027, the AI industry may demand 1.1 to 1.7 trillion gallons of water withdrawal.
According to UNICEF, about two-thirds of the global population currently experiences severe water scarcity for at least one month each year. By 2030, their researchers predict approximately 700 million people could be displaced due to water scarcity.
“We need water to grow food, we need water to feed people, we need water coming out of our taps,” Howells said. “You grab more water to do something like AI, it just makes really bad problems even worse. There’s only so much water to go around, and we don’t have that much more […] to give to a whole new industry.”
AI systems also produce a significant amount of electronic waste. The materials used to construct the computer systems are non-recyclable, so they must be disposed of when they stop working or are rejected. Electronics also aren’t biodegradable and therefore release toxins that negatively impact air, soil and water quality when discarded into regular landfills rather than those designated for toxic waste. AI’s high equipment turnover exacerbates an existing waste problem that is expected to worsen in the future. With a surge in new data centers, the amount of electronic waste, or e-waste, is only growing. According to a 2024 MIT Technology Review article, generative AI could produce five million metric tons of e-waste by 2030.
Despite the significant environmental impact of the industry, many companies haven’t converted to more sustainable practices or followed through with their “Extended Producer Responsibility” — a pledge first developed in the ’90s stating that creators and manufacturers are responsible for telling consumers how to sustainably dispose of their products.
“It’s disappointing because you’re seeing a lot of backtracking from companies who are in the lead on the AI front and have been in the lead on the technology front for many years, retreating back to old sources of energy to drive profits,” Howells said.
Major tech companies like Google and Microsoft are falling behind on their emissions goals due to their recent developments and investments in AI. In 2021, Google set a goal to reach “net zero” carbon emissions by 2030, a milestone the company’s Chief Sustainability Officer labeled ambitious. To reach net zero, the company would have to remove the equivalent amount of greenhouse gases from the atmosphere that it emits.
However, Google’s 2024 environmental report revealed that their emissions had grown by 50% since 2019 due to their large-scale AI development. Despite the increase, Google continues to expand its AI sector, recently announcing the construction of a new data center in Indiana.
Microsoft, a major investor of OpenAI, has pledged to be carbon-negative by 2030, not only balancing out its own emissions but offsetting others’ as well. According to the company’s 2024 sustainability report, however, its emissions have risen by almost 30% due to data center construction. Microsoft is taking steps to reduce its carbon footprint by expanding its use of renewable energy, though it continues to pursue AI development.
Other companies aren’t as transparent about their emissions. OpenAI, for example, doesn’t disclose the amount of carbon Chat-GPT’s operation emits. Some internet users have noted that, even if users ask the chatbot for an exact number of their emission levels, the service won’t output a straightforward answer.
Co-captain of Maryland’s Walt Whitman High School’s Sustainable Scholars Club junior Sarah Rezavi believes that new AI expansions aren’t worth their environmental toll.
“A lot of it is just sort of unnecessary,” Rezavi said. “You don’t need to AI a new photo, you can go on the internet and find pre-existing photographs that people have taken.”
Amidst changes in executive leadership, the federal government hasn’t acted consistently to reverse these worsening problems. On his first day in office, President Donald Trump signed an executive order removing the United States from the Paris Climate Agreement, which sets standards for involved parties to reduce their carbon emissions, as well as systems for reporting and monitoring their progress. Trump has also reversed many environmental action policies established under previous administrations and limited the authority, staff and funding of the U.S. Environmental Protection Agency.
According to Howells, effective change is in the hands of the people. There are still opportunities for citizens to limit the detriments of climate change, despite the big industries and administrations dominating the decision-making.
“It goes back to political will and societal will,” Howells said, “that we can get to a point where we realize we’re all going to be hurting because of the effects of climate change unless we can do something about it pretty quick.”