Environmental toll of generative AI sparks concern
Each interaction with artificial intelligence (AI) may feel seamless, even effortless — but behind the scenes, the environmental costs are escalating.
A single ChatGPT query consumes approximately 3 watt-hours of electricity, about 10 times the energy of a typical Google search. While much of the public discourse around AI focuses on its potential to replace human jobs, far less attention is given to its growing ecological footprint. Tools like ChatGPT, DALL-E and other generative AI systems have transformed industries ranging from education to health care and customer service. However, behind the scenes, the massive energy consumption, water usage and resource depletion required to power these tools raise critical sustainability concerns.
Training and running large-scale AI models requires immense computational power, which draws vast amounts of electricity — much of it from nonrenewable sources like coal, natural gas and oil. According to researchers at MIT, data centers that support AI are among the biggest energy consumers in the tech industry. The International Energy Agency reports that data centers worldwide consumed about 1 percent to 1.5 percent of global electricity in 2022, and this is projected to double by 2026, largely driven by AI workloads. Since much of this electricity still comes from fossil fuels, the carbon emissions tied to AI use can be substantial.
Furthermore, AI infrastructure places a heavy strain on water resources. Data centers generate a massive amount of heat and must be cooled to function efficiently. Many rely on evaporative cooling systems, which consume millions of gallons of water per year. For example, Google’s data center in The Dalles, Oregon, used an estimated 274 million gallons of water in 2021 alone. In areas already grappling with drought or water scarcity, this can worsen local environmental and public health conditions.
Another often-overlooked aspect is the growing tide of electronic waste. As AI technology advances rapidly, frequent upgrades to GPUs, CPUs and other hardware components are required. These outdated components often end up in landfills, contributing to environmental pollution. Compounding the issue is the production of AI hardware, which depends heavily on rare earth minerals such as lithium, cobalt and neodymium. Mining these materials not only depletes finite resources but also disrupts fragile ecosystems and often involves exploitative labor practices.
A report from Harvard Business Review highlights the unequal distribution of AI’s environmental toll. The countries that host massive data centers often bear the brunt of electricity and water usage, while the global population benefits from the services AI provides. This geographic imbalance raises ethical questions about environmental justice and the accountability of large tech corporations.
Experts stress that sustainability must become a core priority in AI development. Promising solutions include improving algorithmic efficiency to reduce computational demand, shifting data centers to renewable energy, adopting liquid or immersion cooling technologies and enhancing hardware recycling programs.
As AI continues to reshape the digital world, its hidden costs must not be ignored. Without proactive and responsible measures, the environmental consequences of AI may soon outweigh its benefits, leaving a lasting impact on both humanity and the planet.
Hannah Yeo is an 11th grader at Choate Rosemary Hall.