Artificial intelligence (AI) is rapidly permeating daily life, reshaping how we manufacture, teach, consult, code, and organize. It’s predicted that generative AI will make many key industries much more productive, displacing an estimated 92 million human jobs by 2030, according to the World Economic Forum. As AI shifts the future of labor and education, it‘s simultaneously being heralded as a tool for climate action.
Artificial intelligence tools with the capacity to enhance data collection and analysis improve energy system management, model climate risks, locate and safeguard carbon sinks, improve agricultural processes, and process environmental information at otherwise unmanageable scales have the potential to drive down greenhouse gas emissions through targeted deployment in high-impact sectors. Beyond improving the management of strategic and environmental data and enabling green transitions, AI is increasingly sought to unlock frontiers once beyond human reach, from decoding animal vocalisations and plant signalling to enabling molecular recycling, ecosystem-level farming, and the tracking of invisible pollutants. Perhaps through these more strategic uses, AI might improve our ability to tackle climate and environmental challenges. Or will it make these challenges worse?
These visionary applications come with a steep environmental price tag. According to the International Energy Agency, data centers in the United States are set to drive nearly half of the growth in electricity demand by 2030, in a quiet shift where the Cloud takes over the grid. And while AI has a global reach, its environmental burden is often local; developing and running datacenters requires immense amounts of water and energy, and the resulting heat must be continuously managed to prevent system failure, writes legal expert Whitni Simpson. This places many of AI’s hidden costs on communities living near water-intensive data centers. The same is true at mining sites that supply the minerals essential for AI hardware, where the extraction and refining are poisonous for the local environment, writes environmental studies and public policy scholar Sophia Kalantzakos.
For example, Erin Brockovich-style reporting is finding that in Memphis, Tennessee, Elon Musk’s xAI supercomputer, Colossus, is worsening air quality due to its massive use of methane-gas turbines, emitting a smog of nitrogen oxides and toxic formaldehyde. Without disclosing the use of these turbines or gaining proper air permits to release these toxins, Colossus was able to come online in 122 days, the quickest an AI supercomputer has been constructed to date. The area in proximity to the datacenter is historically Black and has disproportionately high rates of cancer and asthma. Under the guise of new investment in the area, government officials are permitting the continued operation and build out of Colossus. Residents are responding, by organizing and demanding protection for the health of their community and the environment.
The situation in Memphis isn’t unique. Despite growing awareness of AI’s environmental impact, most discussions and regulations focus narrowly on electricity demand and energy intensity. Far less attention is paid to carbon emissions and water use, in part because power consumption is the most visible metric. What happens behind the walls of data centers, where water is drained for cooling and carbon quietly accumulates, rarely makes headlines. In fact, the water demands of AI infrastructure present growing concerns, particularly where datacenter facilities are co-located in regions already facing scarcity.
More to Explore
What Is Greenwashing?
Datacenters rely on water-intensive cooling to keep servers functional, while the power plants that supply their electricity often demand additional freshwater for generation. Drawing on shared water supplies can exacerbate local stress, with consequences that extend beyond ecosystems to wider social and economic systems.
Moreover, the byproducts of manufacturing and operating digital infrastructure can contain contaminants that, if not properly treated, pose risks to water quality and ecosystems. Discharge of polluted water and discarded electronic components into rivers, groundwater supplies, landfills, or unregulated recycling facilities results in the hazardous materials gradually seeping into surrounding soil and water systems. Over time, these substances may infiltrate groundwater, creating serious long-term health concerns, degrading habitats, and endangering communities that rely on those resources. In areas where water availability is limited, competing demands, such as agriculture or community needs, may be constrained, illustrating how AI’s resource footprint can intersect with broader development challenges.
Weekly Newsletter
Globally, the uptake of AI has an impact on climate change mitigation. Training a single AI model can produce up to five times the carbon emissions of an average car over its entire lifetime. The rapid expansion of AI is already creating challenges for major technology companies in meeting net-zero targets, according to The New York Times. Google’s 2024 sustainability report notes a 48 percent increase in emissions since 2019, largely driven by rising energy demands in its datacenters. The company’s growing dependence on AI is transforming its overall energy profile, pushing demand upward and locking in increased infrastructure needs that complicate efforts to bring overall emissions down.
Have we absorbed the lessons of past technological paradoxes? In an effort to normalize the pace of innovation around AI, the technology has been compared to the steam engine, the calculator, and the first-ever computer. The Jevons Paradox reminds us that gains in efficiency often fuel greater overall consumption. In a world where growth is equated with progress, capitalism breathes through expansion, and consumption becomes its oxygen. The drive to consume, innovate, and develop ever more advanced futures proves difficult to resist, leaving critical impacts under-examined and under-regulated. Opportunities to integrate generative AI must be balanced against the environmental costs of the servers themselves, which span the entire hardware life cycle, from manufacturing and operation to disposal, and include significant carbon, energy, and water footprints. The real test is not whether AI can help us fight climate change, but whether it can do so without reproducing the same injustices and excesses that brought us here.
Support JSTOR Daily! Join our membership program on Patreon today.
