As vast resources are poured into the creation of new AI models, the technology is quickly being integrated into industries across the globe. Unfortunately, the development of this new technology has not been done with social and environmental implications in mind, and policymakers are now trying to catch up and create regulations that will prevent the worst of AI’s impacts. Issues of copyright infringement and misinformation are regularly discussed in public media; however, AI’s relationship to the environment and climate change is talked about less.

Scientists hope that AI will improve climate change mitigation and adaptation by analyzing carbon emissions, forecasting environmental processes, and informing disaster relief. However, the development and continued use of AI has massive detrimental impacts on the environment, both locally and globally. The extensive water and energy use of AI needs to be considered, especially as AI development continues to scale.

AI data centers require massive amounts of energy, and as more data centers are built all over the world, their energy consumption is increasing rapidly. According to the Harvard Business Review, the training of an AI model “can consume thousands of megawatt hours of electricity and emit hundreds of tons of carbon.” An estimate from the International Energy Agency’s Electricity Report forecasted that AI energy demand will have increased tenfold between 2023 and 2026. AI models require vastly more energy than search engines such as Google. For example, the IEA report found that a single request made to the model ChatGPT consumes 10 times as much energy as a Google search. If a large number of users began using AI as a substitute for Google, there would be a spike in energy consumption and emissions.

The detrimental effects of AI data centers extend beyond their energy use and carbon emissions. AI data centers consume massive amounts of fresh water for cooling, much of which is potable. In mid January, researchers from UC Riverside and UT Arlington released a comprehensive article on AI’s water use titled “Making AI Less ‘Thirsty’: Uncovering and Addressing the Secret Water Footprint of AI Models”, which gives an overview of AI water use and data center cooling processes. Cooling towers, which are commonly used for AI data centers, require “continuous clean freshwater replenishment,” as they can only recycle water a handful of times. This means a great deal of local potable water is evaporated. According to “Making AI Less ‘Thirsty’”,  the “global AI demand is projected to account for 4.2 – 6.6 billion cubic meters of water withdrawal in 2027, which is more than the total annual water withdrawal of Denmark.” Chip manufacturing and energy generation for data centers also use large amounts of fresh water. This is especially concerning in the context of the ongoing global water crisis. According to UNESCO, half of the world experiences “severe water scarcity” for at least part of the year, and climate change is exacerbating water scarcity issues.

While the energy use and emissions of AI have a clear global impact, a great deal of the detrimental effects of AI data centers are local. The massive use of fresh water puts strain on local communities already facing water scarcity. Data centers also cause air pollution and thermal water pollution, and stress local energy grids. They can also produce toxic solid waste, which, according to the UN Environmental Program, includes mercury and lead. In 2024, AP News published an article highlighting protests against AI data centers in Virginia, where locals are unhappy with the development of data centers so close to their homes.

Yet, despite its massive detrimental impact on the environment, AI does have the potential to be an incredibly useful tool in climate change adaptation and mitigation. According to Scientific American, AI is becoming a useful tool in natural disaster relief. The program xView2 can use satellite imagery to identify infrastructure damage, reducing risk for human first responders. In an interview transcribed by UPenn’s Environmental Innovations Initiative, professor Cary Coglianese explained that AI can be used to manage complex electric grids by forecasting energy supply and demand. This would make it easier to transition from large fossil fuel plants to a larger number of small energy distributors, such as solar panels. Coglianese also highlighted AI’s ability to create complex forecasts and track progress on climate goals.

AI developers have a moral imperative to design technology that will benefit humanity, rather than simply innovating for the sake of novelty and financial gain. AI has the potential to improve lives all over the world, but is currently being utilized for detrimental causes.

Given the sizable environmental impact of AI’s development and use, it should not be used in ways that do not improve human wellbeing, such as AI models that mimic non-essential creative generation. Instead, regulations should be put into place to promote productive uses of AI technology. 

AI is most effective at combating climate change when utilized in a globally cooperative manner. The 2024 Article “Ecological Footprints, Carbon Emissions, and Energy Transitions: the Impact of Artificial Intelligence”, which analyzed panel data from 67 countries, found that AI development has a positive impact on carbon emissions by way of “efficiency improvements and process optimizations enabled by smart technologies,” but that the “anti-globalization and ‘trade protection’ that have emerged in recent years are not conducive to the sustainable development of all countries.”

AI is currently underregulated. The 2024 UN report Governing AI for Humanity concluded that there is currently “a global governance deficit with respect to AI.” According to the report, the majority of international AI regulations are currently comprised of norms and suggestions, which are only voluntarily adhered to. The report urges the pooling of information, recommends the creation of an international scientific panel on AI, and warns against leaving AI deployment and development to the “whims” of markets. Additionally, the UN Environmental Program published an issue note on artificial intelligence, with recommendations that AI developers and governments globally should adopt. This includes the standardization of measuring and reporting methods relating to the environmental impact of AI models, and the “adoption of green data centers” and renewable energy sources. 

With the proper regulation, goals, and transparency, AI has the potential to be a positive tool for climate change. However, its current environmental impact needs to be considered in development and policy decisions moving forward. Moving forward, AI innovations should be globally cooperative and sustainability focused, keeping in mind humanitarian efforts and human wellbeing.