If you are following the conversation around AI and the environment, it can feel overwhelming. The headlines can be alarming as they talk about massive energy consumption, growing water usage, and technology scaling faster than our ability to understand its full impact. And at the same time, AI is becoming increasingly difficult to avoid in how we operate day to day. This is a real problem for nonprofit leaders trying to reconcile those two realities. They must balance the environmental impact of AI with the need to use it responsibly and ethically. If you’re a nonprofit leader feeling uneasy about using AI, the environmental concern is real, but the levers that actually affect the environment have more to do with policy and regulations than whether or not you use AI. 

The Footprint AI Is Already Leaving 

Research published in the peer-reviewed journal Patterns in 2025 estimated that AI systems alone could be responsible for between 32.6 and 79.7 million tons of CO2 that year, the AI subset of the broader data center footprint. New York City emitted 52.2 million tons of CO2 in 2023, which gives some sense of the scale we are talking about. Google’s own sustainability report acknowledged that its total greenhouse gas emissions increased 13 percent year over year in 2023, driven primarily by data center energy consumption, leaving its emissions nearly 50 percent higher than they were in 2019. Microsoft has reported similar pressure, with total emissions 23.4 percent higher in fiscal year 2024 than its 2020 baseline. 

Beyond carbon, data centers consume enormous amounts of water to cool their servers, either directly on site or through the power plants supplying their electricity. Roughly 20 to 50 AI queries can consume about half a liter of water, and that adds up across hundreds of millions of daily interactions. The 2025 Patterns study projected AI’s total water footprint could reach between 312.5 and 764.6 billion liters in 2025 alone. In Arizona, Nevada, Utah, and parts of Texas where many data centers are concentrated, water is already a contested resource, and the communities living near these facilities are the ones absorbing that pressure. 

The Trajectory We Are On 

If today’s numbers are concerning, what is coming is significantly worse if nothing changes. The IEA projects that data center electricity demand could double by 2026, reaching roughly 1,000 TWh, equivalent to Japan’s entire annual electricity consumption. Global data center emissions currently sit at around 200 million metric tons of CO2 per year. Financial research from Morgan Stanley projects that will triple to 600 million metric tons annually by 2030, driven primarily by AI growth. To put that in perspective, South Korea, the world’s tenth largest economy, emitted approximately 574 million metric tons of CO2 in 2023. Within six years, data centers alone could be emitting more than entire major nations do today. 

Water consumption is on a similar trajectory. Researchers at UC Riverside and UT Arlington project AI’s global water footprint could reach 4.2 to 6.6 billion cubic meters by 2027, roughly ten times what is projected for 2025 and four to six times the annual water usage of a country like Denmark. 

We Have Been Down This Road Before 

These numbers can feel overwhelming, but they are also not unprecedented. Every major technology of the last century came with an environmental cost that society initially underestimated and eventually got better at regulating. Even though we have made some progress, it has not come easily and there is still a long way to go. 

Consider cars, a technology advancement that has propelled society forward but has come with its own environmental costs and battles. Transportation accounts for roughly 28 percent of US greenhouse gas emissions, with passenger vehicles responsible for more than half of that. The US transportation sector alone produces approximately 1.8 billion metric tons of CO2 per year. To put that in perspective, all global data centers combined, including every AI workload running today, currently produce approximately 200 million metric tons of CO2 annually according to Morgan Stanley. The entire US transportation sector alone produces roughly nine times what global data centers do today. AI’s footprint is real and growing, but it operates at a fundamentally different scale than the technology most of us use every day without a second thought. 

The internet itself, before AI became a major workload, has long been a meaningful contributor to global emissions. Estimates of the internet’s total share of global greenhouse gas emissions range from roughly 2 to 4 percent depending on methodology. Streaming video is consistently identified as one of the largest contributors within that footprint. Every time someone watches Netflix, attends a video call, or streams a webinar, there is an environmental cost most people never think about. AI is being built on top of this existing infrastructure and is accelerating demand that was already growing. The uncomfortable truth is AI is just the latest addition to an environmental problem we haven’t been paying enough attention to up until now. 

Learn how nonprofit leaders can approach AI responsibly and effectively

Let’s talk about what we can do for your organization 

Connect

So, what can we do? 

It helps to look at what has worked when other technologies pushed up against environmental limits. The science behind every policy win in this post existed before the policy did. What moved legislators was never the data alone. It was people connecting environmental harm to something tangible in their own lives and communities, building coalitions across unexpected lines, and sustaining pressure over years rather than news cycles. 

Let’s look at the Clean Air Act, which is one of the first pieces of environmental protection legislation that has made a lasting impact on society. An estimated 20 million Americans showed up for the first Earth Day on April 22, 1970, organized by Denis Hayes, a 25-year-old graduate student coordinating from a basement apartment in Washington DC. They did not know they were helping pass the Clean Air Act eight months later. They just showed up. The political impact was swift. By December 1970, Congress had authorized the creation of the EPA and strengthened the Clean Air Act, with the Senate passing it nearly unanimously. Twenty years later, the 1990 amendments that addressed acid rain were driven by decades of advocacy from environmental organizations and the communities in the Adirondacks who watched their lakes and forests die from acid precipitation. The Senate passed the 1990 amendments 89 to 11. It took a generation of sustained effort to get there. 

The results matter as much as the fight. The EPA’s own retrospective analysis found that the 1990 amendments alone prevented 160,000 premature deaths in 2010, a number projected to rise to 230,000 by 2020. Six major air pollutants tracked by the EPA decreased by an average of 78 percent between 1970 and 2020, during the same period that the US economy grew 285 percent. Environmental protection and economic growth were not in opposition. They happened simultaneously. 

The Energy Star program, launched by the EPA in 1992 as a voluntary efficiency certification, took a different approach. Rather than mandating change, it created a market incentive for efficiency that manufacturers and consumers adopted because it made economic sense. Smart policy design, one that aligns environmental goals with economic self-interest, can move faster than regulation alone. Since its inception, Energy Star has saved Americans more than $500 billion in energy costs and prevented 4 billion metric tons of greenhouse gas emissions. 

Fuel economy standards for vehicles tell a similar story of hard-won progress. Tightened significantly between 2007 and 2012 after decades of industry resistance, they passed because environmental advocates built a coalition that crossed traditional lines, including veterans’ groups citing energy independence, public health organizations citing tailpipe emissions, and consumer advocates citing fuel costs. Politicians could not attribute the pressure to a single ideology. The standards drove average new vehicle fuel economy from 25.7 MPG in 2005 to 36.0 MPG by 2022, and the MY 2017-2025 standards are projected to avoid 1.8 billion metric tons of CO2 emissions over the lifetimes of covered vehicles. 

AI follows the same pattern as every transformative technology before it. A real environmental cost, growing faster than the public conversation can keep up with, and an industry that benefits from how slowly accountability tends to catch up. What changed the trajectory in every prior case was not better data. It was people who decided the issue mattered enough to organize around. That is the work in front of us now, and history suggests it works when enough people commit to it. 

Small Actions, Big Impact 

Most of us do not have the bandwidth for full time activism. There are still simple ways to show up for the environment that have consistently made a difference. 

First, educate yourself and understand the issues well enough to speak from a place of knowledge rather than anxiety. Fear-based conversations tend to generate more heat than progress. Having open and understanding conversations where you can validate someone’s fears and help them get to a place of understanding is how we build the connections that can help drive change. 

The second is to find and support organizations that are already doing this work like the NAACP Environment & Climate Justice initiative, Food & Water Watch, and the Climate Justice Alliance. These groups fight for data center disclosure requirements, pushing AI companies to honor their environmental commitments, and representing communities dealing with the direct impacts of unchecked data center growth. You do not have to lead that work to be part of it. Donating, amplifying their message, or simply staying connected keeps the pressure alive. 

The third is your vote. Every policy win described in this post required politicians who were willing to prioritize their communities over corporate interests. The Clean Air Act passed 73 to 0 in the Senate in 1970, which tells you this has never been purely a partisan issue. Vote for politicians who have shown they can hold corporations accountable, and make sure the ones who haven’t, hear about it at the ballot box. 

AI is another piece of technology that will help advance our society, and like every transformative technology before it, it comes with a cost. That cost has largely been placed on society by corporations operating with very limited regulation or oversight, and the trajectory is only going to accelerate if that does not change. That is not an argument against AI or against progress.   Whether any individual nonprofit leader uses AI will not determine AI’s environmental impact. The question is whether enough of us show up for the policies, organizations, and elections that can make a difference. The Earth has never been saved by individual restraint. It has been saved by collective pressure. 

 
 
References 

  1. EPA. (2011). The Benefits and Costs of the Clean Air Act Amendments of 1990. US Environmental Protection Agency. https://www.epa.gov/clean-air-act-overview/benefits-and-costs-clean-air-act  
  1. EPA. (2023). Energy Star Program Overview and Annual Report. US Environmental Protection Agency. https://www.energystar.gov/about/origins_mission  
  1. IEA. (2024). Electricity 2024: Analysis and Forecast to 2026. International Energy Agency. https://www.iea.org/reports/electricity-2024  
  1. Li, P. et al. (2023). Making AI Less Thirsty: Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv:2304.03271. https://arxiv.org/abs/2304.03271  
  1. EPA and NHTSA. (2012). Final Rule: 2017-2025 Light-Duty Vehicle Greenhouse Gas Emission Standards and CAFE Standards. https://www.epa.gov/regulations-emissions-vehicles-and-engines/final-rule-2017-2025-light-duty-vehicle-greenhouse-gas  
  1. University of Michigan Transportation Research Institute. (2022). Average Fuel Economy of New Vehicles Sold in the US. https://umtri.umich.edu/our-results/publications/average-fuel-economy-new-vehicles-sold-united-states  
  1. Google. (2024). Environmental Report 2024. Alphabet Inc. https://sustainability.google/reports/google-2024-environmental-report  
  1. Patterson, D. et al. (2021). Carbon Considerations for Large Language Model Training. Google Research. https://arxiv.org/abs/2104.10350  
  1. De Vries-Gao, A. (2025). The carbon and water footprints of data centers and what this could mean for artificial intelligence. Patterns. https://www.sciencedirect.com/science/article/pii/S2666389925002788  
  1. Microsoft. (2025). 2025 Environmental Sustainability Report. Microsoft Corporation. https://blogs.microsoft.com/on-the-issues/2025/05/29/environmental-sustainability-report  
  1. EPA. (2024). Inventory of U.S. Greenhouse Gas Emissions and Sinks 1990-2022. https://www.epa.gov/ghgemissions/inventory-us-greenhouse-gas-emissions-and-sinks  
  1. Morgan Stanley Equity Research. (2024). Global Data Centers: Sizing and Solving for CO2. Reported by Reuters and Data Center Dynamics, September 2024. https://www.datacenterdynamics.com/en/news/morgan-stanley-data-center-industry-will-emit-25bn-tons-of-co2-by-2030  
  2. EDGAR/JRC. (2024). South Korea CO2 emissions data. https://www.theglobaleconomy.com/South-Korea/carbon_dioxide_emissions