Artificial intelligence (AI) and data centers are reshaping our relationship with water and energy, yet water feels like an afterthought for many tech companies. Public sentiment is full of questions and concerns over environmental impacts, and most voters just don’t know enough to form a strong opinion, according to recent polling from Data for Progress. As we read alarming headlines about drying taps and rising energy bills near new data centers, and more of our partners are wading into local data center decisions, we thought it would be helpful to summarize the facts and conversations happening now.
AI's data centers
Construction of data centers has boomed in the past ten years, and is predicted to exponentially increase with President Trump’s executive order expediting large-scale data center construction. At the same time, the use of generative AI has increased as internet users have turned to OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini, Anthropic’s Claude, and other models.
It’s important to note, AI isn’t a “character” in this story. AI doesn’t build data centers, nor does it decide how they will be powered or what water source they will use. Tech industry leaders from OpenAI, Amazon, X, Microsoft and the like are courted by states with tax incentives worth billions and other financial benefits. Their companies are drawn by cheap electricity rates to fuel increasingly energy-intensive hardware, which contributes to electricity projections doubling or tripling by 2028. Data centers require stable, 24/7 power, and utilities across the country are adding new gas plants and delaying the retirements of coal power plants to satisfy their non-stop demand, exposing a limit to tech companies’ clean energy pledges.
Tech giants consider water a “cheap” resource, but when they dip into communities’ water sources, they can end up paying lower rates than residents. In 2023, it took 66 billion liters of water across the U.S. to cool down data centers for the year. However, these calculations aren’t sufficient for informed decision making. There are additional long-term financial and environmental costs associated with upgrading water infrastructure, developing new water sources, conserving or recycling water in areas dealing with drought, and indirect water usage associated with electricity generation. Without these factors, data centers’ water footprint is understated or even hidden from the public.
Is it a lot of water?
Is 66 billion liters of water a lot? Does a ChatGPT inquiry throw out a bottle of water? Are these statistics accurate? It depends. Water is a hyperlocal issue and numbers without context don’t tell us enough. While 66 billion liters of water in a year sounds high, people in the United States use about 9 billion liters per day watering residential lawns and gardens. To bring this statistic down to earth, consider this: a large data center can use 5 million gallons of water per day, the same as a town of up to 50,000 people.
The main concern regarding AI data center’s water use isn’t simply about the quantity of water, but of how, where, and when water is being used. Since 2022, dozens of data centers have been built in drought-stricken and water scarce areas across the country, including California, Arizona, and Texas. Data centers use drinking water to cool their data centers and when they use more water and power on hot days, it’s a serious strain to already stressed local water supplies and electrical grids.
Opportunities for innovation
There are currently no federal regulations for water use or environmental impact reporting for data centers. Senator Edward Markey of Massachusetts introduced the bipartisan Artificial Intelligence Environmental Impacts Act of 2024, which would set federal standards and voluntary reporting guidelines to measure the environmental footprint of AI data centers. State lawmakers in California, Connecticut, and New Jersey have also introduced their own bills.
Locally, communities are contending with the effects of proposed or constructed data centers. In addition to growing opposition, local leaders are calling on tech companies to invest in recycling and re-using water, or setting up specific policies like water-sharing agreements. Industry speculation and investment in new technologies for waterless cooling methods and electricity generation have emerged out of data centers’ water use concerns. However, such technology has not yet been widely adopted by the industry and is very expensive, begging the questions: who should be on the hook for the cost of developing and implementing new technology?
In order to make informed decisions about AI and data centers, policymakers and communities need to know exactly how much water is being used, where, when, from what source, and at what cost. The lack of reporting requirements or standards enables a lack of transparency from tech companies about their water and energy use and whatever data is released is difficult to combine or compare across AI models, types of inquiries, and data centers.
It’s clear that more communities and organizations are facing decisions around AI use and data centers. An understanding of its water and broader environmental impacts is a glaring gap in public discourse and a barrier to informed decision-making across multiple levels of society. This is also due to the dynamic and rapidly evolving trajectory of AI and data centers, which has emerged as an area of interest to us at the Water Hub as well. This issue briefing focusing on the water realities of AI is only a start; while there’s no clear or uniform approach to address it, community organizations and reporters alike have an important role to bridge the information gap between tech companies and impacted communities.