Your "AI Footprint": The Environmental Costs of AI Data Centers
Data centers focused on AI (referred to here as AI data centers) require substantial amounts of land, energy, and water in order to fulfill the following core functions:
Ensure continuous power.
Provide continuous cooling.
Guarantee continuous operation.
A recent study conducted by researchers at Virginia Tech sheds some light on concerns related to the electricity and water demand of AI data centers:
More specifics on the amount of each resource, including land, required for AI data center operations are provided below.
Land - According to dcmap.us, there are a total of 3,269 operational data centers in the U.S., while an additional 1,244 are being planned and 438 are under construction. These numbers far exceed any other country on the planet by almost 10 times (source). The average size of new data centers in 2025 was 200,000 square feet, double what it was in 2016 and growing rapidly (source). The increase is due to a higher percentage of large and hyperscale AI data centers being built, the size of which can range from 50,000 to 100,000 square feet for a large data center to more than 1,000,000 square feet for a hyper-scale data center similar to those used by Google and Meta. (source)
Electricity - According to a 2024 Berkeley report, U.S. data centers used approximately 176 terawatt-hours (TWh) of electricity in 2023, which represents about 4.4% of total U.S. electricity consumption. Looking into the future, researchers from Berkeley also estimate that total electricity usage by data centers in the year 2030 will increase to between approximately 520 and 840 TWh or about 10 to 15% of total U.S. consumption. (source) According to the video above, it has even been estimated that AI data centers alone will consume as much electricity as 8 New York Cities by the year 2028.
Water - It is difficult to determine the amount of water required to build and operate an AI data center. Water is needed to extract and refine the materials used to build a data center, manufacture the processor chips, generate the electricity required for operation, and provide continuous cooling of equipment. In addition to the concern over total water consumption is the fact that data centers require pristine water, which limits them to using drinking water. This can have extremely detrimental effects on already water-stressed communities that are located nearby. Focusing for now on operations, according to the Environmental and Energy Study Institute, large data centers can consume up to 5 million gallons of water per day or 1.8 billion gallons per year (equivalent of up to 50,000 U.S. households), though data centers solely focused solely on AI can consume much higher amounts. To bring this down to the level of a typical AI user, researchers at the University of California, Riverside, estimate that (1) 10 to 50 simple queries using an AI model such as ChatGPT would require about 500 ml (~17 fl oz) of water and (2) each 100-word AI prompt requires roughly a standard bottle of water. Prompts involving the design of images and other graphic media would require substantially more water. Given that billions of users are prompting AI every minute, water consumption quickly adds up.
Helpful Hint
Attempt to minimize your "AI footprint" by reducing your use of AI chatbots such as ChatGPT and AI agents such as Meta's Muse.
Estimate your current "AI footprint" in terms of electricity and water usage due to AI image and text queries. A simple online calculator can be found here.
Try to continuously reduce your "AI footprint" from week to week as a way to not only help the environment but also to lower your overall dependence on AI. Make sure to use the same calculator when estimating future footprints.
Excellent follow up to the previous post related to Data Centers.
Thank you Jason for this useful information. More environmental concerns will be figured out soon.