|
Aerial view of data centers being built in Leesburg, VA. Credit: Gerville/2024
Overview
The ClassACT HR73 Environment and Climate Change Working Group is acting to warn our classmates and others about the impact of explosive growth of data centers across the United States. Here we concentrate on climate and water impacts. Subsequent alerts will deal with other pollution and health effects, electricity costs, and the added stress on electric grids.
What is Happening Today
Data centers receive, store, and process vast amounts of information. Some data centers are also the flesh and the bones of Artificial Intelligence. They house the servers, storage and network equipment that enable AI activities, such as ChatGPT, to be deployed and to grow in complexity and scale. All data centers consume large amounts of energy. Their cooling systems run without interruption 24 hours per day, 365 days a year.
Champions of data centers in the tech world argue that without a steady increase in their number and physical size, businesses and government functions will falter and AI will never reach its full potential; scientific research will be stymied and national security will be harmed. Federal and state agencies will not be able to store the massive amounts of data essential to fulfill their duties in the 21st century. These champions maintain that ordinary denizens of the digital world will lack adequate cloud storage, and that their online AI queries will go unanswered.
Accelerated by AI developments, data centers are proliferating. With over 5,000 already in the US in 2026, more are under construction and more are in planning. Newer ones often consume more energy than earlier ones. One large center devoted to AI can use as much electricity as a city of 10,000 homes!
At present, half or more of the electricity used by data centers is generated by burning fossil fuels, natural gas and coal. Contributions to climate change and pollution are accelerating. Companies that run data centers talk about building their own generating plants, including nuclear ones, but that has not yet happened. Microsoft says it will pay its own way and not burden communities with added electricity costs. Imagine that—paying for the electricity you actually consume! But it's a novel commitment that no other company has made.
Data center cooling systems often use huge volumes of water, in some rural locations pumped directly from aquifers crucial for public water supply and agriculture. They take up large tracts of land, and clusters of them can dominate communities in places like Northern Virginia.
No single decision at the federal, regional, or state levels can determine the environmental impact of data centers. Multiple factors are in play with trade-offs among them: business interests, construction jobs, maintenance jobs, economic competition across the globe, local economic development, electricity generation and consumption, electricity rates (individual consumer vs. data center vs. industrial), electric grid reliability, water consumption, as well as water, air, and light pollution. The players are huge corporations, state and local governments, citizens groups, public service commissions (where they exist), power companies—and the public.
Adverse Impacts on Climate Change
From 2018 to 2024 data center carbon emissions tripled! At current rate of AI growth, by 2030 an additional 24 to 44 million metric tons of carbon dioxide will be emitted. Despite pledges of giants like Google and Meta to achieve zero carbon emissions by 2030, their zeal to beat each other in the AI race may make that goal impossible.
Depletion of Water and Land
Hyperscale data centers can demand as much as five million gallons per day for cooling – the equivalent of the water used by a city of 50,000 people. They also gobble up massive tracts of land with arable topsoil. The largest hyperscale centers can occupy areas the size of 17 football fields. Increasingly they are placed in rural areas where their stormwater runoff threatens remaining farms, and their thirst for water dries up local wells.
|
