What you need to know about the hyperscale data centers coming to the Midwest | Credit: Illustration by Shira Friedman-Parks
By Siri Chilukuri | Chicago Reader
The explosion of proposed data centers across Illinois has sparked a fierce, bipartisan backlash and even legislation to curb the long-lasting effects on people’s air quality, water quality, and energy bills. In town halls from Pekin to Joliet to Chicago’s southeast side, people have packed rooms and spent hours discussing the impact of potential developments on their communities. As residents grapple with project proposals, it’s never been more important to understand the impact of data centers.
Everything from the emails you send to the photos you save on your phone is stored in a data center. For decades, data centers have been central to the architecture of the Internet, especially the digital lives we lead today. But the rise of artificial intelligence, in particular generative AI—which uses computer models to produce text, images, videos, and more—is driving the development of facilities that use more energy and water than ever before. The size of these so-called hyperscale data centers, and the profit motives driving the rush to get them online quickly, can cause pollution.
Sarah Moskowitz, executive director of the Citizens Utility Board, has been working with utility customers for decades on issues of affordability and climate change. The CUB is an advocacy organization for Illinois ratepayers, which has been watching the rise of data center projects in Illinois and learning more in the process about what they actually do. For Moskowitz, the distinction between next-generation hyperscale data centers that have roiled communities and the existing data center infrastructure is an important one to make. At the most basic level, “a data center is a physical facility containing equipment to store, process, and disseminate digital information,” Moskowitz said. “A hyperscaler is basically a large version of that. Generally, we think of hyperscalers as gigantic facilities, largely deployed to process artificial intelligence.”
Even the term “hyperscale” is imprecise, though, according to Helena Volzer, senior source water policy manager at the Alliance for the Great Lakes. “There’s really no hard and fast definition of what hyperscale even means. It just means a large facility,” Volzer said. Economists at the University of Virginia expect an average hyperscale data center to be about 300,000 square feet by 2030.
Hyperscale data centers are energy intensive not just because of their massive size but also because of what they process. The power used by generative AI is staggering—the computational power needed to train the large language models that power chatbots, such as OpenAI’s ChatGPT or Anthropic’s Claude, and the power needed to perform basic functions once they’re up and running far outpaces the demands of existing data centers. One Goldman Sachs report estimates that AI will spur a 165 percent increase in data center power consumption by 2030.
“What makes the data center problem unique is the fact that these facilities are being built in a specific location, or specific locations, and they are being built en masse,” said Yury Dvorkin, an engineering professor at Johns Hopkins University. “[If] you put a lot of electricity demand in a very constrained location, what happens is that it’s harder for the grid to deliver power in that specific location.”
Report continues here.
Related: “Illinois lawmakers begin days of deep dives on data centers,” “Do you trust our Board of Trustees? We don’t. But you decide for yourself once we have finished. (Part 4),” “Do you trust our Board of Trustees? We don’t. But you decide for yourself once we have finished. (Follow-up),” “Do you trust our Board of Trustees? We don’t. But you decide for yourself once we have finished. (Part 3),” “Do you trust our Board of Trustees? We don’t. But you decide for yourself once we have finished. (Part 2),” “Do you trust our Board of Trustees? We don’t. But you decide for yourself once we have finished. (Part 1),” “110 Acre AI data center campus pitched to Village Board”