Experts discussed the negative environmental impacts of artificial intelligence (AI) and emphasized measures that can be taken to mitigate harmful effects Tuesday at the AI’s Energy Future Forum at the Wisconsin Energy Institute.
Expansion is happening without much public awareness. Two new massive data centers, or buildings to house computer systems, were recently established in Wisconsin. Digital Power Optimization partnered with a subsidiary of Bellerud to build a $200 million AI computing center in Central Wisconsin. And tech giant Microsoft is in the process of building a $3.3 billion dollar data center in Mt. Pleasant.
AI is costly to operate. Large quantities of water are needed to cool the data centers that power AI operation. Meanwhile, vast amounts of energy consumption — 11.7% of the total U.S. power demand — strain the supply of electricity on the nation’s already aging power grid.
ChatGPT, which is widely used by students, consumes 17 ounces of water for every five to 50 prompts.
AI data centers are a cause for concern among climate activists because of their excessive use of water to cool their systems and energy, which doesn’t always come from environmentally friendly sources. Costa Samaras, director of the Scott Institute for Energy Innovation at Carnegie Mellon University, called for power usage to align with environmentally friendly goals during the forum.
“Why do we care about data center energy? [Because] we have to get to zero emissions. We have to get to zero air pollutants. We have to get zero waste,” Samaras said.
Experts also discussed the responsibility of companies who utilize AI data centers in creating their own energy. At any point, an AI provider could switch gears to some newer, less power intensive tech that doesn’t use any of the infrastructure just built. Tyler Huebner, an energy market developer at Google’s central U.S. division, said companies have a responsibility to create their own power when building new data centers.
“Historically, we have tried to find utility companies that have extra power generation,” he said. “Essentially, there are no utility providers like that in the U.S. They say, ‘we have no spare power plant.’ We have to show up; it is our business risk, and [no one else should bear that risk.]”
Google’s current clean energy model endorses building new sources of clean energy near AI data centers. The multinational corporation has invested in geothermal, nuclear and strategic placement of data centers near clean energy sources.
Huebner stated that “64% [of the energy] in the grids we operate on is being matched with clean energy in those hours.”
“We are actively working on that, basically across the globe, at all times,” he said.
Amy Barrilleaux, communications director for the clean energy advocacy organization Clean Wisconsin, echoed similar sentiments. Barrilleaux called on state lawmakers to enforce regulations on data centers trying to work within the state.
“We should demand that any entity that comes into our state brings benefits and does not harm our communities,” Barrilleaux said. “We are finding out about AI data centers after the deal is already done. We need to expect more of our lawmakers. These impacts will last for decades.”
These warnings come as We Energies, an energy provider for Wisconsin and surrounding states, announced a new gas-fired power plant in Kenosha this December.
The project could produce up to 590,000 tons of greenhouse gas emissions annually including carbon monoxide, nitrogen oxides and particulate matter. It could cause 351 premature deaths, 858 asthma cases and 93 heart attacks amounting to $3.6 to $5.7 billion in health costs, according to an analysis from Healthy Climate Wisconsin.
Samaras, however, was optimistic about the possibilities to regulate the boom in consumption, both locally and nationally.
“There are key important policy measures that states and local entities need to be leaders on, Samaras said. “Making sure that our electrified economy works for everybody is doable.”
