IN MARCH The Information reported that Microsoft was negotiating with OpenAI, the creator of ChatGPT, to spend $100 billion on a massive data center in Wisconsin. Code-named “Stargate,” this center would consume five gigawatts of electricity at full capacity, enough to power 3.7 million homes. This is comparable to the output of Plant Vogtle, a nuclear power station in Georgia that cost $30 billion to build.
Stargate is still in the planning stages, but the scale of the proposal highlights a significant issue: AI is an energy hog. For decades, American electricity consumption has hardly grown, partly due to advances in energy efficiency. In 2023, total electricity consumption fell slightly from 2022 levels.
However, a report from Grid Strategies published in December predicts that US energy demand will rise by up to 5 percent over the next five years, partly due to new data centers for AI. To meet this demand, utilities propose building new fossil fuel power plants and dismantling climate legislation.
Environmentalists see this as a step backward. AI, initially seen as a problem-solver, now risks worsening global warming.
Since the internet became an essential part of daily life, computing’s energy demands have surged. Every digital activity is processed in data centers filled with computer hardware. These centers are the physical manifestation of the cloud.
For the past 20 years, data centers have increasingly handled computation. Corporations and governments rent computing services from Big Tech for efficiency and cost-effectiveness. However, as anthropologist Steven Gonzalez Monserrate notes in The Cloud Is Material: On the Environmental Impacts of Computation and Data Storage, “heat is the waste product of computation.” Data centers consume significant energy to keep computer chips cool, and the new generation of AI software is particularly energy-intensive.
Michelle Solomon, an analyst at Energy Innovation, calls the AI power crunch “a litmus test” for a society threatened by climate change. “Are we going to grow the right way?” Solomon asks. “Or will we keep coal plants online and build new gas plants?”
JUST THREE COMPANIES—Microsoft, Amazon, and Alphabet—host two-thirds of the cloud. Known as “hyperscalers,” their data centers consume enormous amounts of power and water. Despite their environmental claims, the negative consequences of AI energy consumption are often unacknowledged.
The hyperscalers have committed to a net-zero future, using the Greenhouse Gas Protocol to track emissions and investing billions in renewable energy. According to S&P Global, they have driven significant renewable energy growth, contracting for about 45 gigawatts globally.
Historically, the data center industry has grown without straining energy resources. A 2020 study found that from 2010 to 2018, data storage increased 26-fold, servers grew by 30 percent, but energy consumption rose only 6 percent. However, Andrew Chien of the University of Chicago argues that the exponential growth of hyperscale data centers, fueled by AI, presents a severe problem.
Big Tech’s electricity consumption has surged. Between 2021 and 2024, the number of data centers globally jumped from around 8,000 to 10,655. A spokesperson for Amazon told The Wall Street Journal that a new data center is being built somewhere in the world every three days.
The cost of training AI software is doubling every few months, making AI extremely expensive. John Hennessy, chairman of Alphabet, told Reuters that it was 10 times more expensive to ask Google AI a question than to use a traditional keyword search.
TODAY, AI-PROPELLED data center growth is booming, especially in Northern Virginia’s Data Center Alley, where energy demand is growing at 20 percent a year. Dominion Energy has asked for permission to build additional gas-fired power plants and supported a proposal to withdraw from the Regional Greenhouse Gas Initiative.
Other regions are seeing similar appeals for extended fossil fuel use. In Ohio, AEP’s CEO announced plans to sue the EPA over new coal plant regulations. Meanwhile, Maryland’s Governor Wes Moore signed legislation exempting backup diesel generators from environmental review.
Despite the negative environmental externalities, Big Tech continues to expand. “If AI raises their revenue high enough,” says computer scientist Roy Schwartz, “they will gladly pay the energy bill.”
The lack of regulation is concerning. In Virginia, 17 bills aiming to regulate the industry were defeated. Senator Ed Markey’s Artificial Intelligence Environmental Impacts Act of 2024 is stalled in committee. Gonzalez Monserrate emphasizes that the cloud is material and has a cost, highlighting the need for regulation.
Original Story at www.sierraclub.org