AI and Climate: Experts Debate Technology's Environmental Impact

The emergence of artificial intelligence holds promise for advancements in clean energy by offering novel approaches to reduce emissions and adapt to climate impacts. However, the computational power required to operate generative AI is itself a substantial energy consumer.

To explore this intricate relationship between AI and climate, prominent experts from technology and energy corporations convened at an event at Newsweek headquarters to engage in a dialogue about whether AI is a climate ally or adversary.

AI in Healthcare: How Machine Learning is Revolutionizing Diagnosis and Treatment

As part of Newsweek's new Horizons event series, the panel discussion took place on Wednesday, September 25, amidst Climate Week in New York City. The panel consisted of five participants and was moderated by Newsweek's Environmental and Sustainability Editor Jeff Young.

Panelists included Microsoft Vice President of Energy Bobby Hollis; Bezos Earth Fund Director of AI and Data Strategies Amen Ra Mashariki; Salesforce Executive Vice President & Chief Impact Officer Suzanne DiBianca; Duke Energy Managing Director, ESG & Sustainability Heather Quinley; and MIT assistant professor Priya Donti.

Their discourse examined how the advantages provided by AI-assisted breakthroughs could potentially offset the associated power demands necessary to generate them.

The topic is gaining relevance daily—the California-based Electric Power Research Institute (EPRI) published a study in May indicating that data center power consumption could account for up to 9 percent of U.S. electricity generation by 2030. This figure is more than twice the amount currently utilized.

Last month, the White House expressed confidence that the nation would be capable of meeting the energy requirements related to AI.

"The rapid growth of AI is posing this significant question: Will we have sufficient energy to power AI?" Energy Secretary Jennifer Granholm told reporters in a briefing on Friday, August 9. "We emphatically affirm yes, we will."

That optimism was echoed by Hollis, who stated AI was "definitely a benefit" regarding its relationship to climate change.

Hollis, who has led Microsoft's global energy team since 2023 and previously held energy-related positions at Meta and Apple, informed the audience at Newsweek that "when you consider wildfire mitigation, when you consider storm impact, when you consider the factors that will truly affect every human being on this planet, that's where you observe AI as a tool that literally can preserve lives."

Microsoft's clean-energy aspirations have been complicated by AI. In May, a Microsoft sustainability report revealed that the company's aggregate emissions for 2023 were up 29 percent compared to 2020—a fact that it attributes largely to the expansion of data centers powering generative AI.

Just last week, Microsoft entered into an agreement with electric utility company Constellation to supply its data centers with nuclear power, a move that may lead to Pennsylvania's Three Mile Island being reactivated.

"Our entire portfolio of AI has accelerated and increased the need for carbon-free energy, so that necessitates us to look beyond conventional solutions, and that for us means examining some of the technologies that have been present for an extended period," Hollis said when questioned about potentially restarting a Three Mile Island reactor, now called the Crane Clean Energy Center. "It's a Pennsylvania energy resource that operated through 2019, had an exceptional operational record and possessed the capability to deliver carbon-free energy."

In his role at Bezos Earth Fund, Mashariki assists groups in utilizing data to address complex environmental issues. Discussing how his organization came to focus on AI, he said, "We didn't intend to fund AI work. We aimed to actually fund climate and nature work."

Mashariki shared a well-received example of how AI is helping assist in the removal of invasive pythons from the Florida Everglades. The process of removing the animals illustrates a broader point about how to employ AI to enhance impact.

"You'll often hear about projects that claim to use satellite imagery and AI deep learning to identify python locations. But for us? We're far less interested in understanding where the pythons are. We're more interested in understanding their locations in a way that facilitates the organizations that would typically intervene [to remove them]," he elaborated. "We're not interested in merely identifying locations and measuring things anymore. We're more focused on measuring and monitoring to drive impact."

DiBianca leads Salesforce's stakeholder capitalism strategy, which encompasses the company's global corporate relations, sustainability and ESG strategy. She stated, "We are developing LLMs. But our mantra is, you don't need to know how to compose poetry to resolve an issue in healthcare. So the models you're observing today are extensive models. They're requiring significant compute power, and you don't always require that much."

Salesforce recently surveyed nearly 500 sustainability professionals on their perspectives regarding AI and found that while 40 percent were concerned about AI's impact on their company's sustainability efforts, 60 percent expressed optimism that AI's benefits would counteract those negative effects.

DiBianca, who described AI as a benefit "100 percent," added that "what I want to avoid is the next digital divide, where educators, public schools, nonprofits lack access to this technology, so we're undertaking substantial work in this area."

At Duke Energy, one of the largest energy holding companies in the U.S., Quinley's team establishes ESG strategy. She viewed AI as "somewhat of both" a benefit and a drawback.

"We've been utilizing AI, leveraging its capabilities since 2017, and it helps enhance our efficiency," Quinley said. "We've been employing it for leak detection and faulty meters, energy theft, all the way to plant maintenance and storm detection and outage management. So it's benefited us, it's benefited our customers."

Quinley, however, also addressed the increased energy demands stemming from the growth in AI.

"What we're observing is our customers are experiencing significant load growth," she explained. "We're witnessing significant load growth from data centers and advanced manufacturing. As we've indicated for data centers in the coming years, they will account for 25 percent of our new projects, and advanced manufacturing—EVs, batteries, semiconductors, etc.—will account for about two-thirds."

MIT's Donti, who is also co-founder and chair of the research group Climate Change AI, provided a response on how AI itself is not yet a benefit or drawback regarding its impact on the environment.

"AI isn't the agent in this equation. People and organizations are. AI is an accelerator of what we do across society," she informed the audience. "If we're committed to combating climate change, it will be beneficial. If we're not, it's going to be detrimental."

Researchers, including Donti at Wednesday night's discussion, have stated more transparency is needed from tech companies to fully examine the issue of how much energy AI models are consuming. Legislation from Massachusetts Senator Ed Markey now pending in Congress calls for additional disclosure.

"I believe that part of the reason the challenge seems so daunting is because the nature of the challenge is, I think, not necessarily yet that well understood," Donti said. "These kinds of dynamics are actually quite difficult to comprehend if you don't have transparent access to data about the uses and then it means that you don't know how to proceed. … I believe further transparency and data sharing from multiple entities on these kinds of matters are going to be really instrumental."