According to Gartner, 81% of HR leaders have already explored or implemented AI solutions for their organizations. Yet, many aren’t honed in on the multi-layered environmental impact of AI, which requires a massive amount of energy and water, while computing centers boast a powerful carbon footprint.
OpenAI CEO Sam Altman warned an audience about this issue at an event happening alongside the World Economic Forum meeting in Davos this week, saying that AI will consume much more energy than initially expected.
“There’s no way to get [to climate-friendly use of AI] without a breakthrough,” said Altman, according to reporting from Reuters. “It motivates us to go invest more in fusion.”
The environmental impact of AI
HR leaders can be viewed as force multipliers, as their decisions are amplified by the hundreds or thousands of employees who utilize workplace technology. Now, these leaders have to consider that power as they address the sustainability implications of adding AI to their tech stack.
“There are many social components that HR deals with,” says Brian Sommer, author of The Executive’s ESG Playbook. “The real twist is to apply that same thought process against suppliers.”
The ecological weight of AI computing has been in the headlines, but far less has been written about this aspect of artificial intelligence than, say, all the things people will do with it. However, the following stories highlight the behind-the-scenes resource burn of artificial intelligence that, experts say, HR leaders should consider as they weigh technology purchases.
Carbon footprint of gen AI
MIT Technology Review put it this way: “Generating 1,000 images with a powerful AI model is responsible for roughly as much carbon dioxide as driving the equivalent of 4.1 miles in an average gasoline-powered car.”
Energy consumption of AI servers
The New York Times recently reported on a peer-reviewed study about the growing energy footprint of artificial intelligence by Alex de Vries, a Ph.D. candidate at the VU Amsterdam School of Business and Economics.
“In a middle-ground scenario, by 2027, AI servers could use between 85 to 134 terawatt hours annually,” according to the article. “That’s similar to what Argentina, the Netherlands and Sweden each use in a year, and is about 0.5% of the world’s current electricity use.”
Water consumption of gen AI
Reporting from the University of California, Riverside, examined one of the world’s most precious resources, fresh water. “Run some 20 to 50 [ChatGPT artificial intelligence] queries, and roughly a half-liter, around 17 ounces, of fresh water from our overtaxed reservoirs is lost in the form of steam emissions.”
Considerations when purchasing HR tech
While most HR leaders aren’t in control of how much stress workplace technology places on the environment, they do influence buying decisions that could limit the purchase of ecologically irresponsible players.
Sommer, one of HRE’s Top 100 HR Tech Influencers, says that all tech purchases have an environmental cost. He notes that HR leaders, as the buyers of tech, need to consider how each potential vendor treats its resources, workers and community.
“In the past, buyers were mostly interested in the immediate, direct environmental impact of their purchases,” he says. This was focused on facilities housed within the walls of the company—for example, the heating cost of an office or the cooling costs of a server room.
“The purchase of cloud-based solutions—including newer AI capabilities—seemed to shift that responsibility to these cloud solution providers, but that assumption is not correct,” says Sommer. He points to a growing emissions concern involving computing data centers, where output continues to increase at an unprecedented rate. According to a 2021 study out of Virginia Tech, data centers are responsible for approximately 0.5% of total U.S. greenhouse gas emissions. (Around one-quarter of the world’s data centers are located in the U.S.)
He also notes the amount of cooling and other energy required to train large language models and process requests. “It’s not green if it’s going to be sucking up the water table or local lakes,” says Sommer. And a data center’s location can significantly affect water consumption, he says. The Virginia Tech study reports that one in five data center servers in the U.S. have a “direct water footprint” tied to water-stressed watersheds, while nearly half of servers are powered by facilities located within water-stressed regions.
ESG and reporting requirements
Sommer advises that environmental, social and governance (ESG) reporting must include that of vendors engaged by the organization—at every tier in the supply chain. Looking forward, he says, HR teams will need to compare the sustainability metrics of the companies providing new tech with AI components. For now, that might not be easy. “Some companies may not have all the data available yet, but that doesn’t mean their track record is good,” warns Sommer.
He sees a future where CHRO and sustainability officers work together more frequently to hash out the environmental impact of tech purchasing deals. He warns that HR and other business leaders can’t delegate responsibility for ESG reporting, even when vendors are providing cloud or artificial intelligence from an external standpoint. “You can’t get away with shoving off your ESG responsibility,” says Sommer.
To help sort this complicated topic, Sommer wrote an article for HRE readers called Are you ready for HR’s increasingly significant ESG role? recently. Sommer also will deliver a keynote address, HR & ESG: A Combination You Must Embrace, at the free, online HR Technology Conference Virtual, Feb. 27-29.