”Difficult to Gain Clarity About the Environmental Impact of AI Bets,” new report from Capgemini shows

"Companies are often groping in the dark when it comes to the hidden climate costs of the AI revolution," according to the new Capgemini study. Generative AI is becoming increasingly popular, but its increasing use comes with a hidden cost: a growing climate footprint. The new Capgemini report reveals that many business leaders are aware of the problem, but few actually have control on how much their own AI bets contribute to emissions.
Estimates based on data previously shared by OpenAI about GPT-3 indicate that training a modern generative AI alone can consume as much energy as 5,000 households use in a year.
Capgemini’s new report “Developing Sustainable Gen AI” shows that almost half of decision-makers at organizations with active AI projects are aware that their use of the technology has contributed to increased greenhouse gas emissions. At the same time, only 12 percent have investigated how much their AI investments contribute to increased emissions.
“But it’s not about disinterest,” says Caroline Segerstéen Runervik, Nordic Head at Capgemini. “The problem is that it is extremely difficult to obtain reliable data on AI’s environmental impact. Public information is often outdated and the tech giants are tight-lipped about their secrets.”
She believes that the difficulty in obtaining actual information about the climate footprint of AI is an important part of the explanation for the fact that only 20 percent of respondents in the survey include environmental impact among the five most important factors when deciding on AI investments. The report also highlights the fact that 74 percent of those surveyed state that they have difficulty finding reliable information due to a lack of transparency from technology giants and AI specialists.
“Here, the industry, together with decision-makers, has a lot of work to do to increase the opportunities for organizations to make well-founded and long-term sustainable decisions about investments in AI,” says Caroline Segerstéen Runervik.