AI’s Energy Consumption Could Surpass Bitcoin Mining by End of Year

Artificial intelligence’s energy use is already making up 20% of global data-center power demand, according to new research published in the journal Joule. By the end of this year, AI’s energy demand could double and account for nearly half of all total data-center electricity consumption worldwide.

The research, led by Alex de Vries-Gao, founder of Digiconomist, a research company evaluating technology’s environmental impact, aims to shed light on AI’s growing energy usage. De Vries-Gao notes that the widespread adoption of large language models like ChatGPT has accelerated this trend, with AI now surpassing bitcoin mining in terms of energy consumption.

Tech giants have acknowledged AI’s significant contribution to their increasing energy use. Google’s greenhouse gas emissions have risen 48% since 2019, complicating the company’s goal of reaching net zero by 2030. Other companies, including Microsoft and Google, are also exploring ways to reduce their energy consumption without compromising performance.

A recent report by the International Energy Agency found that data centers now account for 1.5% of global energy use, with electricity consumption growing four times faster than overall consumption in recent years. The IEA predicts that data center electricity consumption will reach over 900 terawatt-hours by the end of the decade.

However, the exact share of AI’s energy consumption in current data-center configurations remains unclear due to limited publicly available information and varying utilization rates of AI hardware. Researchers acknowledge that disclosing tech giants’ energy expenditure is crucial for accurate calculations.

To estimate AI’s energy demand, de Vries-Gao used analyst estimates, earnings call transcripts, and device details. He calculates that without increased production, AI will consume up to 82 terrawatt-hours of electricity this year – roughly equivalent to the annual electricity consumption of a country like Switzerland.

Source: https://www.wired.com/story/new-research-energy-electricity-artificial-intelligence-ai