Readers of Blognone are well aware that running Generative AI consumes electricity at a higher rate than ordinary processing tasks. However, the issue of increased electricity usage may just be the tip of the iceberg. This information comes from Ami Badani, the head of marketing at Arm, who revealed at the Fortune Brainstorm AI seminar in London that the energy consumption of AI is a significant concern. With the global trend towards more devices capable of running Generative AI, such as laptops and smartphones, the demand for electricity is skyrocketing, posing a risk to insufficient power supply.
Badani noted that examples like ChatGPT consume up to 15 times more electricity than regular web searches, highlighting the seemingly unlimited electricity demands of AI. Even OpenAI’s recent introduction of the Sora model for creating videos involves a staggering amount of processing chips and, consequently, electricity consumption.
Currently, data centers account for only 2% of global electricity consumption. Badani estimates that if AI becomes the primary driving force behind operations, data centers in the United States alone could consume up to 25% of the world’s electricity by 2030. Therefore, the current challenge for AI lies in designing chips that work efficiently with minimal electricity usage.
Source: Fortune
Image Source: Pixabay
TLDR: Generative AI’s energy consumption poses a significant challenge due to the increasing demand for electricity, especially in data centers, prompting the need for efficient chip design.
Leave a Comment