The rapid integration of artificial intelligence into daily life marks a significant shift online, with hundreds of millions of people relying on chatbots and generative tools for everything from research to image creation. But as AI use expands, its growing appetite for energy is becoming clear, prompting new research from the MIT Technology Review to quantify its footprint, and to uncover just how little is publicly known. AI's rise has broken the pattern set by past waves of digital growth, where better efficiency kept data centers' electricity use flat even as their numbers ballooned. However, AI-specialized hardware caused US data centers' electricity consumption to double from 2017 to 2023, accounting for 4.4% of national energy use—a figure expected to triple by 2028 as AI becomes even more embedded.
Training large AI models like GPT-4 can cost more than $100 million and use enough energy to power San Francisco for days. But the majority of AI's energy use now comes from inference, or responding to user prompts; generating high-quality text, images, or video with open-source models especially consumes measurable amounts of electricity. Most large AI providers, like OpenAI and Google, release scant details on this energy use, labeling such data as trade secrets. This lack of transparency makes it hard for scientists and regulators to gauge the true impact. "The precious few numbers that we have may shed a tiny sliver of light on where we stand right now, but all bets are off in the coming years," says AI researcher Sasha Luccioni. "Generative AI tools are getting practically shoved down our throats, and it's getting harder and harder to opt out." More here. (This content was created with the help of AI. Read our AI policy.)
,
uBlock 
,
and
Ghostery 
)
