A study titled “Power Hungry Processing: Watts Driving the Cost of AI Deployment?”, posted online in November 2023 by Sasha Luccioni and Yacine Jernite of Hugging Face and Emma Strubell of Carnegie Mellon University, found that generating an image using a powerful AI model can take as much energy as fully charging a smartphone. The researchers tested 88 models across 30 datasets and 10 tasks, and image generation topped the table by a wide margin. Specifically, the authors wrote: “We found that Stable Diffusion XL uses nearly 1 phone charge worth of energy per generation.” The study measured the cost of using a model after training, not the training itself. The smartphone figure describes Stable Diffusion XL, the least efficient model tested, and the paper had not yet been peer reviewed when reported. Text generation is more efficient: making a thousand text answers used only about 16 per cent of a smartphone charge. The study also found that a large all-purpose model can use about 30 times more energy than a smaller, single-purpose model doing the same simple sorting task. Making a thousand images with Stable Diffusion XL produced about the same carbon as driving 4.1 miles in a petrol car. For one family of models, the authors estimated it would take between 200 and 500 million uses to match the energy spent on training.