The electricity consumption of artificial intelligence systems is rapidly increasing and, according to a new study by economist and data scientist Alex de Vries-Gao, could surpass the energy expenditure of Bitcoin mining by the end of 2025. Various tech and economic media outlets, reported on this on May 30, 2025. The study meticulously analyzes current trends in the exponential growth of computing power required for training and operating large AI models, such as those used in advanced chatbots, image generation systems, and scientific computing. De Vries-Gao, known for his work estimating the energy consumption of cryptocurrencies, points out that the energy intensity of modern AI servers and data centers, especially those equipped with thousands of specialized graphics processing units (GPUs) from Nvidia and other manufacturers, poses a serious and rapidly growing challenge. It is projected that if current AI adoption and scaling rates continue, global energy consumption by the AI sector will not only overtake Bitcoin but could also account for a substantial share of total global electricity consumption, comparable to that of medium-sized countries. This raises acute questions about the strain on existing power grids, the need for accelerated development of more energy-efficient AI algorithms and hardware (e.g., neuromorphic chips), and the overall environmental footprint of the burgeoning AI industry and its contribution to climate change. The study's findings call for more thorough global monitoring, implementation of energy efficiency standards, and potentially, regulation of energy consumption in the AI sector.
Study: AI Energy Consumption to Surpass Bitcoin Mining by End of 2025
