AI's Energy Hunger
The Energy Consumption of AI
This leaves us out of the loop if we are not talking about artificial intelligence (AI) today. Especially after the huge success of ChatGPT, tech companies have significantly increased their investments in AI.
Every day, a new functional AI joins our lives. The future is here and it is being transformed by them. They are affecting every sector, including industries, healthcare, finance, transportation and every other aspect of everyday life.
While AI offers numerous benefits, concerns have been raised about its environmental impact, particularly in terms of energy consumption. The sharp rise in consumption rates triggers multiple layers, including communication infrastructure, data center services and hardware production.
Each individual one requires a unique, comprehensive approach to calculating its energy usage. This includes not only the direct energy consumption during operation, but also the indirect costs associated with data collection, storage and processing. Therefore, it is very difficult to calculate the exact impact of AI.
By 2027, it is estimated that 1.5 million AI servers could be running annually, consuming more than 85.4 terawatt-hours of electricity per year [1] That's equivalent to the annual electricity consumption of Argentina, the Netherlands, and Sweden and about 0.5 percent of the world's current electricity consumption [2]
Training large AI models requires immense computational power, equating to the energy usage of hundreds of homes. Data centers that host these AI models are major contributors to global CO2 emissions, especially when powered by fossil fuels. Efforts to develop more energy-efficient AI hardware and utilize renewable energy sources are crucial steps toward mitigating its environmental impact.
While we await for advances in genAI, we can reap the benefits of AI in an ecological way by prioritizing greener approaches. Ensuring transparency in energy use and implementing sustainability standards will help to establish greenAI.
领英推荐
[1]Alex de Vries,The growing energy footprint of artificial intelligence,Joule,Volume 7, Issue 10,2023,Pages 2191-2194,ISSN 2542-4351,https://doi.org/10.1016/j.joule.2023.09.004.