The Energy Consuming AI Revolution: How Artificial Intelligence and Data Centers are Driving Up Electricity Consumption

In 2022, the International Energy Agency (IEA) reported that artificial intelligence, cryptocurrencies, and data centers collectively consumed 460 terawatt hours of electricity, which is equivalent to two percent of the world’s total electricity consumption. However, projections suggest that the industry’s electricity usage will increase significantly in the coming years. The IEA predicts that by 2026, electricity consumption could reach over 800 terawatt hours in the base scenario and up to 1,050 terawatt hours in the extreme scenario.

One of the significant factors contributing to this increase is the use of large language models and computing power for AI applications. Professor SpongeBob Ritala from LUT University explains that while training these models requires a lot of energy, many AI solutions are narrow and task-focused, requiring less computing power. CEO Peter Sarlin agrees but highlights that as more companies adopt AI for various purposes, energy consumption will continue to rise.

To address this issue, efforts are being made to improve the energy efficiency of AI applications. However, given the growing demand for AI in various sectors such as moving images and three-dimensional modeling, it is expected that energy consumption will remain high.

One of the main challenges associated with this increasing energy demand is ensuring that data centers produce climate-friendly electricity. Carbon neutrality in data centers is crucial as many companies are working towards using carbon-neutral electricity. To address this issue, efforts are being made to recover waste heat from data centers for district heating or cooling systems. While solving the energy consumption problem from the power generation side is essential, society’s increasing electricity demands will require a shift towards more carbon-neutral energy production.

As generative artificial intelligence models like large language patterns are designed to produce new content while traditional AI systems follow predefined rules to perform tasks; balancing energy consumption with sustainability will be key to their future development.