More and more artificial intelligence is being used, which means the energy requirements of data centers are also increasing. According to experts, the electricity consumption of entire countries is at stake.
Researchers warn of a sharp increase in energy consumption in data centers due to artificial intelligence (AI). “AI tools consume a lot of electricity, and the trend is increasing,” said the managing director of the Hasso Plattner Institute (HPI) in Potsdam and head of the artificial intelligence and sustainability department, Ralf Herbrich, to the dpa news agency. Just training a single AI model is an energy-intensive process with complex prediction calculations.
Data scientist Alex de Vries from Amsterdam compares energy consumption with that of entire countries using a search engine that works with AI. The topic is therefore also important for climate protection. Scientists and Internet companies are therefore working to improve the “ecological footprint” of AI.
Consumption could rise to 30 percent
“Data centers today consume four to five percent of global energy consumption,” said Herbrich. If you include the use of digital technologies such as laptops and smartphones, the figure is eight percent. “There are estimates that consumption will increase to 30 percent in the next few years.”
AI expert Herbrich makes the comparison with an oven: To train an AI model, processors from hundreds of graphics cards, each consuming around 1,000 watts, ran for several weeks. “1,000 watts is as much as an oven.”
The topic of artificial intelligence is currently dominating public debates about technology. It had received a lot of attention, among other things, from the text robot ChatGPT from the Californian startup OpenAI. Because AI applications are becoming more and more common: safety technology in cars is also controlled by AI, which should make heating systems more efficient. AI systems exist in healthcare and many companies are already working with them.
How can energy be saved?
How is this energy-intensive technology compatible with climate protection? According to Herbrich from the Hasso Plattner Institute, the research aims to ensure that calculations can be carried out with fewer parameters and therefore less energy use, while at the same time the accuracy of the predictions only decreases minimally. Technology companies also pushed forward research into energy savings with AI. However, it will take a few years to develop solutions.
Researcher De Vries estimates that Google currently processes up to nine billion search queries per day. According to his calculations, if every Google search used AI, around 29.2 terawatt hours of electricity would be required per year. This corresponds to Ireland’s annual electricity consumption.
At the same time, De Vries speaks of an extreme scenario that will not happen in the short term. He referred to high costs due to additionally required AI servers and bottlenecks in the supply chain. For comparison: According to the Federal Network Agency, electricity consumption in Germany was around 484 terawatt hours in 2022.
Use of AI for climate protection
The Internet company Google, which launched the chatbot Bard this year, said that the energy required to operate the technology is increasing much more slowly than many forecasts had predicted. This has been shown by studies and our own experiences.
Google used tried and tested methods to greatly reduce the energy consumption for training a model. The company also points out that it uses AI for climate protection. Examples include “fuel-efficient route planning” on Google Maps and the prediction of river flooding.