Data centers expected to double their power consumption by 2030

The use of artificial intelligence could more than double the global power requirements of data centers by 2030. This is according to a study by US investment bank Goldman Sachs. Compared to consumption in 2023, the analysts expect an increase of 165% by 2030. They expect lower figures in Europe. "We estimate that electricity demand in Europe could increase by around 10 to 15 percent over the next 10 to 15 years," says Alberto Gandolfi, Head of the Pan-European Utilities team at Goldman Sachs Research.

According to the study, the global power consumption of data centers was over 56 gigawatts last quarter. Around 54 percent of this is attributable to cloud computing, 32 percent to services for companies, such as emails or storage, and 14 percent to artificial intelligence. Analysts expect electricity demand to increase to 84 gigawatts by 2027. At the same time, the operation of AI is expected to grow to 27% of consumption, while they cite declining shares of 50% for cloud applications and 23% for business services.

Electricity consumption estimates uncertain

The researchers expect data centers to consume 122 gigawatts of electricity worldwide by 2030. However, they point out an uncertainty in their calculations: if the demand for artificial intelligence grows less quickly than previously forecast, the demand for electricity could be between 9 and 13 gigawatts lower than estimated. This would be the case, for example, if the transition to AI-driven work or the monetization of AI applications does not develop as quickly as expected.

Following a decline in demand for electricity in Europe over the past 15 years, the analysts are now seeing an increase in demand. They attribute this to an exponential increase in the number of connection applications received by electricity distribution operators in recent years. This is an early indicator of the future demand for electricity. They attribute the development mainly to increased demand for data centers. However, this is still below the US level.

Analysts: Data center capacity utilization will increase by 2026

After data center utilization increased from 84% to 92% last year, analysts expect a further increase to 97% by mid-2026. After that, the demand for AI applications is expected to increase at a slower rate and be covered mainly by newly built data centers, meaning that they anticipate a decline in capacity utilization from 2027 onwards. Another factor is the efficiency of AI training and inference. "If we see falling capital expenditure due to increasing efficiency, this would reduce the risk of oversupply from 2027," explains analyst James Schneider.

Investments in the electrical infrastructure are necessary to meet the rising energy requirements resulting from the operation of additional data centers. The analysts put the amount in their study at 720 billion US dollars. "Both the approval and construction of the projects can take several years. This represents another potential bottleneck in the growth of data centers," warns Schneider and calls on those responsible to take action. The impact of DeepSeek on the European market cannot yet be estimated, says Gandolfi. In an operational environment, however, the performance of AI is more important than power consumption, which only one in four companies checks.

(sfe)