The increasing use of AI could triple data center energy consumption by 2030.

“Over the past two decades, no technology has driven the need for accelerated infrastructure development in Europe like AI,” analysts wrote in the McKinsey report. Today, data centers in Europe consume around 62 terawatts per hour (TWh), but this could rise to 150 TWh. That would be five percent of the entire European energy chain.

Big challenge

To cope with this increase, an additional 25 gigawatts of green electricity would be needed. This contradicts the goal of becoming climate neutral by 2040. One possible solution is to locate data centers near clean energy sources. This applies to AI training models that do not need to be near major cities. The more AI applications require computing power and low latency, the closer they need to get to urban areas.

As a long-term solution, McKinsey suggests carbon capture and small on-site nuclear reactors. In addition, diesel generators, fuel cells and batteries can also meet the energy needs of data centers. Surprisingly, McKinsey does not believe energy consumption will increase that much because the EU saves enough energy. Milder winters and deindustrialization due to high energy prices are also contributing to stagnating consumption.

The research institute Gartner predicted that the use of new and different processors would significantly reduce energy consumption. Unfortunately, this reduction is complex and very expensive, so the investment must be worthwhile.