© Federal Ministry for Economic Affairs and Energy, data basis by the IEA and the German Association of Energy and Water Industries (BDEW)

How data centres increase our demand for energy

The energy demand for data centres could double by 2030. This is shown in a forecast by the International Energy Agency (IEA). It is mainly compute-intensive tasks, such as AI, that are driving this hunger for energy. But what does this mean for our energy supply?

Artificial intelligence now has a firm place in our everyday lives: for navigation to medical diagnosis to energy saving. The introduction of bitcoin in 2009 opened up a new era in technology. But all these smart systems use a lot of energy.

Experts say that training an AI model such as ChatGPT requires as much electricity as is used to power an entire city. The reason for this are the immense amounts of data that have to be processed using powerful graphics processing units (GPUs). For this purpose, new data centres with super computers are being built all across the world.

^{Europe will see a strong rise in energy demand for data centres.}

The amounts of energy required by data centres will rise – by how much depends on the uptake of AI. At the same time, computer chips are becoming more efficient. This year, Europe’s very first super computer, called JUPITER, is to be built in Jülich. It is expected to conduct a quintillion operations per second and is to be highly efficient thanks to new processors and a cooling system based on warm water. Moreover, the installation’s waste heat is to be used for heating.

According to the International Energy Agency (IEA), global demand for electricity for data centres may more than double by 2030, to reach 945 terawatt hours. The figure for Europe could almost increase threefold (according to McKinsey).

^{Germany currently Europe’s no.1 for data centres}

In Germany, data centres are currently mainly being built in Frankfurt and Berlin, and the availability of green electricity is increasingly important. Capacity has more than doubled since 2010, to reach 2,730 MW in 2024. According to an expertise by the Federal Ministry for Economic Affairs and Energy, the mark of 4,800 MW is expected to be crossed by 2030. This puts Germany in first place in Europe where data centre capacity is concerned.

According to the IEA, conventional serves will need 316.5 TWh in 2030, whereas AI servers will account for another 304.7 TWh of electricity consumed. Another 137.3 TWh will be required for cooling, 111.7 TWh for IT equipment, and 75.6 TWh for other infrastructure.

^{Use of renewables for crypto currencies and AI}

Crypto currencies and AI are to become much more energy efficient and ideally rely on renewables. Under the AI Act, suppliers of general-purpose AI models need to document their energy consumption. The European Commission will also task standardisation bodies with the development of a reporting and documentation procedure regarding energy consumption. This is to render energy consumption more transparent and raise awareness for this issue.