EU warns on AI: digital is not always green

Brussels – In Ireland, data centres already account for over 20 per cent of the country’s total electricity consumption. This is one of the most visible effects of a transformation that is reshaping European energy consumption and creating tension between the European Union’s environmental ambitions and the global artificial intelligence race. More generally, Europe accounts for 15 per cent of global electricity consumption by data centres, the United States for 45 per cent, and China for 25 per cent. These figures were released today (5 May) by the European Environment Agency (EEA), underlining that, at a time of growing geopolitical competition and economic uncertainty, digital technologies are considered key for the competitiveness and autonomy of the 27 member states.

The EEA stressed that tackling the twin transition – the green and digital transitions – is not only an environmental challenge, but also a strategic one. As stated in the document Navigating Europe’s twin transition — opportunities and challenges of digitalisation in the green transition, digitalisation is not inherently environmentally friendly. The rapid expansion of artificial intelligence, in fact, represents a growing challenge to achieving climate neutrality in Europe. The training of AI models is driving exponential growth in computational resources. According to the report Artificial intelligence and sustainable consumption in Europe, the computing power required for training increased a million-fold between 1960 and 2010 and by approximately a trillion-fold between 2010 and 2025.

 In 2024, data centres accounted for around 1.5 per cent of global electricity consumption and around 3 per cent of European consumption. Their energy demand, however, is growing rapidly and is often concentrated locally. “Driven largely by artificial intelligence, the sharp rise in data centres in Europe is expected to lead to a near doubling of the sector’s energy demand by 2030,” according to documents published by the EEA. Germany has the most facilities. Many data centres are concentrated around major urban hubs, particularly in Frankfurt, London, Amsterdam, Paris, and Dublin. Meanwhile, new markets are emerging in countries such as Spain, Italy, and Poland.

One impact that is sometimes overlooked is on water resources. Artificial intelligence generates demand for water at various stages of its life cycle, including cooling on-site data centres, power generation, and semiconductor production. Current projections indicate that data centres could increase their annual water consumption for cooling and electricity generation to around 1,068 billion litres by 2028, an eleven-fold increase from 2024 estimates.

Artificial intelligence systems, moreover, rely on strategically crucial minerals such as gallium, germanium, indium, palladium, and tantalum. Europe’s high dependence on imports and the global concentration of mining and refining capacity for many of these materials expose AI infrastructure to geopolitical and supply chain risks. 

Another key factor in assessing the impact of artificial intelligence is its role in influencing online behaviour. The consumption footprint of EU citizens is expected to increase by 2030, with online shopping accounting for an ever‑growing share. Therefore, for the EEA, “understanding the potential of AI to foster sustainable consumption is becoming crucial, as it shapes consumer behaviour and the way households relate to sustainability.”

Despite these challenges, the Agency believes that digitalisation could promote sustainability by improving the collection and analysis of environmental data, supporting more efficient industrial processes, and enabling smarter energy and transport systems. The main challenge, therefore, is “to ensure that digitalisation is guided by explicit sustainability objectives, without assuming that environmental benefits will emerge indirectly from efficiency improvements alone.”