The digital infrastructure that powers Artificial Intelligence (AI) has become one of the biggest threats to the European and global energy transition, warn experts in a report by the Environmental Coalition on Standards (ECOS) and Open Future, cited in an analysis published by the 2Celsius Association – an environmental organization.

“If you’re tempted to think that artificial intelligence is a ‘clean’ technology, that the answers you get from a chatbot leave no physical footprint in the real world, the report ‘From Innovation to Overshoot: How Data Centre Expansion Risks Derailing Climate Goals’ will make you reconsider that perception. Published by the environmental coalition ECOS and the European think tank Open Future, the document precisely traces the anatomy of a paradox of the digital age: the more artificial intelligence we use, the more the planet pays – in energy, water, raw materials and CO2 emissions. A data centre is, in essence, a massive building filled with servers, storage equipment and computer networks – the physical infrastructure that makes any digital application possible, from email to large-scale language models like ChatGPT or Gemini. These facilities require uninterrupted power supplies, backup generators and sophisticated cooling systems (…) The digital infrastructure that powers Artificial Intelligence (AI) has become one of the biggest threats to the European and global energy transition,” the analysis states.

According to the cited source, with the explosion of AI, complexes dedicated exclusively to this new technology have also emerged, equipped with thousands or tens of thousands of graphics processing units (GPUs) and tensor processing units (TPUs), connected via ultra-fast networks. Thus, Microsoft has built, for example, a dedicated OpenAI supercomputer, equipped with over 285,000 CPU cores and 10,000 GPUs interconnected via a 400 Gb/s network, while Meta has laid the foundations for the ‘AI Research SuperCluster’ with 16,000 Nvidia A100 GPUs.

“The impact of data centers on global energy systems is already overwhelming and accelerating rapidly. The International Energy Agency (IEA) has explicitly identified AI-powered data centers as a major structural factor in the 4.3% increase in global electricity demand in 2024. Furthermore, data centers contributed almost 60% of the increase in electricity consumption in the buildings sector in the same year. The installed capacity of data centers increased in 2024 by about 20% (about 15 GW), mostly in the United States and China. These facilities are still predominantly powered by fossil fuels, which generate almost 60% of global electricity – coal alone covering 35%”, claims 2Celsius President Raul Cazan, according to Agerpres.

Official data shows that, at the European level, data center consumption was estimated at 45-65 TWh in 2022, equivalent to approximately 1.8% – 2.6% of the total electricity consumption of the European Union. In this context, Ireland is an extreme case, where data center energy demand has tripled since 2015 and now represents approximately 18% of the country’s total electricity consumption. Increases in energy demand were also recorded in: the Netherlands (around 5%), Luxembourg (around 4.8%) and Denmark (4.5%).

A McKinsey report from 2024 also estimates that, in Europe, data center energy consumption will triple by 2030, from approximately 62 TWh to over 150 TWh.

“This would raise the share of data centers in total European electricity consumption from 2% to around 5% – an increase that calls into question the entire decarbonization effort of the continent. AI is not the only culprit, the report’s authors state. Bitcoin mining consumes between 155 and 172 TWh annually – comparable to the energy consumption of Poland. Streaming platforms such as Netflix and YouTube together consume over 200 TWh annually. Cloud gaming can require around 340 watts per user in data centers alone. Social networks, with over 5 billion users, generate continuous data traffic that requires industrial-scale processing and storage infrastructures,” the specialized analysis emphasizes.

In terms of the climate goals of big tech companies, the report documents that Google reported a 48% increase in emissions compared to 2019, Microsoft saw a jump of almost 30% compared to 2020, Meta doubled its emissions compared to 2019, and Amazon, despite reporting a 3% decrease in its carbon footprint, was removed from the Science Based Targets initiative for lacking a credible emissions reduction plan.

Overall, research shows that tech companies’ actual emissions are estimated to be 662% higher than what is publicly reported.

“Beyond energy and emissions, data centers have a significant and often overlooked impact on water resources. Water is used extensively in cooling processes – both in the data centers themselves and throughout the entire lifecycle of the equipment, from raw material extraction to semiconductor manufacturing. A single Google data center consumes an average of about 1.7 million liters of water per day. A data center consumes about 26 million liters of water annually per megawatt of power. If 20% of European data centers used liquid cooling systems, total water consumption would reach 43.2 billion liters annually – enough to fill about 173 million Olympic-sized swimming pools. Advanced AI models also amplify this impact: according to the report, models like ChatGPT consume about two liters of water for every 10-50 queries. At the scale of current global usage, the numbers become astronomical,” notes 2Celsius.

According to environmental experts, the European Union currently lacks a coherent strategy covering the entire value chain of AI-powered data centers. In this context, the Energy Efficiency Directive (EED) on mandatory sustainability reporting for data centers with an IT power of over 500 kW, the Ecodesign Regulation for servers and data storage products, the CSRD and CSDDD Directives, components of the corporate sustainability reporting framework, which should require companies to report GHG emissions and identify adverse human rights impacts along their value chains, as well as the Renewable Energy Directive (RED III) – which aims for at least 42.5% of total energy consumption to come from renewable sources by 2030, but which does not contain mandatory targets specifically for data centers.