The power industry is on high alert as data centers, the digital workhorses of the 21st century, continue to proliferate at an astonishing rate. These facilities, voracious consumers of electricity, are already causing ripples in Europe, where they accounted for 96 TWh, or 3.1% of total power demand in 2024. While this might seem like a modest figure, the distribution and future growth trajectories of this demand are anything but straightforward. Matteo Mazzoni, director of Energy Analytics at ICIS, underscores this complexity, highlighting that over 60% of this demand is concentrated in just five countries: Germany, France, the UK, Ireland, and the Netherlands. Within these countries, a mere five cities—Frankfurt, London, Amsterdam, Paris, and Dublin—account for half of the data center power consumption. These urban powerhouses are the epicenters of data center activity, and they are poised to remain so in the near future, with ICIS forecasting a 75% increase in European data center power demand by 2030, reaching 168 TWh.
However, the story doesn’t stop at these five countries or cities. As Mazzoni points out, the Nordics and southern Europe, particularly Spain, are emerging as intriguing new markets for data centers. This geographic shift could be influenced by several factors, including the availability of renewable energy, favorable regulatory environments, and the ever-evolving landscape of AI and machine learning. The types of data centers being built also play a crucial role in shaping future power demands. Computation-focused data centers, which handle AI, machine learning, and scientific simulations, consume significantly more energy than storage-focused centers. This disparity is further exacerbated by the operational differences between data centers used for AI training and those used for inference. Training, which involves teaching AI models using large datasets, can be conducted in remote locations where resources are abundant. Inference, however, is latency-sensitive and requires proximity to metropolitan hubs to ensure quick response times for user interfaces and applications. This geographical and operational complexity makes predicting future demand a challenging task.
The advent of generative AI is set to revolutionize the data center landscape. As AI models become more complex and powerful, the energy demands of data centers will inevitably increase. For instance, the training of GPT-4, with 100 trillion parameters, required 7,200 MWh, five times the electricity consumed for the previous version, GPT-3. On the inference side, the energy consumption is more modest but still significant, with daily queries resulting in an annual consumption equivalent to the needs of roughly 2,600 European families. This exponential growth in AI capabilities and the associated energy demands pose a formidable challenge for utilities and grid operators.
Yet, there is a glimmer of hope amidst this energy-intensive landscape. Advances in liquid cooling methods and improvements in power usage effectiveness (PUE) could mitigate some of the power demands. Liquid cooling, which offers more efficient heat dissipation, directly translates into lower electricity consumption. The average PUE has already decreased from 2.5 in 2007 to 1.56 currently, with new facilities achieving PUEs as low as 1.3 or even better. Google’s state-of-the-art data centers, for example, reported a global average PUE of 1.09 for the last year. This trend suggests that as technology evolves, data centers could become more energy-efficient, potentially offsetting some of the projected demand growth.
Despite these potential efficiencies, the future of data center power demand remains shrouded in uncertainty. As Mazzoni notes, past forecasts of load growth have often overshot actual demand, a pattern that could repeat with data centers. The rapid pace of technological advancements, coupled with the unpredictable nature of AI development, makes long-term projections inherently uncertain. However, one thing is clear: the data center revolution is here to stay, and its impact on the power industry will be profound. Utilities and grid operators must prepare for a future where data centers are not just another load but a dynamic and evolving force shaping the energy landscape. This dynamic requires a proactive approach, involving investments in renewable energy, grid modernization, and innovative energy management strategies. The power industry must embrace this challenge, not as a threat, but as an opportunity to drive innovation and sustainability in the digital age.
