Europe’s Digital Autonomy Starts With Its Power Grid
The EU’s data centre ambitions could turn clean energy into a competitive advantage, provided grid investment keeps pace and costs are shared fairly. A commentary by Ludovic Subran
As part of Europe’s efforts to secure its digital autonomy and gain competitiveness in artificial intelligence, the European Union is proposing to triple data center capacity within the next five to seven years under the Cloud and AI Development Act, or CADA.
Today, Europe only hosts about 14% of global data-center capacity, less than a quarter of the combined capacity of China and the United States.
There, rapid expansion of the centers has triggered a backlash among citizens and local politicians, largely linked to a jump in electricity prices and pressure on water systems.
This doesn’t have to be Europe’s future.
With its plans to mobilize a combination of private and public financing to complete the project, Europe stands to benefit from the structure of its power market, where cost is borne more widely across a greater area. This helps to keep prices under control.
Europe is also well positioned thanks to its energy mix, which relies more on generation from renewables – especially in the Nordics and Spain.
Europe is already the world’s leading host of efficient data centre capacity. If effectively implemented, the European plan could enable the bloc to assert its digital independence on terms consistent with its commitment to sustainability and its goal of achieving climate neutrality by 2050.
The success of CADA will depend, however, on clear, EU-wide guidelines to ensure burden-sharing and sustainability. Already the plan includes incentives that tie permitting for new centers to energy and water efficiency, and grid availability. These are essential to Europe’s success in winning acceptance from local populations.
Under the CADA proposal, the EU would need to install another 24.8 GW from today’s estimated 12.4 GW, increasing data center power demand from today’s 72 TWh (2.6% of total demand) to 195 TWh (6.5%). This would cause total electricity consumption to rise by 7%.
To avoid critical power market bottlenecks and delays in the build-out, it is important that Europe scales its grid infrastructure and generation capacity together with the data centers themselves. If not, the mismatch between demand and supply can gridlock the expansion, resulting in upward pressure on prices
This dynamic is already visible in the US market where nodal pricing –pricing electricity based on local grid conditions – has led to constraints in areas with significant data center build-out, driving up prices.
Wholesale electricity prices in some grid nodes that have seen significant data center activity increased by up to 267% between 2020 and 2025.
Europe, by contrast, largely relies on zonal pricing, which spreads costs more widely across a larger market (often an entire country) reducing the price for individual households in the local vicinity.
Under this model, prices are spread across consumers in the zone. This system prevents local price shocks like those that have triggered opposition from communities and moratoriums in at least 11 US states and led to the blocking or delay of some $156 billion worth of data center projects in 2025.
While this zonal system keeps price impacts from data centers negligible when the zone is large enough (think of Germany, for example), it is less effective in smaller areas.
Ireland is a case in point.
Here, data centers draw nearly a quarter of the country’s national electricity demand. That share is expected to increase to nearly a third by 2034. To keep costs from spiraling out of control, Irish regulators responded by banning new data center developments around Dublin from 2021 to 2025.
In addition, from this year onward, newly built centers must meet at least 80% of their annual energy demand through new renewable sources within six years of starting production.
Spain is also considering legislation that would require data centers to run at least 80 per cent on renewable energy, along with additional efficiency requirements.
To avoid such bottlenecks occurring, European authorities should focus on spreading data centers more widely across the continent and further integrating Europe’s power markets through increased interconnection capacity.
This can help limit price spikes across regional markets, while enabling renewable generation to be developed where resources are most abundant and transporting it to regions with the greatest need.
Even so, our projections show that the power demand generated by the CADA plan would likely cause an increase in power prices in the low single-digit range over one to five years.
Over time, however, the investment cycle could create an opportunity to accelerate the transition towards a more resilient, renewable-based and lower-cost power system.
Using AI could further reinforce these benefits by improving system management, optimizing efficiency and reducing network losses.
How big of a difference can a cleaner power mix make?
Last year, the training of a sophisticated AI model in the US generated the equivalent of at least 72,816 tons of CO2 — more than the lifetime carbon output of roughly 1,000 average cars. Using that number, had the same model undergone the same training and computation in Sweden, it would have produced just 6,700 tons of emissions.
On average, data centers in Europe produce only around 55% of the emissions per kWh compared with those in the US.
So far, however, debates surrounding CADA have focused on protectionism and sovereignty.
Prioritizing the expansion of domestic players would provide the EU with an opportunity to gain strategic autonomy, while taking part in a global investment wave now rivaling the scale of the entire oil and gas industry.
A robust EU-wide framework on cost-sharing and energy use will determine whether Europe can catch this wave. Failure to do so risks reproducing the kind of backlash emerging in the US, and falling further behind it and China.
IEP Bocconi does not express opinions of its own. The opinions expressed in this publication are those of the authors. Any errors or omissions are the responsibility of the authors.
