Underpowered: how energy limits Europe’s AI future

A data centre opening in Sofia in 2021. Photo: Shutterstock

Attempting to catch up with the US-China dominated field of artificial intelligence, Europe must realize that energy, not regulation, may be its Achilles’ heel. As town-sized data-centres are being built in the vastness of the southern United States and all around China, the EU’s digital ambitions are running into a constraint as old as the industry itself: power.

Training a large language model today demands electricity on the scale of a city’s consumption. Meanwhile, the cost of keeping the lights on has become a competitive handicap in Europe. The EU’s dependence on imported energy has turned into a strategic bottleneck for its digital future.

Energy asymmetry

Europe’s energy geography is uneven and complex. Since Russia’s invasion of Ukraine, the continent has slowly weaned itself off Russian gas, but at a steep price. Spot electricity prices remain two to three times higher than in parts of the United States. Even with subsidies and green-transition funds, industrial consumers – including hyperscale data-centre operators – face volatile costs and an uncertain energy supply.

On the other side of the Atlantic, America’s advantage lies in abundant natural gas, somewhat deregulated energy markets, and a growing mix of low-cost renewables. Data centres in Texas and Virginia enjoy stable baseload power that costs half the European average. China, on the other side, powers its AI boom with cheap coal and an obedient grid, making energy security a matter of state planning, not market risk. Yet regardless of the model – whether driven by markets or by state control – the common denominator is sheer scale. The International Energy Agency expects global data-centre electricity use to double by 2030, reaching around 945 TWh, with AI-related facilities as the biggest users.

Bottlenecks at home

Ironically, Europe – a continent leading the green transition – might struggle to power the technology that could optimize it. In the EU, every potential AI cluster must navigate not just the environmental permits in place, but a patchwork of national grids and cross-border complexities.

In countries like the Netherlands and Ireland, data-centre moratoriums have already been imposed or considered due to the strain on electricity networks and land use concerns. Even Germany, being Europe’s economic engine, faces delays in connecting new industrial users, as utilities prioritize residential needs and struggle with transmission bottlenecks from wind-rich northern regions to energy-hungry southern hubs. The resulting picture is an inhospitable environment for proper AI development.

Politics is further compounding the challenge. The EU’s Green Deal Industrial Plan encourages investment in renewables and clean tech but remains ambivalent about energy-intensive industries like AI. With environmental lobbies wary of digital infrastructure’s carbon footprint, and voters sensitive to rising household energy costs, EU policymakers have little appetite to prioritize energy access for data firms, even when they promise future innovation dividends.

Yet behind these political hesitations, the main issue remains structural. To understand how policy meets physical limits, it is worth examining the continent’s reliance on external energy supplies and the balance between renewables and nuclear power that could sustain its AI growth.

Energy dependency

In 2023, the EU’s energy imports dependency rate was about 58 per cent of gross inland energy consumption, meaning nearly six in ten units of energy used in the EU are net imports. For gas in particular the picture remains stark: in 2024, just four suppliers accounted for the bulk of EU imports (Norway 33 per cent, the United States 16.5 per cent, Algeria 14 per cent, and Russia 11 per cent) despite Moscow’s invasion of Ukraine.

What does this mean for the AI-data-centre agenda? As the data-centre loads are destined to grow, any disruption to energy supply or wholesale electricity markets becomes a direct threat to digital sovereignty goals. External dependency means vulnerability to price shocks, pipeline or shipping interruptions, as well as foreign political leverage. This means that Europe risks being less attractive to the AI front-running companies.

Green but inadequate

On the upside, Europe’s clean-energy push has gathered pace. In 2024, renewables supplied about 47 per cent of the EU’s electricity generation, driven by record additions of wind and solar. For data-centres that trajectory looks promising. A greener grid means lower carbon electricity, less exposure to imported fuels, and tighter links between digital expansion and climate goals. However, it is clear that major constraints remain.

Hyperscale data-centres run continuously, while wind and solar do not. Without additional energy sources, renewables alone cannot guarantee the reliability needed by the AI sector. There are also various grid and connection limits. Europe’s wind and solar assets are often far from population and data-centre hubs. The issue in this case becomes transmission and connection, where underinvestment and long permitting times leave new loads waiting years for access. In sum, renewables are indispensable to Europe’s digital future, but insufficient on their own.

Reliable but slow

It is argued that nuclear power’s uninterrupted output makes it uniquely suited to power the 24-hour demands of AI-scale computing.

In the EU, nuclear already plays a role – supplying roughly 23 to 24 per cent of total electricity in 2024. For data-centre operators, the advantages are evident: a constant power supply free from weather fluctuations, improved grid stability, and reduced dependence on the volatile fossil fuel markets.

Yet nuclear is far from a quick fix. Political resistance and public scepticism persist in several member states, while the costs, regulatory complexity and long construction timelines for new reactors or smaller modular units make it an unlikely solution for the next few years of data-centre expansion. Nuclear power works only if policymakers can pair it with faster permitting, clearer investment signals, and modernized grids.

Power and sovereignty

The race for AI development does not wait. Unless Europe finds a way to square this circle, it may find itself watching the AI race from the sidelines. Without affordable and reliable energy, no amount of regulation, data governance or digital strategy can sustain a true AI ecosystem. Europe’s challenge therefore is not merely to build smarter machines but to ensure the current that keeps them alive.

The continent has long viewed energy as a geopolitical concern. Now it must see it as a foundation of digital independence. In the emerging world order of artificial intelligence, sovereignty will belong to those who can generate, store and distribute power as effectively as they generate data.

Zurab Khutsianidze is a former diplomat with expertise in geopolitics, security and emerging technologies.

New Eastern Europe is a reader supported publication. Please support us and help us reach our goal of $10,000! We are nearly there. Donate by clicking on the button below.

Artificial intelligence, energy, Europe