Can Europe keep the lights on and win the AI race?

EUROPE

The EU and UK want to become AI superpowers, but ageing electricity grids, water shortages and public opposition threaten to leave their ambitions badly behind schedule.

Europe may not have enough electricity to win the global artificial intelligence race, a new Sibylline report warns.

The EU wants to triple its data-centre capacity within five to seven years, while Britain is pursuing a similarly ambitious expansion by 2030. But ageing electricity networks, lengthy connection queues and increasing competition for energy risk pushing delivery far behind those political promises.

“Grid capacity and structural challenges are much more immediately pressing” than hostile-state threats to Europe’s data centres, says Annabelle Walker, a Sibylline Europe analyst and co-author of the report.

Europe accounted for about 15% of global data-centre electricity consumption in 2024. Worldwide consumption is projected to more than double, from 415 terawatt hours to around 945 TWh, by 2030.

Although Europe continues to expand renewable and nuclear generation, electricity demand is also rising from electric vehicles, heat pumps and industrial electrification.

The immediate obstacle, however, is often not generating the power but connecting new facilities to congested networks.

Around 40% of EU distribution grids are more than 40 years old, while grid-connection queues exist in at least 16 member states.

In West London, where electricity has become the biggest constraint on data-centre growth, new facilities may wait seven to 13 years for a connection.

Permitting presents another obstacle: 87% of operators identify it as a major challenge, while 70% say projects typically take between three and seven years to become operational.

Some developers are consequently considering on-site generation to avoid waiting for the grid.

Yet this increases costs and risks undermining net-zero objectives.

“The whole thing becomes this maelstrom of lots of different problems that are all being joined together by this issue of grid connection,” Walker says.

The contradiction becomes sharper during heatwaves.

Data centres require enormous amounts of electricity to operate and cool their servers, just as demand from air conditioning surges and power generation can become less reliable.

In June, the UK’s largest supercomputer, Dawn, was shut down for several days after its cooling system failed during a heatwave.

Cooling systems also place growing pressure on water resources. Global AI-related water consumption could rise from 1.1 billion to 6.6 billion cubic metres by 2027, according to the UK Government Digital Sustainability Alliance.

That is equivalent to more than half of Britain’s annual water consumption.

The combination of heat and drought will make parts of southern Europe particularly vulnerable, even as cheaper land and abundant renewable energy make the region attractive to developers.

As competition for electricity intensifies, governments may also be forced to decide who receives priority.

Walker says it is a realistic possibility that data-centre requirements will increasingly be weighed against those of local communities, businesses and agriculture.

Ireland offers one possible solution. After years of restrictions around Dublin, regulators have allowed new large data centres to connect only if they provide generation or storage matching their maximum import capacity.

At least 80% of their annual electricity demand must eventually be matched by additional renewable generation in Ireland.

Similar rules could relieve pressure on national grids, but would impose further costs on developers and potentially make European markets less attractive.

They could also favour the biggest US hyperscalers, which possess far greater capital and resources than most European competitors.

This creates an awkward paradox. Europe is investing in sovereign AI to reduce its reliance on foreign technology, yet its data centres remain dependent upon US-designed chips and cloud services, Asian semiconductor manufacturing and Chinese-controlled critical minerals.

Delays to European capacity could therefore prolong reliance on the same American providers Brussels hopes to compete with.

The strategic importance of AI infrastructure is already growing.

Under a new agreement with Kyiv, Britain will give private companies access to four years of Ukrainian battlefield data to train AI systems capable of detecting threats to defence sites, railways and energy infrastructure.

The technology, initially being tested at a UK defence site, will use sensors in buried fibre-optic cables to distinguish between protesters, vehicles and hostile-state activity.

If successful, it could later be deployed at airports, prisons and other sensitive locations.

But the plan also illustrates why data centres are becoming focal points for environmental, anti-capitalist, anti-technology and anti-militarist campaigners.

In the Netherlands, Extinction Rebellion activists attacked a Microsoft data-centre construction site with balloons containing a corrosive mixture in July.

Walker says the incident represents an escalation beyond the civil disobedience normally associated with the group.

More serious attacks have already taken place in Germany.

The left-wing extremist Volcano Group, which claimed responsibility for sabotaging Berlin’s Lichterfelde power plant in January, partly justified the attack by citing the “insatiable” energy demands of AI technology.

Data centres have become “not only symbolic but also quite a strategic facility to target”, Walker says.

Although opposition in Britain remains largely confined to planning challenges and NIMBY activism, increasingly destructive tactics elsewhere could spread.

The greater danger is that concern over jobs, energy prices and scarce resources erupts following a major infrastructure failure.

A widespread power or water outage affecting hospitals, schools or vulnerable residents could be blamed on a nearby hyperscale facility, whether or not it was responsible.

Walker assesses that the speed at which AI is transforming society could deepen a sense of powerlessness among those who feel decisions are being made above them by powerful technology companies rather than accountable governments.

Europe will continue building data centres, but Sibylline assesses that expansion will be slower and more uneven than headline targets suggest.

Governments must strengthen grids, secure water supplies, streamline planning and persuade communities that they will share AI’s benefits rather than simply absorb its costs.

Failure to do so could transform today’s planning disputes into something far more dangerous.

If a serious power or water failure harms hospitals, schools or vulnerable residents, a nearby hyperscale facility could be blamed whether or not it was responsible.

Walker warns: “That has the potential to create a perfect storm that could trigger major domestic unrest.”