Europe wants sovereign AI infrastructure, more cloud capacity and substantially greater computing power. Yet the global data-centre race may ultimately be decided by something far less digital: electricity. Germany is rapidly expanding renewable generation but remains a comparatively expensive location for power-intensive infrastructure. France is exploiting its nuclear fleet, the Nordic countries combine low-carbon electricity with lower prices, while the United States and several Asian economies are linking data-centre expansion much more directly to new generation capacity. The question for Europe is becoming uncomfortable: can ambitious sustainability rules and fragmented national energy systems deliver the speed, resilience and cost base required to compete globally?

The defining constraint on the next generation of data centres may not be access to processors, land or capital. Increasingly, it is access to electricity that can be delivered in sufficient quantities, at a predictable price and with the reliability expected from infrastructure supporting financial systems, cloud services, public administration and artificial intelligence.
That shift is already visible in investment decisions. A BCS Consultancy survey of more than 3,000 industry participants found that 91 per cent now regard power availability as one of the two most important site-selection criteria, with 70 per cent placing it first. The European Data Centre Association reaches a similar conclusion in its 2026 market assessment: power availability has become the single most important constraint on growth, particularly in established metropolitan hubs.
This creates a strategic problem for Europe. The continent wants to increase its computing sovereignty at precisely the moment when data centres are moving from large electricity consumers to infrastructure capable of reshaping regional power systems.
The real competition is therefore no longer simply between Frankfurt, Paris, Dublin or Helsinki. It is increasingly between different energy models.

Germany’s renewable paradox

Germany illustrates the contradiction particularly clearly. The country has made substantial progress in decarbonising its electricity system. Renewable sources supplied 55.1 per cent of gross electricity consumption in 2025, rising to around 57 per cent in the first half of 2026. Solar capacity continues to grow rapidly, while wind remains a central source of generation.
For operators seeking low-carbon power, that development matters. German data-centre companies also make extensive use of renewable power-purchase agreements. The German Datacenter Association has reported that a large majority of electricity consumed by surveyed colocation operators is already sourced from renewables.
But renewable sourcing and international competitiveness are not the same thing.
The European Commission’s own 2026 impact assessment for the proposed Cloud and AI Development Act exposes the scale of the cost problem. Its comparison of industrial electricity prices places Germany at around €0.275 per kWh, against an EU average of €0.190, €0.121 in Sweden, €0.101 in Finland, €0.090 in the United States and €0.081 in China, with the Chinese figure based on 2024 data. These are broad industrial benchmarks rather than negotiated hyperscale data-centre tariffs, but the disparity is too large to dismiss.
Eurostat points in the same direction. For medium-sized non-household consumers in the second half of 2025, Germany recorded one of the highest electricity prices in the EU, while Finland and Sweden were among the cheapest. Again, the consumption band is not representative of a hyperscale campus, but it illustrates the structural differences within the European electricity market.
This is Germany’s central dilemma. A high share of renewable generation improves the carbon profile of digital infrastructure, but it does not automatically provide cheap, dispatchable electricity at the precise grid node where a 200 MW or 500 MW AI campus wants to connect.
Wind and solar also create a different planning requirement from continuously available generation. Data centres operate around the clock. PPAs can balance renewable consumption contractually over time, but they do not by themselves guarantee that renewable output will physically match a facility’s load every hour of every day. Networks, storage, flexible generation, interconnection and backup capacity still have to close that gap.
The German Datacenter Association has therefore become increasingly explicit about electricity prices and grid access. It argues that excluding data centres from current industrial electricity-relief mechanisms creates a competitive disadvantage and warns that investment could migrate to other countries if digital infrastructure cannot access internationally competitive energy conditions.
Germany may consequently discover that being able to advertise a greener electricity mix is not sufficient. For an AI investor, the more difficult questions are: When will 300 MW actually be available? At what delivered cost? And how firm is that capacity?

France offers a very different proposition

Across the border, France is constructing an almost opposite argument. Rather than relying predominantly on the rapid expansion of variable renewable generation, France is marketing its existing nuclear fleet as an advantage for AI and data-centre investment: large volumes of relatively stable, low-carbon electricity supported by an extensive high-voltage network.
The French government has made the connection explicit. It describes abundant, decarbonised and stable electricity as one of the principal reasons to locate AI infrastructure in France and has identified dozens of sites suitable for new data centres. By January 2026, the government said 63 sites had been identified, 26 had already been secured by project developers and five fast-track locations offered more than 700 MW of connection potential within approximately three to four years.
There are signs that operators are prepared to build their energy procurement around that nuclear base. Data4 has signed a 12-year agreement with EDF allocating 40 MW from France’s existing nuclear fleet, corresponding to an expected annual volume of roughly 230 GWh from 2026.
France also recorded very strong electricity exports in the first half of 2026 as nuclear availability improved, reinforcing the government’s argument that its generation mix constitutes a strategic industrial advantage.
The comparison with Germany is therefore awkward. Both countries can offer low-carbon electricity. But from the perspective of a data centre requiring constant high loads, France can currently make a stronger case that a substantial share of its low-carbon supply is also firm. Nuclear power is not a universal answer. New reactors require long planning and construction cycles, existing fleets require costly maintenance, and concentration on one generation technology carries its own resilience risks. But for projects being developed now, France benefits from infrastructure that already exists.
That advantage is helping to influence investment geography. France has attracted a series of major AI and data-centre commitments and is actively coupling power availability with accelerated connection and permitting processes.
The lesson for Europe is not necessarily that every state should adopt nuclear power. It is that investors increasingly reward countries able to connect energy policy, grid planning and digital industrial policy into one proposition.

The Nordic alternative: cheap, clean – but not unlimited

Finland and Sweden provide a third model. Their attraction is based on abundant low-carbon electricity, comparatively favourable prices, cooler climates and an increasing level of international connectivity. The price differential with Germany is substantial enough to affect location decisions for workloads that do not need to sit immediately beside continental Europe’s largest population centres.
The scale of investor interest in Finland is extraordinary. Fingrid says new electricity-consumption connection enquiries now exceed 100 GW, with more than half associated with potential data-centre projects. The Finnish grid operator correctly cautions that these are preliminary, non-binding enquiries and only a fraction will ultimately be built. Nevertheless, they demonstrate how strongly power availability is redirecting digital infrastructure towards northern Europe.
The Nordics are not a complete substitute for Frankfurt, Paris or Amsterdam. Latency, fibre routes, proximity to customers, cloud availability zones and financial-market requirements still favour established hubs for certain services. Extremely large clusters also risk reproducing the same grid bottlenecks that pushed investment northwards in the first place.
But for AI training and other less latency-sensitive workloads, geography is becoming more flexible. The European Data Centre Association notes that hyperscale development is increasingly moving beyond traditional hubs towards the Nordics, Southern Europe and other regions where energy and land are easier to secure.
Europe’s data-centre map is starting to follow the electricity map.

Ireland shows what happens when digital success outruns the grid

Ireland provides the warning. The country successfully attracted large concentrations of international cloud infrastructure, particularly around Dublin. But that success became an energy-system problem.
EirGrid has reported around 2,000 MVA of demand capacity contracted to data centres and other new technology loads at transmission level, with another roughly 300 MVA contracted at distribution level. New data-centre connection applications are now assessed against requirements that include the ability to provide dispatchable on-site generation or storage sufficient to support their demand.
Ireland therefore demonstrates the limit of treating data centres merely as economic-development projects. Once a cluster becomes large enough, its energy requirements can alter national capacity planning.
This matters for Germany and other countries now pursuing rapid AI expansion. Grid constraints do not disappear because a project is strategically valuable. They become more politically difficult.

Can Brussels solve a problem rooted in national power systems?

The European Commission increasingly understands the scale of the challenge. Its proposed Cloud and AI Development Act, published in June 2026, aims to at least triple EU data-centre capacity within five to seven years. The proposal identifies limited access to energy, land and financing, together with slow permitting, as major barriers to European cloud and AI expansion.
At the same time, the EU has launched a tender for up to seven AI Gigafactories, supported by as much as €10 billion in European and national public funding and intended to mobilise at least €20 billion in additional private investment. These facilities are expected to combine more than 100,000 advanced AI processors with high-speed networking and energy-efficient data-centre infrastructure.
This is an important policy shift. Europe has recognised that compute capacity itself is strategic infrastructure.
But there is a limit to what Brussels can deliver.
The EU can harmonise permitting frameworks, establish sustainability rules, coordinate investment, support cross-border grids, provide financing and prevent 27 entirely incompatible regulatory regimes from emerging. The EUDCA has welcomed CADA precisely because stronger European coordination could improve the investment environment.
What Brussels cannot do quickly is create 500 MW at a constrained substation outside Frankfurt.
Generation mixes, grid investment, planning law, local authorities and many energy taxes remain substantially national. Even the June agreement between the Commission, EUDCA, grid organisations and energy associations is framed around developing common principles that Member States may use to coordinate data centres, energy infrastructure and local government.
The unavoidable conclusion is that Europe needs both levels.
The EU can remove fragmentation. Member States still have to deliver the megawatts.
The danger is that Europe develops an increasingly sophisticated common regulatory framework while the locations able to build infrastructure fastest remain those where national governments have already solved the energy question.
France may therefore matter as much as Brussels in determining whether Europe can compete.

The United States is prepared to build power around computing

The global comparison makes the challenge sharper. US data centres face serious grid constraints of their own. The Department of Energy expects their share of US electricity consumption to rise dramatically, and its latest modelling places data centres at between 9.5 and 15.3 per cent of total US electricity use by 2030.
Transmission queues and local shortages are already creating problems. America is not a frictionless energy market.
But the US response demonstrates substantially greater willingness to treat new generation and new data centres as a single industrial project.
At the Savannah River Site in South Carolina, for example, the US Department of Energy selected Amentum in July to negotiate a project combining a 1 GW AI data centre with around 2 GW of dedicated on-site energy generation, initially using natural gas with a planned transition towards nuclear power.
Separate Department of Energy projects are examining microgrids, battery storage and on-site generation as ways of integrating large data-centre loads while reducing stress on existing networks.
Europe would understandably question parts of this model, particularly the use of additional gas generation. Yet from the perspective of investment speed, the contrast is important. The US is increasingly asking: What generation must be built with the data centre?
Europe still often asks: How can the data centre fit into the electricity system that already exists?
The first approach can increase emissions if poorly designed. The second can preserve climate discipline but create years of delay.
Finding a European answer between those extremes is becoming urgent.
Cost also matters. Average US industrial electricity prices were 8.62 US cents per kWh in 2025. Regional differences are considerable, but the broad gap with many European markets remains a structural advantage.

Asia is not following one model either

Asia offers several further alternatives. China combines extensive state-directed infrastructure investment with substantially lower industrial electricity prices than the EU according to the Commission’s CADA impact assessment. The comparison is imperfect – market structures, energy mixes and state support differ fundamentally – but a European AI company deciding where compute should be located cannot ignore the resulting cost gap.
Singapore, by contrast, has too little land and energy to compete through unlimited expansion. Its response has been selective growth. The government is releasing at least 200 MW of additional data-centre capacity while imposing increasingly demanding energy-efficiency standards and linking further expansion to innovative green-energy pathways.
Malaysia is pursuing scale more aggressively. Government investment policy now targets substantial dedicated data-centre capacity, expanded renewable generation and direct access to green electricity through its Corporate Renewable Energy Supply Scheme. A dedicated Data Centre Task Force is also intended to screen projects for resource availability and execution credibility rather than approve speculative capacity requests.
Japan may offer the most interesting strategic comparison for Europe. Its so-called watt-bit collaboration explicitly seeks to plan electricity networks, telecommunications infrastructure and data-centre locations together. Rather than treating digital infrastructure as a new load that arrives after the grid has been designed, Japan wants computing capacity to follow the geography of carbon-free power – including renewables and nuclear generation – while grid and fibre investment are coordinated with it.
That approach comes remarkably close to the problem Europe is now trying to solve.

Energy security is becoming digital sovereignty

This competition is not simply about attracting foreign investment. For the security sector, the more consequential issue is dependence. Rechenzentren are already treated under NIS2 and the Critical Entities Resilience framework as part of Europe’s critical digital infrastructure. If Europe cannot provide economically viable power for those facilities, more computing workloads will inevitably be located elsewhere. That creates a paradox.
Europe can regulate cloud sovereignty, require high levels of cybersecurity and seek greater control over sensitive data. But digital sovereignty becomes difficult to sustain if the underlying compute infrastructure is economically more attractive in jurisdictions where energy is cheaper and deployment faster.
The energy question is therefore also a security question.
A facility powered by cheap electricity but dependent on a fragile grid is not resilient. A data centre running entirely on contractual renewable energy but unable to obtain additional grid capacity is not scalable. A highly efficient European facility that is built five years after its American or Asian competitor has already entered service may be sustainable but strategically irrelevant.
Europe needs to optimise all three variables simultaneously: carbon intensity, security of supply and cost.
Germany currently excels at only part of that equation. Its renewable transformation gives it a credible path towards increasingly low-carbon computing. Yet high electricity costs, grid congestion and complex national requirements weaken the investment proposition. France currently appears better positioned to combine low-carbon power with firm generation. The Nordics can compete on price and clean energy, while Southern European markets may benefit from renewable abundance if grids and water constraints are addressed. There may therefore be no single European energy model for data centres.

What Europe does need is a common strategic principle: computing capacity must be planned together with the energy system that will support it. The European Commission can establish that framework. It can accelerate permits, finance AI Gigafactories, coordinate grids and reduce regulatory fragmentation. But the decisive competition will still take place at national and regional level, where an operator asks a much simpler question: Can you give me the power, when I need it, at a price that allows me to compete?

France, Finland, the United States, Japan and parts of South-East Asia are increasingly building policy around that question.
If Germany – and Europe more broadly – cannot answer it convincingly, the continent risks an uncomfortable outcome: becoming one of the world’s strongest regulators of AI infrastructure while allowing an increasing share of the infrastructure itself to be built somewhere else.