Summary
- Christine Lagarde says Europe already lacks sufficient data-centre capacity for its AI ambitions.
- The European computing-capacity shortfall could grow sixfold over the next decade if current trends continue.
- Closing the gap depends on physical infrastructure including power, land, cooling, connectivity, and construction capacity.
European Central Bank president Christine Lagarde has called for a substantial expansion of European data-centre and artificial-intelligence computing capacity, warning that the continent’s dependence on overseas technology leaves it exposed to economic and geopolitical risk.
Speaking in Vienna on 14 September, Lagarde said Europe already lacks sufficient data-centre capacity for its requirements and warned that the shortfall could grow sixfold over the next decade if present trends continue.
The warning places physical computing infrastructure alongside semiconductors, software, capital, and AI models in the debate over European technological sovereignty. European businesses may be increasing their use of AI, but much of the underlying compute remains concentrated outside the region, particularly in the United States.
Lagarde argued that Europe needs to become more capable of producing and hosting the technologies on which its economy is increasingly dependent. She also said faster adoption of AI could lift productivity by as much as 4% over a decade.
Building more domestic compute capacity, however, is not principally a software problem. Large AI facilities require electricity connections measured in tens or hundreds of megawatts, suitable land, substations, high-capacity networks, resilient backup systems, and mechanical infrastructure able to remove increasingly concentrated heat loads.
Those constraints are already shaping development across Europe. Power availability has become a limiting factor in several established data-centre markets, while planning scrutiny is expanding beyond building design to include water use, grid reinforcement, backup generation, environmental effects, and the cumulative impact of multiple large projects.
The electrical requirement is particularly difficult to compress into the short development cycles associated with computing hardware. Servers can be replaced in a few years, but substations, transmission lines, and major utility connections are planned over much longer periods.
That mismatch becomes more pronounced as AI hardware pushes rack densities higher. A facility designed around conventional enterprise or cloud loads cannot automatically accommodate dense accelerator clusters without changes to internal electrical distribution and thermal systems.
Liquid cooling is increasingly being designed into new AI-capable sites, while operators of existing facilities are assessing whether older halls can be retrofitted economically. Direct-to-chip systems, coolant distribution units, additional pipework, and revised heat-rejection arrangements can all change the plant required outside the IT room.
Europe’s capacity problem therefore extends beyond the number of data-centre buildings available. The location, power envelope, cooling architecture, network connectivity, and ability to accommodate future hardware densities determine whether nominal capacity is useful for the workloads driving new demand.
Lagarde’s intervention also places the data-centre build-out inside Europe’s broader capital-allocation debate. The continent is attempting to increase investment in AI at the same time as energy networks, defence, transport, housing, and industrial decarbonisation are competing for capital, equipment, land, and skilled labour.
New computing capacity will therefore have to demonstrate more than demand from prospective tenants. Projects need credible power positions, financing, planning pathways, and construction programmes if they are to become operational infrastructure rather than remain capacity in development pipelines.
Greater European hosting capacity would not eliminate dependence on global technology supply chains. Advanced processors remain highly concentrated geographically, and the largest cloud platforms are dominated by US companies. But the location of the physical infrastructure still affects resilience, jurisdiction, data location, latency, and the degree of control available to European governments and businesses.
The ECB has previously identified a widening gap between European and US digital investment, including data-centre construction. Lagarde’s latest warning turns that gap into a strategic infrastructure question.
If Europe wants a materially larger share of AI activity to run on infrastructure located within the region, the required investment will extend well beyond servers. The constraint will be whether enough power, land, cooling capacity, connectivity, construction resource, and capital can be assembled quickly enough to turn demand into operating capacity.
