- As part of its digital omnibus, the European Commission has proposed granting major artificial intelligence developers greater freedom to use certain categories of personal data for training models, while at the same time postponing the implementation of the landmark Artificial Intelligence Act.
- Europe, which remains far behind the United States, which accounts for around two-thirds of current and projected data-center computing capacity, is struggling with growing competition from Asian players in the race for artificial intelligence, in particular China, whose current operational capacity (4.5 GW) is now equivalent to the combined capacity of the main European markets, which are grappling with high construction costs and regulatory and energy constraints.
- Europe has thousands of data centers, but only a few of them have the high computing power required to meet AI needs. Allianz Trade notes that Europe has only two areas with data-center computing capacity above 1 GW, compared with four in Asia (two of which are in China) and seven in the United States.
- In parallel, the EU has just launched a new investigation into US cloud-computing giants amid growing concerns about the protection of European consumers’ data – even though it depends on US hyperscalers to close its technology gap.
- In addition to the EUR 200bn roadmap announced last year, Allianz Trade estimates that a further EUR 100bn is needed to build new capacity and achieve the official goal of tripling European data-center capacity over the next 5–7 years.
Europe is torn between accelerating technological development and protecting consumers. The EU finds itself in an increasingly ambivalent position as it presents its digital omnibus, a comprehensive package designed to streamline regulations and strengthen the competitiveness of domestic SMEs in the face of accelerating digitalization and the growing integration of artificial intelligence into routine operations. The Commission’s proposal grants major AI developers greater freedom to use certain categories of personal data for training models, in order to help European firms keep pace with global competitors, while simultaneously postponing by 16 months the implementation of several requirements introduced by the Artificial Intelligence Act. In parallel, the EU continues to closely monitor the dominance of US cloud-computing giants. In the wake of frequent digital outages, the latest investigation launched this week against large US providers highlights mounting concerns about their ability to protect European consumers’ data. As part of this investigation, in Allianz Trade’s view, officials may tighten mandatory requirements for foreign cloud-service providers, who have so far benefited from a relatively lenient application of the Digital Markets Act (DMA), by increasing interoperability with local software companies and improving data portability for consumers.
Although Brussels is trying to regulate their impact more tightly, Europe is dependent on US hyperscalers to close the technology gap. The European cloud-services market has grown eightfold since 2017, but the rising demand is absorbed largely by US hyperscalers: the three largest providers hold a 70% market share. At the same time, the market share of local providers has halved, as they struggle to match their US competitors in terms of technology deployment and investment capacity. This dynamic deepens Europe’s dependence on foreign firms and raises concerns not only about digital sovereignty but also about security, especially as control over European data is becoming a strategic asset for businesses. Despite initiatives such as GAIA-X or national sovereign-cloud programs, Europe’s own cloud-computing capacity remains insufficient to meet the rapidly growing demand for computing and AI infrastructure. Europe has thousands of data centers, but only a few of them have the high computing power required to meet the exponential needs of energy-intensive AI-based technologies. Europe has only two areas with data-center computing capacity above 1 GW, compared with four in Asia (two of which are in China) and seven in the United States. Meanwhile, according to Allianz Trade, Europe’s position in the global AI race is increasingly constrained by its structural weakness in operational computing capacity, where it lags far behind the United States (by around a factor of three), while also facing growing competition from the Asia-Pacific region, driven mainly by China’s rapid expansion. Computing capacities within Europe itself are unevenly distributed, with only five countries accounting for around half of the region’s total capacity, which limits the benefits of scale and integration. Although current plans suggest around 13 GW of additional capacity in the medium term, this remains almost 30% below the EU’s ambition to triple data-center capacity over the next five to seven years. Despite the EU’s EUR 200bn plan to build new capacity and close the AI-development gap with competitors, Member States are still courting foreign investment, as illustrated by the EUR 15bn worth of contracts signed this month with large US cloud-service providers.
Chart 1: The 20 largest regions by reported data-center computing capacity (in GW, as of H1 2025)
Sources: Cushman & Wakefield (update for the first half of 2025), Datacentermap.com, Allianz Research
Chart 2: Regional breakdown of data-center computing capacity (current and planned) / Revenues of the European cloud-services market and the market share of local providers
Sources: Cushman & Wakefield (update for the first half of 2025), Datacentermap.com, Allianz Trade Research
High construction costs and regulatory and energy constraints are slowing project delivery, and rising costs are holding back the expansion of data centers in Europe. The costs of building data centers in Europe are among the highest in the world and have been rising much faster than in competing regions – more than twice as fast as on major US markets over the last three years – due to higher land, energy and labor costs. Second, administrative hurdles and extensive permitting procedures lengthen project timelines: in many cases, the average time needed to obtain permits is around 18 months, but in some complex cases it can extend to as much as 48 months, which is roughly the average duration of the construction process itself. Third, Europe lacks harmonized electricity-pricing frameworks, exposing operators to high price volatility and complicating long-term planning. Fourth, the strong concentration of data-center clusters has overloaded grids, slowing new investments in hubs such as Amsterdam and Dublin and pushing up operating costs. Grid-congestion-related costs exceeded EUR 4bn in 2024. Similar problems are emerging in Spain and Finland, where data-center demand may reach or exceed 10% of grid capacity during peak-load periods. Finally, Europe’s ambition to create a carbon-free data-center ecosystem – already powered 94% by renewable energy sources – brings additional challenges related to grid stability. This is by far the main issue highlighted by data-center developers in Europe, as difficulties in connecting new data centers to the existing power grid or securing a stable power supply during peak-demand periods negatively affect overall performance and hinder the execution of investment plans. All these constraints have a major impact on the continent’s ability to expand digital infrastructure at the pace required by growing demand.
Chart 3: Regional comparison of data-center construction costs / The five biggest challenges for data-center operations in Europe over the next three years
*Single data point for China. Sources: Turner & Townsend, State of European Data Centers 2025 (EUDCA report), Allianz Trade Research
In this context, Europe must redefine its priorities if it is to meet its ambitious goal. Europe faces a tough strategic choice as it tries to reconcile ambitious objectives for AI and cloud-computing development with the realities of the current investment landscape. The EU’s EUR 200bn action plan is almost certainly already insufficient to achieve the goal of tripling data-center computing capacity by 2030. According to our estimates, EUR 280–300bn will be needed to close the infrastructure-capacity gap.
Among other things, further investment in modernizing the power grid and harmonizing electricity prices in Europe will be essential to improve both efficiency and functionality. Given this gap, external support will be unavoidable – whether in the form of foreign direct investment or service contracts with non-European providers that can help balance emerging supply-and-demand pressures. However, the limited pool of local options in Europe and the relatively less dynamic venture-capital market constrain its ability to scale while retaining full control over the new digital infrastructure. This is underscored by the recent announcement of EUR 15bn in investments by US hyperscalers in developing data centers in Germany and Portugal. Stricter regulations on large US tech firms would certainly help reduce systemic risk, but they could also have undesirable side effects, such as discouraging foreign investors who are crucial for strengthening Europe’s computing base. As a result, policymakers face a difficult challenge: rapidly boosting computing capacity while maintaining strategic control. Achieving both objectives at the same time will be difficult, and some concessions on sovereignty will be unavoidable in the absence of alternative domestic solutions, although this should not come at the expense of consumer protection.
Chart 4: Total construction cost required to triple data-center computing capacity in Europe over the next five years (optimistic scenario) to seven years (pessimistic scenario)
Source: Allianz Research
