Europe’s AI sovereignty moment

For Europe, AI sovereignty is becoming a practical question: where will the region host, power and control the computing capacity needed for the next phase of AI adoption? The proposed Cloud and AI Development Act (CADA), published in June 2026, is a sign that policymakers are starting to answer that question in infrastructure terms. Its most tangible ambition is to at least triple EU data-centre capacity over the next five to seven years, supported by faster permitting and better access to energy, land, water and financing.

France shows how this policy direction can translate into projects. Around the 2025 AI Action Summit, the country announced more than €109bn of AI-related infrastructure investment plans. SoftBank’s plan to develop up to 5GW of AI data-centre capacity in France, with investment of up to €75bn, gives the theme greater scale and visibility.

The geopolitical backdrop also matters. The recent Fable 5 and Mythos 5 export-control episode showed that access to frontier AI can be shaped by jurisdiction as much as by technology. Europe may not build full self-sufficiency overnight, but the direction is clear. If the region wants more control over AI capacity, it requires more cloud, compute and data-centre infrastructure on European soil.

The capacity gap behind Europe’s AI ambitions

Europe starts from a much lower base than the US and China. Europe is estimated to have only around 8GW of data-centre capacity at the end of 2025, far behind the US and China. Yet this gap may also create opportunities for companies involved in supporting future infrastructure expansion. If Europe wants to support AI sovereignty and more local compute capacity, it will likely require a meaningful increase in data-centre investment over the next decade. Secondary markets such as Northern France, Northern Spain and parts of the UK may help on this goal. The top 15 announced European projects alone could theoretically total 28GW¹, more than three times current European capacity.

Figure 1: Total European data centre capacity and growth lag both the US and China

Source: Morgan Stanley. For illustrative purposes only. Forecasts, estimates and project announcements are not guarantees of future outcomes.

This does not mean every project will be built. Data centres are hard assets with hard constraints. Announcements often outpace reality because projects still need access to power, planning permission, equipment and viable economics. In many markets, the question is not whether cloud providers want more capacity, but whether they can build it on an acceptable timetable. That is why the investable story may lie less with data centre operators and more with the companies that supply the infrastructure needed for the build-out.

Beyond GPUs: the infrastructure cost of data centres

AI data centres are often discussed through the lens of semiconductors. Graphics processing units (GPUs) dominate the headline cost and remain critical to the AI supply chain. But the infrastructure wallet is still large.

According to a recent estimation, the total cost of a US 1GW AI data centre is about $60bn, splitting between roughly $41bn of IT hardware and $19bn of infrastructure. That means close to one-third of the cost sits outside the semiconductor stack. It includes powered land, the building shell, electrical equipment, and grid connection.

Figure 2: AI data centre cost breakdown (1GW, gas-powered, USA)

Source: Stanford University, Cushman & Wakefield. Estimates are illustrative and subject to change.

In Europe, the infrastructure share could prove higher in some cases. Grid connections can take years. Suitable powered land is scarce in mature hubs. Labour is another constraint. As a result, ‘time to power’ is becoming one of the most important variables in the AI infrastructure race. A data centre without a reliable power connection is just a stranded building.

This is where data centre construction fits naturally into a European infrastructure investment theme. The opportunity is not to own the entire AI value chain. It is to capture the physical build-out required to host, power and connect AI capacity in Europe.

Where the infrastructure opportunity may sit

The potential beneficiaries can be grouped into three broad areas.

The first group is power and grid infrastructure. Data centres need large, reliable and increasingly low-carbon electricity supply, so the opportunity is not only in grid connections but also in new power capacity. Europe’s focus on energy security and decarbonisation is already supporting renewable generation, storage and grid reinforcement, and data-centre growth adds another source of demand. Renewable infrastructure suppliers may benefit where new capacity is built for power-intensive users, while cable makers and grid equipment providers are needed to connect these sites. Companies such as Prysmian, Nexans and NKT are relevant through cables and grid connections, while Siemens Energy adds exposure through grid technology and power equipment.

The second group is construction and engineering. A data centre is not a standard commercial building. It needs heavy foundations, secure access and complex mechanical and electrical works. The construction phase also must meet tight delivery schedules. According to company guidance, ACS looks particularly relevant given that around 25%² of its revenue is expected from data centre construction in 2026, while Skanska also has meaningful data centre construction experience. For broader construction groups, the opportunity may be more incremental, often through fit-out and energy integration rather than the full building shell.

The third group is electrical systems inside data centres. Once power reaches the site, it must be converted and distributed inside the facility with very high reliability. This is becoming more important as AI racks require much higher power density. Traditional low- and medium-voltage systems remain essential, while new architectures such as 800 VDC highlighted by Nvidia could change how power is delivered inside future AI data centres. Companies like Schneider Electric, ABB and Legrand are relevant because they supply critical electrical systems that help keep facilities running safely and efficiently.

Conclusion

Europe’s AI sovereignty agenda is not only about models, chips or regulation. It also depends on the physical infrastructure needed to build and operate AI capacity within the region. Data-centre development could create opportunities across selected infrastructure segments, including construction and engineering, grid and power equipment, and electrical systems. This adds a clear growth layer to Europe’s broader infrastructure build-out.

What WisdomTree Offers

The WisdomTree Europe Infrastructure UCITS ETF (WBLD) is designed to capture this opportunity through companies involved in building and supplying Europe’s next generation of infrastructure. Its value-chain approach targets exposure to construction firms and key component suppliers, avoids pure operators, and aligns the portfolio more closely with the areas where public funding, procurement rules and industrial policy may drive growth. The portfolio allocates meaningful exposure to Construction Materials, Electrical Equipment, and Construction & Engineering industries, which are directly relevant to data centre build-out in Europe.

Figure 3: WisdomTree Europe Infrastructure UCITS ETF’s coverage in infrastructure value chain

Source: WisdomTree.

¹ Morgan Stanley: European Data Centers: Growth lags the US, but long-term outlook still strong (15/06/2026).
² Source: Morgan Stanley, ‘European Data Centers: Growth lags the US, but long-term outlook still strong’.