A Chat With Siyi Hao, Principal At Roland Berger On Europe’s AI Data Centre Developments

 Why do you think Europe is losing ground in AI data centre development?

Europe is not losing ground because demand is lacking. The reality is that structural constraints make AI infrastructure significantly harder and slower to build than in competing regions. 

Grid connections can take up to seven years in parts of Europe, electricity prices remain high and developers must navigate a highly fragmented regulatory landscape. At the same time, the US benefits from the concentration of hyperscale’s and frontier AI labs attracting the largest infrastructure investment. 

In the US, developers have been bypassing the grids with solutions such as on-site generation with gas turbines or gas reciprocating engines, whilst in Europe, local planning, environmental regulations and high gas prices make this difficult

As a result, Europe’s share of global installed data centre capacity is projected to decline from roughly 13% today to around 10% by 2030, while the US continues to strengthen its position. 

What policy and infrastructure changes do you think Europe needs to stay competitive?

The priority is structural reform and technology innovations rather than simply offering more subsidies. 

Roland Berger’s research shows that grid connectivity, energy availability and regulatory complexity are the biggest barriers to deployment. Faster grid access streamlined approval processes and more predictable planning frameworks would have the greatest impact. For example, the UK Ofgem’s review on speculative connections by applying an application fee could help speeding up connections for the more advanced projects.

More innovative energy solutions should also be looked at. A new network pattern has started to appear in Europe, with Nordics and Iberia starting to form as two training hubs, due to the abundance of renewable electricity. However, there is still expected to be a lot of demand for the traditional FLAPD (Frankfurt, London, Amsterdam, Paris and Dublin) market, where new grid connections are significantly harder to obtain, and energy technology innovations (such as on-site generation, battery storage, and flexible demand) can be leveraged to generate space, with an appropriate longer-term environmental strategy.

Financial initiatives can only help if the underlying infrastructure bottlenecks are addressed. Europe must demonstrate that it intends not only to regulate AI, but also to build the infrastructure that powers it.
 

How can Europe’s supplier base capture more value from the global AI infrastructure market?

Europe should build on the strengths it already has. For example, Europe has a sophisticated data centre supplier base, including for power distribution, cooling and power generation such as gas turbines and gas reciprocating engines, and there is also a strong underlying supply chain to these businesses. There is also a strong adjacent industrial and automotive supplier base that can participate in the data centre value chain with their capabilities, such as 800 VDC.

AI infrastructure is global in technology but local in execution. Companies that can solve challenges around power availability, efficiency, sustainability, resilience and compliance are well positioned to capture value not only within Europe but internationally. 

What do you think this means for Europe’s wider AI and industrial ambitions?

Data centres are the foundation of the AI economy. Without competitive domestic compute capacity, it becomes harder to build and scale AI businesses, attract investment and support innovation.

The implications go beyond technology. AI is expected to be a major driver of future productivity growth and countries with strong access to compute are likely to capture a disproportionate share of those gains.

What are the biggest barriers stopping Europe from building more AI infrastructure?

Our report identifies three primary barriers: power and grid access, electricity costs and regulatory complexity. 

Securing power and timely grid connections remains a major challenge, while high electricity prices create a competitive disadvantage.

These are structural challenges that subsidies alone will not compensate for.

If Europe does not act soon, what do you think the long-term consequences could be for businesses and the economy?

The risk is that Europe becomes increasingly dependent on infrastructure, platforms and frontier AI capabilities developed elsewhere. That has implications not only for competitiveness, but also for resilience, sovereignty and economic growth. 

Our report argues that AI-driven productivity gains will be an increasingly important contributor to GDP growth over the next decade. Without sufficient domestic compute infrastructure, more of that value creation, along with associated software, R&D and high-productivity jobs, is likely to concentrate in other regions. 

Local compute market is also critical for the continued innovation and development of Europe’s existing supplier base to the data centre market.

Every year of delay widens the gap and makes it more difficult to close.