Energy The shape of AI in 2025

The choice of locations for data centres and AI

A number of countries have identified AI as a strategic and economic priority.

The US has reiterated its commitment to foster a thriving environment for AI, and competition to be at the forefront of the AI sector in Europe is heating up.

The UK government published an AI Opportunities Action Plan in January, while France announced significant AI investments at an AI summit in Paris in February. Both countries have identified the role of data centres as core to their aspirations.

Choosing where to locate data centres is driven by physical factors such as power and telecommunications infrastructure and land availability, commercial factors such as power prices, and regulatory factors such as grid connections and energy sourcing requirements. The data centre use case determines which factors are weighted more highly.

For example, data can be stored and served a long way from the user, so data centres to support AI do not necessarily need to be located in the same country as the demand. In the EU, there are agreements that allow data to be stored outside the EEA if certain GDPR conditions are met. Nevertheless, there is an acknowledgement that locating data centres within the EEA has the benefit of bolstering resilience and avoiding overreliance on foreign tech companies.

However, distance increases the cost of connectivity and creates latency (the time delay between a user’s request and the system’s response). This can be a concern for AI inference use cases that are latency sensitive, such as fintech, automatic trading, high-frequency trading and some audiovisual services, including Netflix, YouTube, and Facebook. Proximity is a factor for these applications.

Attracting data centres

Governments have less direct control over the relevant commercial and physical factors, at least in the short term, but have levers they can pull to create an attractive regulatory environment for investors.

We recently reviewed the key regulations that data centre operators face to understand how the regulatory environment could deter them from locating in a given country. Our work focused on three countries - the UK, the Netherlands, and Germany - and we reviewed regulations across five categories.

Grid connection and energy system integration
Energy sourcing and utilisation of waste heat
Sustainability and energy efficiency
Data protection and GDPR
Financial incentives

Our findings are shown in the chart below, where the position on the barrier spectrum from low to high shows a relative assessment of the role of regulation in the three countries. Obtaining a grid connection is a challenge in all countries (with specific exceptions), but other regulations are more of a factor in some places than others.

In Germany, the requirement that at least 10% of the heat generated by a data centre must be reused by 2026, increasing to 20% by 2028, could be a barrier to locating data centres in that country. Germany’s stricter data protection regulations are seen in some quarters as positive; however, at the same time, they can be a disadvantage for the development of related AI products and services. In the Netherlands, data centres have to comply with zoning regulations that vary between municipalities. Currently, there are few areas where they can realistically be built.

The UK has historically taken a lighter-touch approach to regulating data centres. The government’s recent action plan takes this one step further by establishing a proactive set of measures to incentivise data centre operators to choose the UK. Data centres that are located in designated AI growth zones will have priority access to a grid connection. However, power prices for large energy users (especially those not in the scope of the Energy Intensive Industries (EEIs) exemption) are significantly higher than in other proactive European countries. In France, for example, where power prices for large energy users are much lower, the government has already identified 35 key sites as ready to use for data centres.

What is clear is that no one country can claim to have the best environment for data centres across all criteria. Investors will thus need to carefully weigh up the pros and cons depending on their specific needs and expectations of future strategy and regulation.

A passive approach is risky

In the competition to be at the forefront of AI, countries take very different approaches, and the takeaway from our research is that regulations of data centres matter. Countries are in global competition for large-scale data centres and AI clusters. Those that take a passive approach - unlike the UK, France and the US - risk losing out in the race to attract and benefit from the AI sector.