Rewiring for growth

Deploying new data centres often requires substantial available grid capacity at a single location – an increasingly scarce resource. However, strategic choices by developers and grid operators can help overcome these constraints, speeding up the deployment of new facilities, while also reducing the required infrastructure expansion to accommodate the growing sector.

Time to market is crucial for competitiveness, especially in the fast moving tech industry. Wait time for grid connection is therefore an important criterion for site selection of new data centres. But long connection queues are becoming increasingly common across Europe, as system operators struggle to allocate capacity to new users, deterring data centre developers.

Building transmission infrastructure – which takes a long time – is not the only solution. Alternative forms of grid connections could secure a faster process while also bringing benefits for the grid. Such options are described in the following sections.

2.1.1 Countries could attract 20% more data centre growth by halving grid connection time

The average time to build a new data centre in Europe is approximately four years but varies widely across countries, particularly impacted by availability of grid capacity. In the FLAP-D markets, the historically high concentration of data centres in specific locations has led to a build up of grid congestion, with substations unable to support demand for additional capacities. New facilities wait an average of seven to ten years, or even up to 13, in some primary markets.

Countries with long grid connection queues are losing out on potential growth in their data centre markets – which is being reaped by others whose grids are better prepared to accommodate such demands. Currently, countries with half the connection time of key markets are set to attract around 20% more data centre growth by 2030. 

Italy – where data centres wait as little as three years in the grid connection queue – stands out as a key market poised for significant growth over the next five years, alongside the Nordics and Belgium. This is reflected in the huge capacity of grid connection requests for data centres received by the Italian TSO Terna. By the end of 2024, the capacity of data centre projects in the queue stood at a staggering 30 GW, with almost 80% of these requests being registered in the last 12 months, highlighting the magnitude of project developer interest in newer, more accessible markets. 30 GW is equivalent to around 40% of Italy’s current peak electricity demand, underscoring the scale of pressure that could be placed on the grid if even a fraction of these projects move forward.

2.1.2 Phased grid connections enable collaborative infrastructure build up

Available data suggests that data centres currently utilise only a fraction of their grid connection capacity, maintaining headroom to accommodate future growth. IEA and McKinsey data suggests that, on average, data centres in Europe only used 44% of their annual nominal power rating in 2024. The Irish TSO EirGrid reports that the average load currently drawn by data centre customers is approximately 34% of their contracted capacity. This is similar to reported figures for Norway where, out of the 500 MW connected to the grid, only 150 MW of that capacity was utilised in 2023 – a utilisation rate of just 30%. 

Low utilisation of connection capacity makes data centres ideal candidates for a phased connection approach. Data centres could apply for an immediate connection capacity corresponding to around 50% of their needs, with the remaining 50% being delivered over an extended period of time. This provides both the developer and the grid operator with a clear path to scale, a welcome development in sectors both exposed to significant uncertainties. In the US, this phased approach could purportedly reduce time in the connection queue by 80%, from 5–8 years to 12–18 months.

2.1.3 Non-firm grid connections accelerate access to constrained networks

Another relevant option for new data centres is a non-firm or flexible grid connection – where users agree, in advance, with system operators that their consumption may be interrupted under certain conditions such as extreme weather events. This allows new customers to connect to a constrained grid much faster and may bring additional benefits such as lower network tariffs. This setup is distinct from demand response, which is typically a market-based mechanism rather than a condition specified in a grid connection agreement.

The idea of flexibility for data centres is gaining traction, stemming primarily from a need to overcome growing grid constraints. Adopting a flexible grid connection model can unlock capacity in grid constrained areas and shorten connection wait. A notable global initiative is DCFlex, launched in 2024 by the Electric Power Research Institute (EPRI), in collaboration with leading data centre investors and system operators, which aims to promote flexible grid integration for data centres. 

Data centres are typically assumed to be firm loads, but real life data shows that even for colocation facilities, electricity consumption can drop or increase up to 30% between days, depending on the IT job performed. Some seasonal variation is present due to cooling needs as well. In the case of AI data centres, the electricity demand depends on model training and varies with the usage of the model by individuals and businesses. The load profiles can be optimised through the batching of AI queries and different execution schedules that align with electricity prices, grid congestion and on-site generation. This presents an opportunity for more flexible power consumption, with benefits both for the grid and data centre operators.

The IEA estimates that if European data centres offered just 30 hours of flexibility annually, the grid capacity available for them could more than double. This would not mean that a data centre would be occasionally completely disconnected from the grid – the IEA model assumes that in “80% of the hours of grid stress, more than half of the usual grid electricity supply to the data centre would still be available. In 50% of these hours, around four-fifths of the grid capacity would be available”. The side effects of the flexibility scheme could be offset further by the use of on-site backup generation and battery storage. 

Data on data centre power consumption profiles is scarce, but assuming that this profile is completely flat, which is often done by grid operators, can lead to inaccurate modelling results. It also leaves no room for the assessment of data centre flexibility options, potentially overestimating grid investment needs. In contrast, Belgium’s long-term electricity plan explores high levels of data centre flexibility – ranging from 0% to 50% and up to 75% depending on the scenario. With grid availability becoming a scarce resource, data centre flexibility might soon become mainstream in grid planning exercises.

Transmission grid access is a decisive factor in data centre site selection – yet building new lines and substations can take years. Fortunately, both grid operators and data centre developers can take actions to optimise their planning, reducing grid connection times and investment needs.

In the short term, grid operators can steer data centre developers towards priority zones with available grid capacity – e.g. post-industrial regions or areas with abundant clean electricity. Several countries are already piloting this approach. In France, EDF is offering ready-to-use industrial sites with a grid connection capacity of approximately 2 GW for data centre development through a call for expressions of interest, with the aim of reducing project completion timelines by several years. Similarly, the UK government issued an expression of interest in February 2025 to enable identification of AI Growth Zones (AIGZ), with more than 200 responses received. Apart from identifying the most suitable sites, the call also sought to cluster the developments, looking for locations able to host or developers prepared to establish 500 MW+ of AI infrastructure (equivalent to enough energy to power 2 million homes). Italy’s Terna is also looking to cluster requests from data centres for grid connection to minimise the risk of oversizing infrastructure.

In the longer term, strategic spatial planning comes into play, aligning grid investments with the needs of new generation assets, such as renewables, and demand assets, such as data centres. This requires collaboration between all involved parties, as showcased by the UK’s NESO within the Strategic Spatial Energy Planning process. Such an approach might be challenging due to different planning horizons. TSOs typically require long-term offtake commitments to justify infrastructure investments, while data centre developers operate on shorter cycles, driven by rapidly evolving technology, shifting workloads and uncertain demand growth. Yet if executed correctly, a joint planning effort can provide benefits for all actors: reducing grid investment costs and optimising the use of existing grid infrastructure, and fast-tracking permitting and grid connection for data centre and renewables developers. 

In the most advanced format of joint planning, data centres could be co-located with generation and storage assets, minimising the impact on transmission grids and potentially reducing network charges. But to unlock the efficiency gains that come with optimised energy system planning, grid operators need to first publish grid capacity maps. The availability of these would further attract investment, as data centre developers would identify areas where new capacity can be added quickly.