Data center construction is expanding at record pace, and they require enormous amounts of power. The good news is that, for the most part, enough power currently exists. At peak times, renewables generate more electricity than Europe can use, and at historically low cost. The bad news is that unprepared and outdated grid infrastructure is a serious bottleneck for the power reaching where it needs to go. The grid cannot absorb and utilize the energy that renewables produce at peak times, and battery storage remains insufficient at scale.

This is a problem all over Europe, and all the major data center hubs – Frankfurt, London, Amsterdam, Paris, Dublin – are hitting capacity limits. The IEA estimates that around 20 percent of planned data center projects are at risk of delays due to electricity supply bottlenecks.

Major energy projects being unable to connect to the grid is not a new phenomenon. Solar generation capacity in Europe expanded rapidly over recent years, but faster than the investment in grid infrastructure, storage, or local demand could keep up.

Supply ended up outpacing the system’s ability to absorb and distribute it, so prices turned negative during peak solar hours, and curtailment rose sharply – in Spain, around 16 percent of solar generation was curtailed in the first quarter of 2026 alone, double the level from a year earlier, and investors began offloading assets at a fraction of their original valuations.

Solar deployment is now slowing across Europe, not because demand for clean power has fallen, but because projects were built without sufficient coordination with the systems around them.

The data center buildout is following the same pattern, but at greater speed and higher stakes. The lesson from solar is that capacity alone is not enough. Generation, demand, and infrastructure have to be developed together, or the system breaks down.

One such risk of failing to power data centers is falling behind in the global AI race, something Europe cannot afford. Without a faster infrastructure buildout, Europe risks constraining not only the growth of its economies but its ability to compete in the technologies that will drive future productivity and innovation. As investment follows reliable infrastructure, so too will the world’s leading AI companies and talent.

Another risk is that data centers will turn to other power sources, undermining Europe’s expanding renewable capacity. Looking at North America for direction of travel among tech giants, Microsoft has structured a 20-year agreement to restart a nuclear facility in the United States. Amazon has acquired a nuclear-adjacent campus to secure direct, dedicated power supply in West Texas. Google has committed to small modular reactors in Tennessee with first power targeted around 2030.

Meanwhile in the UK, more than 100 data centers have already submitted requests for gas connections, in some cases as primary supply rather than backup (The Guardian, 2026). While nuclear power produces very low greenhouse gas emissions, it remains politically contentious because of its high upfront costs, long development timelines, and legacy safety concerns.

In short, Europe faces a growing infrastructure gap. Data centers are expanding rapidly to support the AI economy, but grid constraints are preventing access to clean power. Without significant investment in grid capacity, operators will increasingly turn to alternative power sources such as gas and nuclear, undermining both climate objectives and Europe’s long-term competitiveness.

One practical response to this problem is the energy hub model. Rather than adding several large uncoordinated connections to an already strained grid, energy hubs co-locate solar generation, battery storage, and data center load at a single grid node. Essentially, rather than adding several large connections to the grid, energy hubs bundle and balance power assets locally, connecting them to the grid at a single point. This presents network operators with a manageable profile rather than multiple connections.

The flexibility this creates significantly increases effective grid capacity. A Duke University study found that grids could absorb up to 100GW of new load without any infrastructure upgrades, if large users are willing to reduce consumption for less than one percent of the time. The energy hub model is designed to make exactly that possible.

The concept is already working in practice. In Ireland, a data center developer, an energy company, and the national grid operator EirGrid pooled resources to build a shared substation that serves both a data center and an offshore wind farm. By planning and funding it together rather than separately, both projects connected to the grid two years faster than they would have otherwise.

Other large technology companies are investing in this type of connection. Alphabet acquired energy developer Intersect Power in 2025 for $4.75bn to develop power generation alongside new data center capacity. Further, Meta, Amazon, Google and Microsoft together accounted for roughly half of the 56GW of clean power purchase agreements signed globally in 2025.

The solution is not to build more renewable energy in isolation, or to default to nuclear because grids cannot keep up. Europe already generates more clean electricity than its grids can handle at peak times. The urgent problem is coordination. Right now, generation, storage and demand are planned separately and added to the grid one at a time. If instead they were designed together and connected at a single point, each new data center would make the grid easier to manage rather than harder.

The businesses and markets that figure this out first will be better placed to attract investment, meet their climate commitments and power the next generation of AI infrastructure.