The rise of artificial intelligence (AI) has skyrocketed the demand for datacentre infrastructure. Many existing facilities simply can’t handle this new wave of energy and capacity needs. We’re not just talking about a little extra space; AI-ready datacentre capacity is extremely limited, especially with hyperscalers and cloud providers already taking most of it.
Hyperscalers like Amazon Web Services and Microsoft dominate the datacentre scene in Europe. They’re not just content with what they have. They’re scrambling to secure even more space for their expanding digital services, trying to stay ahead of their competition. This is ramping up pressure on datacentre developers to ramp up supply throughout Europe.
AI technologies, especially generative AI, tap into far more power than traditional workloads. Creating a training model that needs to analyze countless images to learn facial features demands energy that far exceeds what typical computing setups require. As countries and industries rapidly adopt AI, the demand for datacentre space is only increasing. The International Energy Agency predicts that electricity needed for AI-powered online searches could skyrocket by ten times. By 2026, they expect that datacentre electricity consumption will leap from about 460 TWh in 2022 to over 1,000 TWh.
To keep up, datacentre designs are evolving. These facilities now need to accommodate heavier processing power, which generates extra heat. That means cooling must adapt as well. Datacentre operators are now investing in liquid cooling systems—think direct-to-chip solutions or immersion cooling—to handle these higher power needs. Installing these advanced cooling setups requires more space, so many are opting to construct entirely new liquid-cooled datacentres from the ground up.
In Europe’s metro areas, where most colocation datacentres are situated, finding enough power is tough. Operators face limited grid capacity and must navigate sustainability issues, often competing with residential developers for energy. Land near high-speed network infrastructure is also getting scarce. This scarcity drives prices up, pushing datacentre operators to look further for suitable sites.
The race to meet the demand for datacentre capacity is intense, and the pressure is mounting from next-generation AI applications. As a result, available datacentre space in major European cities is dwindling.
What about existing colocation facilities? They can support AI workloads if retrofitted with the right hardware and cooling solutions. But if AI usage keeps skyrocketing, we’ll need new datacentre capacity. That’s not likely to happen on a large scale in major hubs like Frankfurt, London, Amsterdam, Paris, and Dublin, where land and power are hard to come by.
To meet AI’s requirements, the industry must shift its development strategies. Datacentre operators may need to look beyond traditional markets and explore smaller, secondary locations in places like the UK or France, where power and land could be more accessible.
In the UK, we’re starting to see this shift. Investors, hyperscalers, and datacentre providers are eyeing land for AI-ready facilities. Currently, about 56% of the UK’s colocation datacentres are within 30 miles of London, but that’s starting to change. For instance, Virtus recently acquired land in Saunderton, where they intend to add 75MW of AI-related capacity.
Low-latency connectivity will also become crucial as inference AI rolls out. While datacentres are already integrating equipment for AI training, the next step will focus on inference capabilities.
The AI boom has undeniably shaken up the datacentre market. Not only are we facing a capacity crisis, but building new AI-ready datacentres is fraught with complexity, particularly as European power grids struggle to keep up. Alternatives like Small Modular Reactors (SMRs) and renewable energy sources—wind and solar—are being considered, but they aren’t ready for widespread deployment yet.
Finding new datacentre sites with both adequate power and high-speed network access is proving challenging. To meet AI’s growing demands, operators will undoubtedly need to venture beyond traditional datacentre markets.