Europe, with its fragmented markets, is often said to operate in the shadow of the US and China when it comes to scaling AI.
But the very factors challenging its growth as a major player could give it an edge when it comes to future-proofing the critical warehouses powering the AI ​​boom.
The world is racing to double, if not triple, the total data center capacity built over the past forty years, Pankaj Sachdeva, a McKinsey senior partner in technology, told CNBC, as McKinsey estimated build-out costs. $7 trillion by 2030.
He expects that the US Europe accounts for the lion’s share of activity, but “will continue to build at a fairly meaningful rate” to nearly double its existing capacity.
“Europe is really participating in building this infrastructure, and is actually keeping pace, or we think it will keep pace,” Sachdeva added.
To get there, the bloc must overcome key chokeholds in access to power and regulation, experts told CNBC.
Winners and losers
A critical bottleneck for Europe is access to electricity, with energy costs and availability shaping investment flows across the region. The Nordics and Spain have seen increased appetite for data center builds due to energy increases due to hydropower and renewables, while Germany and the UK may be less attractive due to energy supply constraints.
In terms of grid congestion, Italy is one country on the winning side. It has connection time for three years That compares with a four-year European average, according to energy think tank Amber.
On the losing side again are Germany, the UK, Ireland and the Netherlands, “where we basically don’t have grid capacity right now or we have such a shortage in the system that there’s effectively a moratorium for the foreseeable future,” Jags Walia, head of global listed infrastructure at the Van Lenscott campaign, told CNBC.
Walia said that while the differences between European countries are significant, it will ultimately be “difficult” to catch up to the US in the short term – where deregulation and massive investment enable a very rapid build-out. Most European countries have about 200 to 300 data centers, he added, but “the US has about 5,400.”
Some diversification away from the traditional FLAP-D markets of Frankfurt, London, Amsterdam, Paris and Dublin results in constraints and drives investment in data centers where resources are abundant and stable.
Where Europe, from my perspective, stands out as quite interesting, it looks like a much safer investment case
Seb Dooley
Senior Fund Manager at Principal Asset Management
There have also been some attempts to speed up the project. For example, in the UK, there have been cases of central government overruling local government to allow data centers that were previously denied. country last year Designated data centers Critical National Infrastructure, highlighting their importance in its economic agenda.
A powerful obstacle
Energy consumption from power-hungry data centers could double from 460 TWh in 2022 to 1,000 terawatt-hours (TWh) in 2026, driven largely by AI, according to the International Energy Agency.
According to Walia, the largest cost component of a data center is electricity, although new, state-of-the-art facilities can reduce the burden.
This is a particularly sticky problem for Europe, which saw its energy bills skyrocket when Russia invaded Ukraine. The UK has the highest energy costs in Europe, which is around 75% more than before A full scale attack.
While this can be a deterrent to setting up shop in a particular location, operators aim to balance it with grid congestion times.
According to Kevin Restivo, CBRE’s European data center research lead, grid congestion has also fueled discussions about how to get power to Europe.
“You get a lot of speculators in the queue, and those speculators make it more difficult because they have no intention of building data centers. They just want the power, maybe, to flip it to somebody else,” Restivo told CNBC.
The UK, for example, operated on a first-come-first-served basis, meaning that project importance was not taken into account in deciding who got the authority first.
However, the system currently ‘First ready, first connected,’ process where completed projects can jump forward A connection queue, which was designed in part to combat speculation. Reforms show how building energy and infrastructure forces older systems to evolve and sets the stage for further innovation.
At the same time, the steady pace of change allows developers to be more deliberate about what, where and how they build – meaning Europe can put more emphasis on state-of-the-art facilities.
The fastest way for Europe to tackle these challenges is not to wait for a new grid connection but ‘where do I currently have a good grid connection with declining industry?’, Walia said, as such sites can be repurposed from industrial to tech hubs.
Opportunity in AI prediction
It is unlikely that Europe will lead the way in building facilities for AI hyperscalers or for training AI — That race is all but won — but the general consensus is that it can excel in small, cloud-centric and connectivity-style facilities that need large amounts of fiber going in and out, as well as those designed for AI inference.
Indeed, there are a few basic model developers on the continent, with France’s Mistral being the best known, but McKinsey sees 70% All AI demand comes from prediction.
According to Seb Dooley, senior fund manager at Principal Asset Management, there aren’t “a lot” of AI-related giant data center sites being announced in the continent, nor are they “of a slightly overpriced nature”.
“So, actually, you’re looking for these areas, from our perspective, that are well protected from a potential oversupply bubble that could go through,” he added, because the cloud is well established.
It is largely powered by AI, but non-AI workloads are also expected to tick upwards
Principal Asset Management expects AI prediction to take place in facilities similar to the cloud, which is already the case at some of its US cloud sites. This gives investors “pretty nice upside” without the speculative risk that comes with other AI investments, the fund manager said.
It is also an opportunity for Europe. Speculation must exist within European borders, Dooley said, driven by the wider Push for sovereign AI. However, it has different technical requirements; Density is higher than the 20 kilowatt racks for traditional cloud, meaning data centers that want to do both should be a factor. The forecast also requires different cooling systems.
“That means you have to design these facilities to be flexible and robust so you can change between two different systems as needs change,” Dooley added.
Enjoying a slower and more considered pace in Europe, therefore, is the time to think about such things.
Risk of stranded assets
The pace of AI development has led to widespread chatter of a bubble, which would result in piles of trapped wealth if it were to pop. If AI keeps its pace, which many believe it will, there is still a risk that a data center built today will not be suitable in the future as the technological needs of AI change.
To help, investors are focusing on securing clients before the ground breaks. Restivo said speculatively-built data centers are “mostly a relic of the past.” Developer-operators often lock customers into 10-to-15-year terms, he added, which also spells obsolescence.
However, it is a different case if the tenant itself is a startup or a young company. Neo-cloud providers, for example, carry “significant risk” and short tenures of five to seven years, Restivo said.
“These are companies that haven’t returned capital to shareholders, have unproven business models, and are in dire need of capacity in the short term,” he said, adding that there is “a lot of skin in the game for developer-operators” who work with neo-clouds. However, some debt financiers and developers are “increasingly comfortable” with these terms, Restivo added.
There can be problems repurposing brownfield sites, however, if the data centers are replacing industrial plants that are still in operation – meaning job losses. European policy requires developers to report energy and water consumption of data centers as well as justification for a specific location.
Some member states go further. Walia pointed out Proposed stability requirements In Spain, in which data center developers will report the socio-economic impact. “Nobody asks about it in the US,” he said.
But Dooley expects the tighter rules to work in Europe’s favor in the long run, as data centers will be integrated into local communities “rather than being the complete blight on everyone’s lives that they can be sometimes,” he said, adding that sustainability is an area where the bloc is “very good at innovating.”
“Where Europe, from my perspective, is very interesting, if we look more from the capital markets side than the US, it looks like a safer investment case,” Dooley said.
“A lot of it comes from the fact that it’s hard to make it in Europe. We have a lot of barriers, but, in reality, the more difficult something is to replicate, the more long-term value you have, the more likely people are to reuse, come up with creative solutions to reuse assets,” he added.
Ultimately, investors and developers may have no choice in the matter but to support Europe’s sovereign AI — an “unpredictable” driver of data center build, Premier Mitton Global Infrastructure Income Fund manager Jim Wright told CNBC.
Overall, Europe has an opportunity to innovate and create long-term value for both investors and citizens. Scarcity increases profitability and resilience for the former, while regulation encourages sustainable and creative production for the latter.
However, there will not be a one-size-fits-all approach to building data centers in Europe. “The industry is still very much ‘figuring out exactly what it needs’ at the moment,” added Dooley.
