Europe’s slow and steady approach to AI could be its edge

Europe, with its fragmented markets, is often said to be working in the shadow of the US and China when it comes to scaling AI.

But factors challenging its growth as a major player could give it an advantage when it comes to future-proofing the critical silos powering the AI ​​boom.

The world is racing to double or even triple all data center capacity built in the last four decades, and McKinsey estimates the cost of that construction will cost as much as $1.5 million, Pankaj Sachdeva, McKinsey’s senior partner in technology, told CNBC. $7 trillion by 2030.

He expects the United States to take the lion’s share of activity, but Europe will “continue to develop at a fairly meaningful rate” to nearly double its current capacity.

“Europe is actually participating in this infrastructure building and keeping up with it, or we think it will keep up with it,” Sachdeva said. he added.

To get to that point, the bloc must overcome major hurdles in access to power and regulation, experts told CNBC.

In terms of network congestion, Italy is one of the countries on the winning side. There is connection time up to three years That compares with Europe’s four-year average, according to energy think tank Ember.

On the losing side are Germany, the UK, Ireland and the Netherlands again; “Here, we either don’t have grid capacity right now, or there is such a deficiency in the system that there is a de facto moratorium for the foreseeable future.”

Walia said that while the differences between European countries are significant, it “will be difficult” to catch up with the U.S. in the short term — where deregulation and massive investment are driving much faster growth. There are about 200 to 300 data centers in most European countries, he added, but “there are around 5,400 in the US.”

The restrictions are causing some movement away from the traditional FLAP-D markets of Frankfurt, London, Amsterdam, Paris and Dublin and investment in data centers where resources are abundant and stable.

Some efforts were also made to develop projects faster. For example, there have been instances in the UK where central government has overridden local government in approving data centers that had previously been rejected. Last year, the country designated data centers as Critical National Infrastructure and emphasized their importance to its economic agenda.

This is a particularly challenging problem for Europe, where energy bills are soaring after Russia invaded Ukraine. The UK has the highest energy costs in Europe. 75% higher than before full scale attack.

While this may be a disincentive to set up shop in a particular location, operators aim to offset this with network congestion times.

Grid congestion has also fueled debate about how to deliver energy in Europe, according to Kevin Restivo, CBRE’s European data center research leader.

“There’s a lot of speculators coming into the queue, and those speculators are making it even more difficult because they have no intention of building data centers. Maybe they want to hand over the power to someone else,” Restivo told CNBC.

The United Kingdom, for example, operated on a first-come, first-served basis; this meant that the importance of the project was not taken into account in the decision on who would receive power first.

However, the system is currently ”first ready, first connectionA process through which completed projects can progress in the future Link queue designed in part to combat speculation. The reforms show how energy and infrastructure structures are forcing legacy systems to evolve and paving the way for further innovation.

At the same time, the steady pace of change is allowing developers to be more intentional about what, where and how to build; This means Europe can place greater emphasis on state-of-the-art facilities.

Walia said the quickest way for Europe to overcome these challenges is not to wait for new grid connectivity but to ask “where I have good grid connectivity to a sector that is currently in decline,” as such sites can be transformed from industrial hubs into technology hubs.

Europe is unlikely to take the lead in building facilities for AI hyperscalers or training AI. this race is considered won – but the general consensus is that it may succeed in smaller, cloud-focused and connectivity-style facilities that require large amounts of fiber to go in and out, and in facilities designed for AI inference.

In fact, there are few basic model developers on the continent; France’s Mistral is the best known, but McKinsey sees 70 percent of all AI demands from inference.

There don’t appear to be “too many” massive AI-related data center sites announced on the continent, or that they are “slightly overpriced”, according to Seb Dooley, senior fund manager at Principal Asset Management.

“So actually from our perspective you’re seeing these areas are well protected from any potential oversupply bubble that could arise,” he added, as the cloud is well established.

Principal Asset Management expects AI inference to occur in the same facilities as the cloud; This has already happened on some cloud sites in the US. This gives investors a “pretty good upside” without the speculative risk that comes with other AI investments, the fund manager said.

This is also an opportunity for Europe. Inference will likely need to exist within European borders, Dooley said, driven by a broader crackdown on sovereign AI. However, they have different technical requirements; density tends to be higher than 20 kilowatts per rack for traditional cloud; This means data centers that want to do both need to take this into account. Inference also requires different cooling systems.

“That means you have to design these facilities to be somewhat flexible and robust so that you can switch between two different systems as requirements change,” Dooley added.

So the joy of moving slower and more carefully in Europe is that there is time to think about such things.

Some member states go further. Walia pointed out recommended sustainability requirements In Spain, data center developers will be seen reporting socio-economic impact. “No one asks this in the United States,” he said.

But Dooley hopes tight regulations will work in Europe’s favor in the long run as data centers are integrated into local communities “rather than completely destroying everyone’s life sometimes”, noting that sustainability is one area where the bloc is “very good at innovating”.

“From my perspective, Europe stands out as quite interesting; it feels like a much safer investment situation from a capital market perspective than the US,” Dooley said.

“A lot of that is because it’s hard to build in Europe. We have a lot of constraints, but really, the harder something is to copy, the more long-term value what you have, the more likely people are to reuse and find creative solutions to reuse assets,” he added.

Ultimately, investors and developers may have no choice but to back Europe thanks to dominant AI being an “underrated” driver of data center construction, Jim Wright, manager of the Premier Miton Global Infrastructure Income Fund, told CNBC.

As a result, Europe has the 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 constructive construction for the latter.

But there will be no one-size-fits-all approach to building data centers in Europe. “The industry right now is still in the ‘figuring out exactly what it needs’ phase,” Dooley added.