AI data centres are following the power: could that help wind?
The AI boom is fuelling a shift in data centre growth beyond current hubs and future smart planning could take it where the green power is, says one expert
The energy demand boom to serve Europe’s AI-fuelled data centre industry is only just getting started, but it’s already helping pick winners for investments worth billions and driving a shift that could open new opportunities for future wind and other green power development.
New research points to a move in highest data centre power demand growth away from what’s known in the IT industry as FLAP-D – Frankfurt, London, Amsterdam, Paris and Dublin – to new regions elsewhere in the continent.
The reason is all about access to the grid, exacerbated by the mismatch between a data centre sector moving at light speed and the far slower energy industry, said Elisabeth Cremona, a senior analyst at energy research group Ember.
While data centres will drive up power demand everywhere, the FLAP-D locations will see expansion between 1.5-2.5 times between 2024 and 2035, according to Ember’s calculations in a new study of the sector. That contrasts to growth predicted up to fivefold in Nordic markets and a tripling or more in southern European countries such as Greece and Italy.
It’s no coincidence that with the exception of France, the FLAP-D markets are already suffering chronic grid congestion, with waits for a connection of between seven and ten years. Ireland’s grid operator, for example, has had what amounts to a moratorium on new data centres in place since 2022, such is the pressure on the system.
For the big data centre operators – which can build a facility in as little as six months – the delay is excruciating as they face the looming demands of AI, which will help drive up European data centre power demand by 150% by mid next decade.
For the markets seeing slower growth in data centre demand, the result is a blow to their ambitions in one of the most prized sectors of economic activity, said Cremona.
“It's very clear that investors for big tech in the IT industry are not going to wait around for the infrastructure to be built,” she said. "They will simply move."
"There are a number of countries where this infrastructure is ready to go, and they prefer to be located there. It's those countries that are going to cash in on the economic opportunities that actually come with data centres. If we look at the Netherlands, 27% of all jobs are directly or indirectly related to [the data centre sector].”
“It was very clear to me that we were seeing this change in investment trends, which is something we don't often think about – that grids are actually an economic tool, a strategic pull for investment. At the core of it, grids are now deciding effectively where that investment will go. It's quite a big revolution.”
According to Cremona, the markets set for the highest data centre growth aren’t there by accident but thanks to what she calls “innovative and dynamic transmission system operators (TSOs)”.
In Denmark, for example, “I found plans as early as 2017… where they were looking to position themselves as a data centre hub. They started building substations at least as early as that date to attract this kind of investment.
“And I think now they're expecting something like a tripling of data centres in the country in the next five years. So, we're looking at very, very fast development, huge investment, but it's coming out of a deliberate decision from a TSO that chose to do that.”
Could wind power for AI 'training' be a match made in heaven?
For wind and other green power developers, the trends flagged by Ember’s research could potentially unlock a major new offtake opportunity down the line as the EU, national governments and TSOs race to maximise their shares of the data centre cake.
That includes a recognition of the need for smarter spatial planning that would strategically locate power-hungry facilities near to large-scale renewable generation – already the case in wind-rich Finland, which is tipped to see a fourfold increase in data centre power demand on Ember’s figures.
AI ‘training’ data centres, where both the computing workload and energy needs are the highest, do not face the same imperative to be nearer data demand centres as other classes of facilities, where latency – the lag experienced by data users due to geographical distance – is more of a consideration.
Tapping Scotland's wind power for $5.5bn data centre
Plans have been tabled to transform an iconic UK post-industrial site into a £4bn ($5.5bn) AI data centre served by Scotland’s abundant wind power.
Developer Apatura hopes the site at Ravenscraig near Glasgow – previously home to one of Britain’s largest steelworks – will be selected under the UK Government’s AI Growth Zone initiative, which aims to identify priority areas for development.
The plan would tap 550MW of grid links coming online by 2030 and “constrained wind and solar generation that’s often wasted on windy, sunny days – helping to reduce grid congestion and lower consumer bills throughout the UK”, Apatura says.
The developer reckons Ravenscraig data centre would contribute an additional 0.4% to Scotland’s annual GDP once operational.
Apatura has a total of 1.6GW of grid connections earmarked in Scotland for development of five data centre sites.
The initiative enjoys strong political support. Local member of parliament Pamela Nash said the scheme could become “a world-class digital and AI hub powered by Scotland’s renewable strength, creating thousands of local jobs during construction, driving billions in economic value, and transforming post-industrial land into the beating heart of a green, future-ready economy”.
That opens the potential for European regions such as Scotland, with its vast wind resources, to host a new class of heavy power user.
The Ember report says: "Locating large electricity demand nodes close to clean energy supply infrastructure lowers grid investment needs. It also allows data centres to stabilise their electricity costs through near-site PPAs, which can deliver network charge savings as well."
Cremona added: "There is the possibility, especially for I think, something like an AI training type of data centre, to be located in more rural places where there is that renewable capacity readily available.
"That's always a trade-off [and] it depends on the type of data centre you would be building. So, there is that potential, and it could be a great way to utilise the resources that you have there."
(Copyright)