The UK is becoming a hotbed for data centre development as it continues to capture the attention of global investors. But will its growing status as an AI superpower come at a cost? BCIS chief economist, Dr David Crosthwaite, explores the economic and cost implications of the data centre craze.
UK data centres: economic goldmine or cost headache?
Advances in AI continue to supercharge demand for data centres worldwide. The UK is at the forefront of investor interest with US tech firm, Equinix, recently acquiring a major data centre site in Hertfordshire in a deal worth nearly £4 billion.
It’s estimated there are more than 470 data centres and counting in the UK with as many as 100 in the pipeline according to analysis of planning applications shared with BBC News(1).
International appetite for data centres in the UK makes them a useful vehicle for growth, but with a creaking grid and competing infrastructure priorities, are they more cost headache than economic goldmine?
Analysis by non-profit techUK estimates the construction and operation of data centres could deliver an additional £44 billion in Gross Value Added annually to the economy between 2030 and 2035(2).
In a bid to support this and respond to the burgeoning demand for a domestic AI ecosystem, the UK government has taken credible steps to expand the nation’s data centre network – sealing the £31 billion ‘Tech Prosperity Deal’ with US tech firms and announcing a new taskforce to fast-track progress on the North East’s AI Growth Zone.
When the time comes, planning reforms introduced in the last year should help to fast-track data centre construction too.
However, making a new generation of data centres operational is another ball game entirely.
Data centres operate 24/7 and currently account for 2.5% of the UK’s total electricity consumption(3).
Growing AI adoption means more high-performance accelerated servers (hardware that supports AI functionality) are being deployed which is increasing the amount of power data centres are burning through.
This is a problem of both how much energy data centres need, and the concentration of that energy demand geographically.
The UK’s grid is old and as it stands, could not effectively support a new generation of power-hungry AI data centres, particularly where clusters of them create localised strain.
The obvious answer is to build new grid infrastructure to improve the integration of data centres. National Grid is building a new substation site in Buckinghamshire for this very purpose(4). The site will feature two substations and connect more than a dozen data centres to National Grid’s network.
Grid upgrades and the construction of data centres themselves are inherently good for the economy, helping to boost construction output, increase job opportunities and encourage investment in local economies.
But timing is everything and the arrival of more data centres well in advance of the UK’s energy upgrade could bring unwanted cost headaches.
One of the biggest cost risks is uncertainty over the demand you’re building for.
It’s widely recognised that data centres are and will continue to load pressure onto the grid but there doesn’t appear to be a consensus on what upgrades will be needed and by when.
Regular, unprecedented demand spikes could spur reactive investments in new grid upgrades – the cost of which would likely be paid for by data centre operators but could indirectly fall into the lap of everyday consumers and disrupt spending and economic growth as a result.
There’s also the added issue of delays connecting organisations to the grid which has led to data centres absorbing grid capacity at the cost of other development progress.
This is a big concern given the immediate economic value in increasing construction output amid a stagflationary environment.
The good news is these are problems with a lot of eyes on them.
Later this year, National Grid and partner, Emerald AI, will trial the mediation of a data centre’s connection to the grid using an AI-powered platform(5). The smart mediator, known as ‘Emerald Conductor’, will hopefully adjust the data centre’s energy consumption and prove usage flexibility is possible.
The transformation of Cottam Power Station into the UK’s first nuclear-powered data centre is hugely promising too.
Energy stop-gaps such as on-site small modular reactors are an opportunity to reduce data centre grid reliance sooner, and together with the potential ability to dial down consumption at peak times, are a really viable answer to managing the AI boom.
There’s also the possibility of including data centres in policy exemptions that reduce electricity bills for energy-intensive industries.
The government is currently consulting on the fine print of its new British Industrial Competitiveness Scheme and factoring data centres into this could help to improve the UK’s current reputation for high energy costs and the impact these have on project viability.
Grid connection delays are still a sticking point – it’s a tricky balance of maintaining the flow of development and ensuring prospective data centre investors remain engaged – but the government’s new Connections Accelerator Service could be a much-needed lifeline.
The service, which is due to be rolled out by the year’s end, is intended to help ‘demand projects’ (those creating high-quality jobs and delivering the most economic value) secure grid connections.
It’s not really clear how the service will calculate the value of projects but any reduction in the connection queue would be welcome.
Energy security remains the biggest growing pain for UK data centres but even with this drawback, the sector continues to attract strong international investment.
The UK’s place at the heart of the AI gold rush will ultimately hinge on how well and how quickly energy use, grid connections, and new supply solutions are managed.
To keep up to date with the latest industry news and insights from BCIS, register for our newsletter here.