Reflections on the U.K. government’s pledge to turbocharge the data centre industry

“This is the nation of Babbage, Lovelace and Turing”

The UK government has announced ambitious plans this week for the UK data centre industry. British Prime Minister Kier Starmer set out how he intends to put a stake in the ground in AI, which will require development of numerous hyperscale data centres of 100MW or more. His aim is that the UK will become a leader as opposed to a follower in the pack of the global data centre industry - by its nature an industry where economies in different regions are in direct competition with one other. From a relatively short statement by the Prime Minister containing some emotive themes, we have extracted a few key commitments, and a few key omissions too. 

AIGZs - the new enterprise zones of the 2020s? 

The government has announced the creation of several AI Growth Zones. The first of these will be Culham in Oxfordshire, already home to several advanced scientific research companies. There is little detail on what benefits will exist from a business locating itself in an AIGZ. If it will be anything like previous waves of enterprise zones, we can expect tax incentives.

Businesses situated in enterprise zones created in 2011 benefitted from a business rates discount and enhanced capital allowances on construction costs to reduce their corporation tax bill. The previous wave of enterprise zones in the 1980s brought about a simplified planning process for those within the zone in addition to similar tax incentives. The same is true of the eight freeports created in 2021, which also bestow employment tax incentives on businesses located within their boundaries.

Speeding up the grant of planning permission

The announcement contains a statement that the grant of planning permission will be sped up. It remains to be seen what this means. Under basic principles of the planning regime in England and Wales, a local planning authority has discretion on whether to approve or reject an application to build a data centre within its boundaries. The local planning authority is legally obligated to have regard to the National Planning Policy Framework, its own local plan and other lawful considerations. So long as it does this properly, the decision will ultimately remain the local planning authority’s call. A different regime exists for a nationally significant infrastructure project (NSIP). As things stand, the construction of a data centre of any size does not count as an NSIP and the Secretary of State does not have the power to add data centres onto the list. The construction of a data centre is not a project related to transport, waste, water or energy under the criteria of the Planning Act 2008. The expansion of NSIP projects to include hyperscale data centres could be achieved with new primary legislation introduced by the Labour government.

Adding data centres into the list of NSIPs would take away some of the barriers of parochial reluctance. That does not necessarily mean that planning consent for a data centre would be granted quicker than under the local planning regime. The statutory process of determining an NSIP application is not meant to take more than 17 months from the date of submission. There are statutory time limits for each of the pre-examination, examination and decision stages, however each of these can be extended by the Secretary of State. The foregoing timeframe also does not take into account the significant time needed for public consultation required prior to submission. 

The introduction of a permitted development right to convert a brownfield site into a data centre site is another possibility to speed up the planning process. This would not necessarily obviate the decision-making role of a local planning authority on aspects of a proposed data centre. However, it would by and large take away a local planning authority’s ability to veto a project meeting the initial requirements of the permitted development right. It would therefore be a more radical and effective way to speed up the planning process assuming suitable buildings to convert can be found in the right locations.  

The energy demand of data centres

Energy demand of data centres is an increasingly well-known hurdle to the Prime Minister’s ambitious plans. Running a ChatGPT query requires 3 watt-hours of electricity. That is ten times the amount required to run an ordinary Google search (albeit one could query whether a ChatGPT response is more satisfying, necessitating less follow up searches and therefore energy neutral). It takes 12 watt-hours to generate a single image using AI, which is the same amount it takes to fully charge a smartphone. If we are to harness AI, as the Prime Minister says, to predict every no-show doctor’s appointment, plan every primary school lesson, and spot every pothole up and down the country before it appears, that is going to need a whole lot of extra energy. It will require even more than the exponential growth in demand currently expected.   

The Prime Minister said in his speech that if a data centre needs better access to power (which is like asking a boat if it needs access to water), he will get it a grid connection at speed. Getting a grid connection is just one piece of the puzzle for a data centre owner. Under section 16(1) of the Electricity Act 1989, an electricity distributor (DNO) is under a duty to make a connection between the distribution system and premises when required to do so, save for excepted circumstances. However, DNOs in the UK are already massively overstretched and several commentators have voiced concern that it is a pipe dream for data centre owners to obtain sufficient security of power supply from the grid. This creates a major obstacle to the swift expansion of data centres. 

An alternative option for the hyperscaling of data centres is for owners and operators to procure their own private electricity supply. This could be done through on-site generation or alternative power sources. It is no coincidence that the Department for Science Innovation and Technology recommends that AIGZs are post-industrial towns with existing energy capacity or locations in coastal Scotland. On-site wind and solar generation form a useful part of the energy mix for data centres but for hyperscale sites, this will not be sufficient. Nuclear power is touted as an alternative, which could be generated on or near site via a small modular reactor (SMR).

Currently there are no small modular reactors operating on land in a commercial setting. The technology is being developed simultaneously in several countries. The first SMR is expected to be commissioned in 2026 in Hainan province, China and the reactor will have a capacity of 125 megawatt-hours. With the rapid fall in the cost of renewables, some experts believe that SMRs will never be economically suitable for municipal energy supply. However, SMRs are very promising in the context of a data centre for several reasons. The first is that an SMR can generate a huge amount of stable, reliable, always on energy from a small amount of land. For example, a NuScale Voygr-12 model SMR could generate 924 megawatts of energy on a 0.13 square kilometer site. To generate that amount of power with solar panels would require a site the size of 6,000 football pitches. 

Another advantage of SMRs in the context of data centres lies in the word ‘modular’. Data centre operators tell us that it is very hard for them to predict how much power they will need in five or ten years’ time. The beauty of many SMR models is that they can be scaled up as and when a growing data centre’s needs expand. The same cannot be said of a wind farm, a gas power station or even a large scale nuclear plant. It is uneconomical to return to a generating site every few years to erect a few more wind turbines or other additional plant needed to increase generating capacity. If SMRs are to become a reality for data centres, then the UK government will need to think of ways to ensure that their deployment can keep pace with the explosion in demand for AI services themselves. We predict that future of AI and nuclear is intertwined. Data centre demand might just be the economic fuel that ignites the sustainable SMR development chain reaction the world needs.