Pros and cons of Small Nuclear Reactors in the data centre construction industry

Following a competitive two-year bidding process with two US rivals, Rolls-Royce was recently awarded the contract by state-owned Great British Energy - Nuclear: to become the first company to build small modular nuclear reactors (SMRs) in the UK. 

Alongside £14.2bn of investment already pledged for the Sizewell C power station in Suffolk, the advent of domestic SMRs, to which the government is pledging over £2.5 billion, is part of its wider effort to position Britain at the forefront of nuclear energy technology. It is anticipated that Sizewell C will produce 3.2 GW of power, while each of the three Rolls-Royce’s SMRs will provide about 470 MW. 

SMRs are seen as a potential power source for datacentres because they provide a reliable, low-carbon, and scalable energy supply. Globally, the exponential growth in demand for AI-related services is driving demand for AI data centres that require consistent power delivered by SMRs. 

Companies such as Google, Amazon and Oracle have recently announced plans to use SMR energy to run their data centres. To manage the huge technological demands, AI needs bigger, more complex and power-hungry data centres, which supply chains and national power grids may find challenging to meet. 

The news about Rolls-Royce building SMRs in the UK is welcome. Based on our project teams’ on-site experience across EU member states, and our preconstruction activities, such as viability, feasibility, and buildability analysis, it does raise several challenges in relation to how they will work in practice to service datacentres.

Critically, the SMR model is yet to be commercially demonstrated: no sites are yet fully operational anywhere in the world.

The first point of concern is the temporal disconnect that exists in project timelines. The construction and commissioning of nuclear facilities involve lengthy processes - typically, ten to 15 years - whereas the construction of hyperscale facilities operates on a much shorter timeline. In an ambitious statement, the UK government hopes ‘to connect the SMR projects to the grid in the mid-2030s.’

Hyperscale facilities are designed to handle enormous amounts of data processing, storage, and computing power. In our experience, they usually have a three-to-five-year planning to completion cycle. By the time the first SMR is scheduled for completion in the UK, the data centre’s tech stack may be more than a decade old, making it potentially less aligned with future requirements.

The next point to consider is data centre consumption, which is based on dynamically fluctuating loads. This means it can transform from being idle to spiking upwards at any point: they are designed to scale up or down rapidly in order to meet fluctuating demand.

By contrast, the SMRs provide a consistent and constant output. Because their capacity to supply these types of facilities has not yet been demonstrated, it is therefore advisable that the load capabilities of SMRs are investigated during the early stages of design. Notably, nuclear energy may complement existing backup systems, such as diesel gensets, UPS systems, and battery backups.

Cost and return on investment (ROI) are always critical challenges in the evaluation of any large-scale infrastructure project: determined by the profitability of an investment proportionate to the total costs involved. The construction of data centres is predicated on a rapid ROI model, whereas nuclear energy is based on long-term, high capital expenditure. Without long-term power purchase agreements (PPAs) or state-backed incentives, it is challenging to synchronise the respective investment cycles of nuclear developers and hyperscale clients because they are so different.

Public perception remains an ongoing topic of consideration in the nuclear industry. History demonstrates that nuclear power has generated public scepticism over time: fuelled by concerns over safety, due to potential risks during operation and the management of the waste, and potential links to nuclear weapons proliferation.

Manifestly, there will be challenges in maintaining public confidence when adopting nuclear power as a source of energy for datacentres, which can often be contentious within the communities where they are built. For example, public demonstrations have recently occurred in the Netherlands, involving local farmers and environmental groups, over proposed government net zero policies aimed at reducing nitrogen emissions and expanding nuclear energy. 

Location is another key issue. Any decision to locate data centres adjacent to small nuclear reactors introduces important considerations and, since the UK lacks clear rules for direct (behind-the-meter) connections, several outstanding details to be clarified: who supplies the power, how are costs set, and who holds liability? 

Having sufficient space to guarantee safety is a crucial consideration. SMRs need to create safety zones, cooling systems, and security buffers: the space that is required to fulfil these objectives makes it more complex to integrate close to dense urban or fibre-rich sites where data centres usually operate, such as edge data centres.

As outlined above, waste management is a central focus in relation to nuclear projects. 

All reactors, including the planned SMRs, produce nuclear waste. Long-term storage and handling (even if it is reduced) will require national policy alignment and appropriate facility agreements to be implemented. 

Elsewhere, alternatives to nuclear power for use on datacentres are being investigated. The Netherlands is looking at the potential adoption of hydrogen energy, because its combustion produces only water, without any direct carbon emissions. 

As a key part of its transition to sustainable energy, the Netherlands is investing heavily in hydrogen energy, particularly green hydrogen. This would enable excess power from local wind turbines to be used in order to convert water to its component elements - hydrogen and oxygen. Because the Netherlands has an abundance of both wind and water, it could become a major competitor in the field of alternative energy sources, particularly for datacentres. 

Although there are logistical and technical considerations for implementing these green benefits, the use of hydrogen energy clearly has significant potential to reduce greenhouse gas emissions. The Dutch government’s vision also aligns with many of our tech clients’ goals to become entirely carbon neutral, or even negative, within the next decade or two. Equally, any scientific advances in energy output that could potentially mitigate nitrogen oxides emissions on datacentres will also help the industry to comply with EU legislation and regulations.

Although it requires thoughtful planning relating to waste management, and the substantial cost and long-term planning required for construction, nuclear power is also a low-carbon energy source: it does not produce greenhouse gas emissions during electricity generation. 

Even though key advantage of nuclear is that it only requires a small amount of fuel to provide power for a significant duration: as a fuel source, it could potentially outlive the lifespan of server rack inside a datacentre to which it provides energy. 

SMRs will be much smaller than conventional nuclear reactors. Potentially, they can be built in factories. As a faster and more cost-effective way to deploy nuclear power, their modular, factory-based construction is a key advantage. Occupying a smaller footprint than traditional nuclear plants, they can be modularly scaled, making them more adaptable to phased datacentre expansion plans. Co-located SMRs will also reduce reliance on national grids, improving resilience against outages, grid congestion, or power price volatility.

Although cheaper and quicker to deploy than large nuclear power plants, multiple SMR designs exist. 68 are currently listed by the International Atomic Energy Agency (IAEA). Ultimately, their individual success will depend upon demonstrating that they are fit for purpose.