The Small Modular Opportunity

Since 1956, Britain’s nuclear plants have been large-scale projects requiring extensive civil works. To get Hinkley’s workforce in each day, EDF were forced to build the second largest bus station in Europe. Hinkley Point C will use 1.8 million cubic metres of ‘nuclear grade; concrete and 22,000 tonnes of steel. In return for all of this, Britain will get a nuclear plant capable of generating enough energy to power 6 million homes.

Yet large projects – megaprojects – are expensive and often can only be financed with large guarantees from the state. Given the large risk involved, it is understandable that there has never been a private sector large-scale nuclear deployment. It is also understandable that HM Treasury is reluctant to commit to a fleet of projects when cost overruns can run into the tens of billions. Hinkley Point C will cost more than seven times the annual DESNZ budget to build. In short, there are major barriers to Britain adopting the fleet model for large-scale nuclear.

Small Modular Reactors (SMRs) could revolutionise nuclear. Built in a factory and assembled on-site, SMRs have the potential to unlock the fleet approach in Britain. Unlike large-scale reactors, SMRs can be built on a plot of land no larger than a football pitch. Each unit produces significantly less power than a Hinkley reactor (e.g. Rolls-Royce SMR’s - 470MW), and in some cases, a lot less (Last Energy PWR - 20MW).

The attraction of SMRs is their size, scale and repeatability. Unlike Hinkley-style megaprojects, SMRs are much better suited to the fleet approach. Cost overruns on a single project are smaller and easier to manage. If fleets are ordered, then the gains from production learning and the ability to spread large capital investments over multiple projects could be more than large enough to offset the economies of scale lost from building smaller plants.

A key advantage of SMR’s lower-per-unit costs is that they open up new opportunities for finance. Unlike large-scale projects, SMRs could be privately financed once proven. This could be via a Contract for Difference as is the case with solar and wind, or via a power purchase agreement (PPA) where the SMR developers sell power directly to an industrial customer. Most of the demand from industry so far has come from tech companies. For instance, Microsoft, Google, and Meta are all pursuing deals with SMR and Micro-SMR developers seeking reliable clean baseload to power their data centres. However, there is scope for a much wider range of industrial customers. Not only can SMRs provide reliable power for electric arc furnaces to produce, some Advanced Modular Reactors can also produce low-carbon heat for hard-to-decarbonise industrial processes such as chemical processing.

Britain could lead the world in SMRs. Rolls-Royce have been designing and making small modular reactors for nuclear submarines for decades. Not only are Rolls-Royce SMR in negotiations with the British Government over a fleet order as part of the Great British Nuclear competition, they are also in discussion with a number of European governments including Czechia’s over sales. A silver lining to having the highest industrial energy prices in the world is that for SMR developers Britain is an attractive and potentially profitable market for their first deployment.

AI needs nuclear (and will pay for it)

AI is the defining technology of the 21st century. Progress is rapid. In 2022, AI couldn’t talk like a real person, see and describe images, or work with long documents or code. Today, it can. Senior politicians such as Technology Secretary Peter Kyle MP now talk of reaching Artificial General Intelligence by the end of this Parliament. If Britain is to harness the full potential of AI, then we will need access to compute. In other words, Britain will need to build more data-centres and the infrastructure to power them.

The Government’s AI Opportunities Action Plan – authored by entrepreneur Matt Clifford – set out a vision of the UK as an ‘AI maker, not a taker’. Britain has enormous strengths in AI. London hosts Google DeepMind (founded by Nobel Prize winning scientist Demis Hassabis) andOpenAI, Anthropic, Microsoft and Meta AI all have major offices in the capital. Britain is home to some of the world’s leading universities – magnets for top global engineering and science talent. Yet, Britain is falling behind in one key area: compute.

The Department for Science, Innovation, and Technology’s (DSIT) Compute Roadmap set a target of Britain having six gigawatts (GW) of AI-ready data centre capacity by 2030 – more than triple Britain’s current capacity. This will require over 50% more electricity as the higher end of Britain’s National Energy Systems Operator higher range forecast from 2022.⁷⁵ By 2035, DSIT predict that total energy demand from AI data centres could nearly double again to 11 GW.

To meet this target, Britain will need to overcome two key barriers.

• Britain has the world’s highest industrial electricity prices, four times those in the US.
• Britain’s grid lacks capacity. Projects can face waits of over a decade to gain a connection to the National Grid.⁷⁶

Recent reforms to the Grid’s connections queue should cut lengthy waits, yet almost all additional capacity has been allocated to connecting solar and wind farms to deliver Clean Power by 2030. If data centres are able to connect to the grid, they are hit with extremely high electricity prices. A single 75MW AI data centre could face an annual electricity bill of £120m - four times the cost in the state of Virginia.⁷⁷

In some cases, tech giants will be willing to bear the cost in order to serve lucrative markets such as the M4 Corridor, yet high industrial electricity prices are a major deterrent to the scale of investment being made in the US and China,

One option, being pursued in the US and elsewhere, is for AI businesses to cut out the middle man and strike Power Purchase Agreements (PPAs) directly with generators. In some cases, businesses are working around grid constraints to build their own private transmission lines.

Amazon, Microsoft, and Meta are the largest corporate PPA buyers. Meta has signed two long-term PPAs totaling 650 MW of solar in Texas and Kansas with AES to power its U.S. data centers.⁷⁸ Microsoft is providing a customer for Constellation Energy who are planning to invest $1.6 billion to restart Unit 1 of Three Mile Island Site.⁷⁹ Amazon is now connected via a private wire to Susquehanna Steam Electric Station, a nuclear power station in Pennsylvania.⁸⁰ The deal gives Amazon’s data centre first-dibs on any power generated from the reactor with excess generation sold to the grid.

In theory, PPAs with private wires connected to new clean generation are a logical option for data centres in Britain. Meta, Google, and OpenAI could avoid a ten-plus year wait in the grid connection queue and lock in prices below Britain’s ultra-high industrial electricity price. Last Energy has already signed in-principle deals to build 24 micro-SMRs in the UK with four large industrial customers.⁸¹

Yet this model faces major barriers in the UK:

• Access to the grid is still essential for stability, to export excess power, and as backup when nuclear plants are being refuelled.
• SMR builders want certainty and long-term contracts, but there’s a risk AI or industrial customers will not be around in 10 years.
• Most tech majors have ambitious Net Zero goals. Yet nuclear energy is not covered by Renewable Energy Guarantee of Origin (REGO) certificates and access to green finance for such projects is limited.

What needs to change

Create alternative routes to market for commercial nuclear deployment

Up until today nuclear projects have been so large and expensive that they needed state-backed financing and guarantees. Small Modular Reactors (SMRs) and Micro-SMRs match the scale of other privately-financed energy projects. Britain has been able to deploy billions of pounds worth of grid-scale solar and wind using Contracts for Difference (CfD), which guarantee a fixed power price. When there’s a glut of power and prices are low, the CfD tops up the generator. When they’re high, the generator only receives the pre-agreed price.

CfDs have clear advantages: they are simple, they protect consumers from construction overruns, and provide certainty to investors. CfDs are well-suited to nuclear projects that must run near constantly to recoup investment costs and cannot easily ramp generation up and down like gas. In fact, CfDs are arguably better suited to nuclear than renewables because they lack the market cannibalisation problem where wind and solar produce at the same time causing prices to crash and requiring large top ups.

Yet nuclear is the only low-carbon energy source to be excluded from CfD auctions. This was initially due to scale – one single nuclear project could produce more power than the entire auction budget pot. This no longer applies as Micro-SMR and SMR technology develops. Pot 2 CfD auctions are open to less mature innovative technologies such as floating offshore wind and tidal stream. Projects range from 12-400 MW and are estimated to cost between £185-£228 per MWh (in 2025 prices). SMR and micro-SMR nuclear projects are of similar scale and could offer lower prices. For context, a 100 MW SMR at a £150/MWh strike would have an indicative CfD budget impact of £52.5 million a year – less than half of the AR 6 pot for emerging tech.⁸² This could make the UK the best place in the world to build for SMR startups looking to build their first deployment. The Department for Energy Security and Net Zero should open future Pot 2 CfD auctions to SMR and micro-SMR projects.

Advanced Market Commitments are binding pledges to buy new technologies before they are developed. This was used famously to procure large quantities of the coronavirus vaccine even before it was approved for use. It is also being used to stimulate the demand for carbon removal technologies. This approach should be applied to nuclear to stimulate investment in micro-SMRs.

The Department for Energy Security and Net Zero should offer a bilateral (e.g. non-auction) CfD for innovative new nuclear (Micro-SMRs and AMRs) deployed in the next five to ten years. The model should reward faster deployments with 5GW available before 2030 and 10GW by 2035. If a project is delivered before 2030, receive a generous strike price, while one in 2035 should get a lower strike price. This would be a powerful signal to target the UK for first deployments.

Power-Purchase Agreements (PPAs) are deals between power generators and heavy-power users such as data centres. In the US, corporate buyers have contracted about 75 GW of renewable capacity via PPAs to date, and deals are increasingly being struck between tech companies and nuclear plants.⁸³ A new Industry Growth CfD would support this model in the UK. It would sit beneath PPAs as a backup and offer a guaranteed power price to SMRs operators if their industrial customer goes bust or sharply cuts demand. The Department for Energy Security and Net Zero should create a new Industry Growth CfD to reduce counter-party risk for PPAs. The fallback CfD should be pegged to a forward rolling average of the wholesale price at the moment the PPA falls down.

Many large corporations have set ambitious science-based targets for emissions reductions. In some cases, the targets are substantially more ambitious than the UK's Net Zero by 2050 target. However, nuclear is not benefitting from this wave of green private investment because Renewable Energy Guarantees of Origin (REGO) certificates exclude nuclear and do not reflect the benefits of always-on nuclear power. For example, REGOs ignore periods when wind and solar lulls mean industrial customers are likely running on fossil fuels. The Government should update REGOs to explicitly include nuclear power and use real-world 30 minute data on fuel mixes to make SMR PPAs attractive for businesses seeking to cut their emissions.

Projects with a strong decarbonisation element can attract specialist green finance including green gilts and NS&I Green Savings Bonds. Green investors factor in a project’s environmental benefits and accept lower returns. This, in turn, makes financing cheaper for clean energy projects. However, nuclear is currently excluded from the UK’s Green Financing Framework despite having the lowest life-cycle carbon emissions of any source of energy and playing a vital role in our future energy mix. The Government should update the Green Financing Framework to include investment in civil nuclear power.

Update grid regulations to unlock investment in co-located nuclear power for industry

Without massive investment in compute, Britain will fall behind in AI. DSIT’s ambitious target of 6GW of AI-capable data centre capacity on the grid by the end of the decade risks being missed unless two key barriers can be addressed: ultra-high industrial electricity costs and a lack of access to the grid. In response to the AI Opportunities Action Plan, the Government committed to create AI Growth Zones. Data centres (and supporting infrastructure) built within AI Growth Zones receive fast-track planning and easier access to power.

Even if tech companies secure behind-the-meter access to new clean power, data centres will still need access to the grid for stability and balancing. Usually this would involve building a new Grid Supply Point to step-down power from the high-voltage grid to local networks. However, building a new Grid Supply Point is a complex and lengthy process because they’re typically designed to serve a whole region, not a single cluster. This entire process can take up to ten years. In short, it is unfit for a world where we need to more than triple our AI data centre power demand by 2030.

One alternative would be to allow private companies to fund and build their own direct connections to the grid within AI Growth Zones. As the infrastructure only needs to serve a single power cluster and not a wider region, it could be delivered faster, three years not ten, and for substantially less cash.

The problem is the Electricity Act 1989 bans anyone from transmitting electricity without a licence, leaving National Grid the monopoly builder of 275 kV / 400 kV substations. There are exemptions, but they only apply to cables used to take offshore wind power to land. To unlock tech investment into co-located nuclear, the Government should create a new exemption to the Electricity Act 1989 to allow privately owned and operated substations to connect to the national transmission system to serve co-located clean power generation with AI data centre demand within AI Growth Zones.

Billions of pounds worth of investment in new data centres is being held up by years long waits for grid connections. To solve this, NESO should allow data centres (and other heavy industrial users) to buy a ‘non-firm’ grid connection when they have reliable on-site generation.⁸⁴ SMRs provide reliable clean power and are offline only rarely for planned maintenance, so access to the wider grid is there mainly for system support and scheduled downtime.

This arrangement should be governed by clear and reliable rules. Access to the grid should be there 95% of the time and data centres should receive forecasts of when cutbacks are likely to be needed. This would allow projects to go live without waiting years for a full, always-on transmission connection.