Where things stand
The one active Western SMR construction, Ontario Power Generation’s BWRX-300 programme at Darlington, is funded from OPG’s own balance sheet with costs recovered through regulated rates, which is utility finance rather than project finance.
The UK’s only recent nuclear financial close, Sizewell C in November 2025, shows what it took for a large plant: the Regulated Asset Base model, the UK Government holding 44.9 per cent of the equity, National Wealth Fund debt, and a £5 billion commercial bank facility made possible by export credit cover from Bpifrance Assurance Export. At Wylfa, Great British Energy – Nuclear’s contract with Rolls-Royce SMR, signed on 13 April 2026, takes the programme to a final investment decision expected in 2029. No revenue model for the Wylfa units has yet been announced.
Why the usual route is closed
Conventional infrastructure becomes bankable because construction risk is transferred to a contractor under a fixed-price turnkey wrap and lenders lend against that wrap. Nuclear has largely lost that option. The IAEA notes that a lump-sum turnkey contract is the most expensive form available because the contractor must price the risk it absorbs, and in practice tier-one contractors decline to absorb it. In September 2026 TerraPower announced it would rebid the EPC contract for its roughly $4 billion Natrium project after it and Bechtel decided not to proceed together into full EPC.
Lenders also cannot yet price the cost curve. Published nuclear learning rates range from around 2 to 15 per cent, negative learning has been observed, and the OECD Nuclear Energy Agency has warned that most published SMR cost estimates are implicitly nth-of-a-kind figures. Until real units have been built at a known cost, a credit committee has nothing to calibrate against.
So the route to bankability cannot be a single transaction at FID. It has to be built in stages, with each stage financed by the kind of capital able to carry the risk that remains, and each stage designed to produce the evidence the next kind of capital needs.
Stage one: fund the technology risk with development capital
Before a project exists, the risks are technical and regulatory: will the design work, will the regulator accept it, and what will it really cost? Commercial debt has no place here. The right capital is equity and quasi-equity from investors who expect to carry technology risk: venture and strategic investors, state development banks, and increasingly public markets.
Finland’s Steady Energy is a clear current example of this stage done deliberately. Spun out of the VTT Technical Research Centre in 2023, it is developing the LDR-50, a 50 MW heat-only reactor operating at around 150°C and low pressure for district heating. In September 2026 the European Investment Bank committed up to €40 million as a senior unsecured convertible loan, its first investment in SMR technology, explicitly to fund research, testing and licensing between 2026 and 2028 and excluding any installation. Alongside it, the company combined with 3North Partners and listed on Nasdaq First North Finland on 2 October 2026, with around €69.8 million of committed investment from Finnish pension insurers, Fortum, Orlen VC and the state investor Tesi.
The point is the match between capital and risk. Patient, convertible and equity money is paying for design, a non-nuclear pilot facility in Helsinki and regulatory review, while no lender is being asked to take construction risk on a design that has not yet been licensed. The UK equivalent is the National Wealth Fund’s facility of up to £599 million to Rolls-Royce SMR for its generic design.
What this stage must produce is the evidence base for the next one: regulatory acceptance of the design, such as a Design Acceptance Confirmation from Generic Design Assessment; a frozen design; a bottom-up cost estimate; and a first site.
Stage two: fix the revenue before anything else
Everything a lender assesses follows from how the project gets paid, so the revenue model has to be settled before FID rather than assumed. Rolls-Royce SMR has told Parliament that financing its projects requires a long-term government agreement, through either a Contract for Difference or the RAB model.
The two distribute risk very differently. A Contract for Difference pays only once the plant generates, so all construction risk sits with the developer and its financiers until first power. The RAB model allows a licensed company to earn a regulated return during construction, which lowers financing cost by sharing overrun risk with consumers and the state. For a fleet the design questions are specific: whether each unit is its own regulated asset or the programme is one, how allowed costs are set for later units, and how learning is passed back to bill-payers.
The buyer matters as much as the mechanism. A sixty-year asset should be contracted to the most creditworthy counterparty available: the state, a regulated network, a large utility or a technology company with an investment-grade balance sheet. Steady Energy’s market illustrates a different answer. Its customers are municipal and utility district heating companies in European cities of around 50,000 people or more, buyers with long-established heat businesses, and it offers either a turnkey plant or a heat-as-a-service model in which Steady Energy owns and operates the reactor and retains the licensing, operations and waste responsibilities, so the customer buys heat without taking on a nuclear liability. WNN reports agreements covering 15 reactors in Finland; these are not yet construction contracts, and construction of the first plant is targeted for 2029.
Stage two also has to turn the pipeline into commitments. A credit committee will give weight to a binding multi-unit order, to capped pricing for long-lead components across several units, and to a defined decommissioning and waste funding arrangement accepted by the regulator. Memoranda of understanding do not count.
Stage three: put a ceiling on first-of-a-kind construction risk
If construction risk cannot be wrapped, it has to be allocated. Lenders will finance a first project only where the downside has a ceiling and every overrun scenario has a named owner with the money to meet it. The tools are well known: overrun sharing through a regulated model; a government backstop or cost-overrun facility above a defined threshold; sponsor completion support; contingent equity sized against an independent estimate of plausible overruns; and export credit cover for exported content, which is what made Sizewell C’s commercial bank tranche possible.
On a first project, a large share of the capital will be public or quasi-public: state equity, development bank debt and export credit. That should be planned for openly. The objective is to confine it to the risk that genuinely cannot yet be priced, and to set out at the start how it steps back.
Stage four: prove the plant, then refinance
The single largest change in a nuclear project’s risk profile happens at commercial operation. Much of the construction risk falls away and what remains is an operating asset with long-term contracted revenue, which is exactly what pension funds, insurers and infrastructure investors want to own. Refinancing an operating first unit into infrastructure capital at infrastructure rates releases the state and development money that carried the construction risk, so it can be recycled into the next units.
That needs designing in from the start. The revenue contract, ownership structure and debt terms on unit one should be built to be refinanced, with no change-of-control or prepayment terms that make it expensive for long-term investors to step in.
Stage five: earn the learning credit
Lenders will price fleet learning once they can see it. The evidence that moves them is actual outturn: cost and schedule achieved on units one, two and three against the estimates made at FID, under a stable design and a single regulatory acceptance. The NEA’s work suggests most designs need five to seven units before supply chain learning is realised. Once that evidence exists, later units can be financed with a smaller public share and more conventional project finance, and the pricing of the fourth or fifth unit becomes the real test of whether the programme has worked.
Six things that shorten the route
Standardise the deal as well as the reactor. Hunton’s analysis points out that when every SMR project is negotiated from scratch with bespoke terms, transaction costs stay high and timelines stretch. A fleet needs template supply, construction and revenue contracts, agreed once with lenders and reused.
Treat design change as a financial event. Every change after FID moves cost and schedule away from the numbers lenders underwrote. Change control is a credit discipline.
Choose a unit size and product that suit the capital available. Smaller units can mean smaller individual cheques; shorter construction and faster accumulation of cost evidence depend on design maturity and repeatable delivery. A simpler product can change the financing problem itself, which is the bet Steady Energy is making with a low-pressure, heat-only design sold to existing heat utilities.
Use each instrument for what it does best. Development banks and quasi-equity for technology and licensing risk; regulated revenue for construction; export credit for exported content; infrastructure capital for operating assets.
Make the evidence public. A programme that reports its actual cost and schedule outturn transparently gives later lenders something to price against, and may lower the cost of capital for subsequent units.
Start the financing work early. The structure that makes a project fundable has to shape procurement, contracts and corporate structure from the outset.
What this means for Wylfa
For the UK’s SMR programme the window between now and the 2029 investment decision is where most of these choices will be made. The revenue mechanism for a fleet, the allocation of overrun risk on the first units, a commitment that runs beyond three units, a refinancing route once the first unit operates, and UK Export Finance support for exported units all need settling well before FID if they are to shape the project rather than be retrofitted to it.
The honest position
The sceptics’ case deserves a hearing. NuScale’s Carbon Free Power Project in Utah was cancelled in November 2023 after its target power price rose from around $55–58 per MWh to about $89 per MWh and it failed to secure enough subscribers. IEEFA argues that SMRs remain too expensive, too slow and too risky. Every SMR project that has moved forward has done so behind a regulated revenue stream, a state-owned utility, government equity, a development bank or an export credit agency.
The staged route does not dispute that. It accepts that the first projects need public capital and sets out how that capital is meant to step back as the evidence builds. Whether it can, by the fourth or fifth unit, is the question the first fleets will answer, and the structuring decisions being taken now will largely determine the answer.
NCL advises developers, investors and public bodies on nuclear investment cases, risk allocation, programme structure and the regulatory pathway. If you are working out how to take an SMR proposition from development to financial close, talk to us.