Small Modular Reactors: A Real Investment Case, or the Next NuScale?

The investment thesis for small modular reactors is simple enough to fit in a sentence: standardised factory construction strips out the execution risk that has made conventional nuclear effectively uninvestable for thirty years. The thesis is also, on the current evidence base, partially refuted. NuScale's flagship Carbon Free Power Project was cancelled in November 2023 after costs ran past the threshold the consortium of municipal utilities could absorb. France's NUWARD design was scaled back in late 2024 to a simpler reference reactor using established technology. Two of the highest-profile Western SMR programmes have therefore failed to clear their first commercial bar.

That is the context the policy momentum has to be measured against. The 33-country pledge at COP30 in Belém to triple nuclear capacity by 2050 makes the politics easier; it does not make the engineering economics easier. The question for investors is whether the surviving SMR programmes — chiefly Rolls-Royce in the UK and China National Nuclear Corporation's Linglong-1 — are credible enough to underwrite at current valuations.

Why the AI Hyperscalers Are Now in the Story

The constraint that has changed in the last 18 months is demand. Wind plants run at full output for an average of nine hours per day. Solar plants run for six. Nuclear runs for closer to 24, and at the size of facility a hyperscaler wants to colocate with a data centre, no other low-carbon source compares on a duty-cycle basis. Meta has already signed long-dated PPAs with Vistra, TerraPower and Oklo to lock in baseload supply for its AI build-out. Microsoft has gone further, agreeing to restart the Three Mile Island Unit 1 reactor.

What this tells us is that the buyers of first-of-a-kind SMR capacity are no longer regulated utilities subject to public-sector procurement timelines. They are private capital with multi-year compute commitments and balance sheets that can absorb cost overruns the original municipal-utility buyers could not. That is a meaningful upgrade to the financing assumption sitting under the SMR investment thesis.

The UK Programme

In June 2025, the government selected Rolls-Royce over Westinghouse to partner with Great British Energy-Nuclear on the UK's SMR programme, backed by £2.5 billion of state funding and a target of first units operational by the mid-2030s. The choice was partly political — Westinghouse is US-controlled and the government had a clear industrial-strategy preference for a domestic prime — but the engineering case is also defensible. Rolls-Royce has been running reactor production lines for the Royal Navy's nuclear submarine programme for decades; the skill base and supply chain are not being built from scratch.

The export pipeline is where the equity story sits. Rolls-Royce has signed an agreement to deliver the first Czech SMR alongside ČEZ, and a memorandum with Jordan that contemplates pairing SMR generation with desalination for the 4 million residents of Amman. Neither commits to a binding offtake at this stage. But they establish the template — a UK-designed, UK-manufactured reactor sold to a foreign sovereign customer — that has to work for the economics of the wider programme to scale.

Where the Investment Case Fails

There are three identifiable failure modes, and investors should be honest about which ones matter.

The first is cost. Modular construction reduces, but does not eliminate, the unit-economic uncertainty of first-of-a-kind nuclear builds. NuScale's cancellation was driven by exactly this — the levelised cost of energy projection drifted high enough that the offtakers walked. Until Rolls-Royce has poured concrete on a paying customer site, the LCOE numbers underwriting the programme are model outputs, not field data.

The second is waste. A 2022 study led by Stanford estimated that SMRs could generate roughly nine times more neutron-activated steel per unit of electricity than conventional gigawatt-scale reactors, because of the higher surface-area-to-volume ratio in smaller cores. That steel becomes low-level radioactive waste with disposal costs the LCOE models do not always fully internalise. The technology question is not whether this can be managed; it is whether the cost of managing it is in the bid price.

The third is timeline-versus-demand. The AI power problem is acute now. The earliest UK Rolls-Royce SMR comes online in the mid-2030s. Hyperscalers facing a 2027 grid-capacity wall are not going to wait — and the substitute they have already chosen is restarted legacy nuclear, pipeline expansion, and in some cases gas. SMRs may arrive after the demand spike they were supposed to serve.

Our View

We see Rolls-Royce as the best-positioned listed UK pure play on the SMR thesis. The combination of state backing, an active export pipeline and a manufacturing base that is genuinely capable of the build separates it from the more speculative US-listed SMR comparables — Oklo and NuScale — both of which carry materially higher execution risk and trade more on narrative than on contracted revenue.

That said, the SMR story remains a long-duration industrial bet, not an AI play with a 2026 catalyst. Investors sizing positions should be calibrated to a five-to-ten year horizon, and should expect at least one further high-profile programme cancellation somewhere in the Western pipeline before the survivors clear the cost bar. The asymmetry — large upside if modular nuclear works, manageable downside given Rolls-Royce's underlying defence business — favours patient capital. It does not favour anyone treating it as a thematic momentum trade.

Sources

Educational information only. This is not personal financial advice.