In partnership with
In Ontario, the first small modular reactor in a G7 country has a price the public can read. In Wales, the first British units have a contract, a loan and no price. In Utah, the buyers of the one American project that published its number walked away. The silence, not the reactor, is the story.
Ontario Power Generation's first GE Vernova Hitachi BWRX-300 at Darlington carries a release-quality budget of C$7.7bn, about £4.2bn ($5.6bn, €4.9bn). That covers the C$6.1bn reactor, about £3.3bn ($4.4bn, €3.9bn). Shared site infrastructure adds C$1.6bn, about £870m ($1.2bn, €1.0bn). The four-unit programme is C$20.9bn, or £11bn ($15bn, €13bn). The figure includes interest, escalation and contingency. A small modular reactor (SMR) is a unit of up to about 300 MW. Among the programmes racing to follow, none has published a committed figure of that kind.
The reactors are not the bottleneck. The numbers are.
The opening is one number. Darlington's first unit costs C$7.7bn, or £4.2bn, with every cost included
The headline is a machine. The substance is a spreadsheet. Ontario Power Generation (OPG) calls its estimate "release quality", meaning engineering, procurement and schedule are mature enough to commit. It includes licensing, engineering, procurement, construction, operations readiness, contingency, interest and escalation. OPG projects a cost to consumers of about 14.9 Canadian cents per kWh over a 60-year operating life. That is about 8.1p (10.8 US cents, 9.5 euro cents). A buyer can test that price against wind, solar and storage. It is also the first unit's price, which is the hardest number in nuclear to reveal.
Only one SMR carries a board-approved price. Every other number is an estimate
Set that against the field. Westinghouse launched the AP300 in 2023 with a target cost per unit of £752m ($1bn, €880m). A target is not a price. Oklo and X-energy have not published a first-unit project cost. The Tennessee Valley Authority's Clinch River unit carries a reported figure of around $5.4bn, or £4.1bn ($5.4bn, €4.8bn). No final investment decision stands behind it. Each is a number of a different kind. Only one is a commitment a board has signed.
Darlington's C$20.9bn counts interest, escalation and contingency. Vendor targets do not
The difference is what a number includes. An "overnight cost" imagines the plant built in a single night, with no interest charged. Vendor targets usually exclude owner's costs, financing and inflation. Darlington's C$20.9bn, or £11bn ($15bn, €13bn), includes all of them, plus a contingency OPG has not itemised. Comparing an all-in price with an overnight target flatters the target by design. Buyers who miss that distinction misprice the risk. The ladder of disclosure runs from all-in to nothing at all. Most programmes sit near the bottom.
You are buying a first-of-a-kind SMR. Which number do you insist on before signing?
NuScale published a price once. The project died at £15,100 per kilowatt
The one American project that published its all-in number is instructive. NuScale's plant for Utah's municipal utilities rose from $5.3bn to $9.3bn in January 2023. That is a rise from about £4.0bn ($5.3bn, €4.7bn) to £7.0bn ($9.3bn, €8.2bn) for 462 MW. The final figure is roughly £15,100 ($20,100, €17,700) per kilowatt. Its target power price climbed from £44 ($58, €51) to £67 ($89, €78) per MWh. Ten months later the partners terminated the project for lack of subscribers. Transparency did not sink it. The number did. Disclosure exposed a cost the market would not carry, which is exactly what disclosure is for.
Rolls-Royce has a contract and a £599m loan. It has no public unit price
Britain shows the other pattern. Rolls-Royce SMR signed a contract in April 2026 for three units at Wylfa. A National Wealth Fund loan facility of up to £599m ($797m, €701m) backs the work. The release names jobs, sites and suppliers. It does not name a unit price. The company also holds a commitment for up to six units in Czechia. A firm contract without a public figure leaves every other buyer guessing what a Rolls-Royce unit costs. That is the gap this piece is about.
Hyperscalers have signed up to 13 GW. The financial terms are not disclosed
Demand is not the problem. The Carnegie Endowment counts announced hyperscaler nuclear agreements at up to 13 GW. About half are power purchase agreements, most of them on existing plants. The rest are new-build partnerships worth about 6.1 GW by the mid-2030s, if SMR developers deliver. Carnegie notes the financial terms of the largest deals were not disclosed. The buyers with the deepest pockets have signed without a public price. That tells vendors the silence is affordable. It is not affordable for the next buyer, who cannot absorb an unknown cost.
The nth-of-a-kind claim: Darlington prices its fourth unit 33% below the first
The industry's answer is the learning curve. OPG prices its fourth unit at C$4.1bn, about £2.2bn ($2.9bn, €2.6bn), 33% below the first unit's C$6.1bn reactor cost. Its refurbishment programme returned its second unit to the grid 169 days ahead of schedule. That is the nth-of-a-kind claim in a form a buyer can audit. The learning has to be captured somewhere. Increasingly that is a reactor twin, the digital model where fleet learning is recorded and proven. Without it, the fourth unit's saving is a promise. With it, the saving is evidence a regulator and a lender can inspect.
Poland, Korea and Tennessee show a programme without a committed price
Three programmes show a build advancing without a number to hold. Poland's first BWRX-300 carried a 2022 estimate of €1.1bn, about £940m ($1.3bn, €1.1bn). No committed price has replaced it. Korea Hydro and Nuclear Power (KHNP) bought into TerraPower in January 2026 with no unit price attached. The US Nuclear Regulatory Commission's staff recommended a construction permit for Clinch River in June 2026. The commissioners still decide. Each is real progress. None yet gives a buyer a figure to test.
The investment implication: Darlington has a number for lenders, regulators and buyers. No rival does
The capital lesson is direct. A reactor's novelty does not set its deployment pace. Its price transparency does. Lenders size debt against a number they can stress. Regulators approve cost recovery against a number they can inspect. Buyers sign offtake against a number they can compare. Darlington has all three, which is why it is under construction. Programmes without a public figure move at the speed of trust, which is slower. The International Atomic Energy Agency (IAEA) General Conference meets in Vienna this week. Its audience is the vendor set that can change that. So was last week's World Nuclear Symposium in London.
The bottom line
One small reactor has a public price. The rest have targets, envelopes or silence. That gap, not reactor design, now decides which programmes get financed and built. Disclosure sank one project and is building another. From Ontario to Wales to Utah, the number is the product. Intelligence can map which programmes have one. The vendor still has to publish it.
Next week: Gas, nuclear and grid all booked out. AI's bottleneck moved from chips to power.
AI made PMs faster. Multiplayer mode is still broken.
A PM can summarize research, draft a PRD, and mock up a prototype before lunch. The hard part starts when the team has to decide what actually gets built.
Jira Product Discovery gives product teams one place to capture insights, prioritize ideas with consistent frameworks, and build living roadmaps stakeholders can rally around.
And because it’s connected to Jira, the context behind every decision stays with the work—so developers and their agents know not just what to build, but why.
AI helps PMs move faster. Jira Product Discovery helps the whole team build with confidence.