Nuclear Construction Costs: Overnight Cost, Financing and SMR Scale

60-second answer: Most nuclear cost claims that look contradictory are measuring different things. Overnight cost prices the plant as if built instantly, with no interest; total construction cost adds financing and escalation over the build; LCOE spreads everything over lifetime electricity at a chosen discount rate (WNA, economics of nuclear power). The gaps between them are large: $1 billion financed entirely for five years at 8% grows to about $1.47 billion before any overrun, and the IEA/NEA's standard study shows nuclear's LCOE roughly doubling between a 3% and 10% discount rate (Projected Costs of Generating Electricity 2020). So the checklist below normalizes any cost claim (currency, price year, scope, capacity basis, financing) before comparison, and applies the same discipline to SMR estimates, where almost every number is still a target rather than an outcome.

Fuel is the cost nuclear operators barely notice; capital is the cost that decides whether plants get built. For uranium investors the construction bill matters because it gates the demand: announced reactors become fuel buyers only if someone finances the concrete.

Normalize before comparing

Five attributes, checked in order, resolve most "conflicting" cost figures:

1. Currency and price year. A 2015-dollar estimate and a 2026-dollar estimate differ by years of inflation before any real difference appears.
2. Scope. Does the figure include owner's costs (land, transmission, project development), contingency, and first fuel? "EPC cost" and "all-in cost" can differ by a third without either being wrong.
3. Capacity basis. Dollars per kW of net or gross output; for cogeneration designs, electrical or thermal capacity (why MWe vs MWth matters).
4. Financing inclusion. Overnight or financed. A frequent apples-to-oranges error: comparing one project's overnight estimate with another's financed actual.
5. Project stage. A published target, a contracted price, or an as-built outcome. These are different evidence classes, and only the last one is a fact about construction.

Overnight cost versus the cost of time

Overnight cost strips out time so that technologies and designs can be compared on engineering content. Real projects then add the cost of time back in, and for nuclear the addition is large because builds are long and capital-heavy. The mechanics in one illustration: finance $1 billion for five years at 8% and compounding alone takes it to $1.469 billion, a 47% premium with zero overrun. Real projects draw money in stages rather than all at once, so the effective premium is smaller than the worst case, but the sensitivity is the point: every year of delay compounds the financed total, which is why schedule slip, not concrete, sank the economics of several Western builds. The same arithmetic runs in reverse for anything that shortens builds, which is the economic argument for factory fabrication.

LCOE: useful, and easy to misuse

Levelized cost divides lifetime discounted costs by lifetime discounted electricity. It is the standard way to compare generation technologies, and its output is dominated by an input that has nothing to do with engineering: the discount rate. The IEA/NEA study is explicit about this, publishing nuclear LCOE across 3%, 7% and 10% rates precisely because capital-intensive, long-lived plants swing hardest with financing assumptions (IEA/NEA 2020). Three cautions carry over from its method: its figures are study assumptions with a price year, not current bids; LCOE excludes system effects (grid value, firmness) by construction; and a single-number LCOE claim without its discount rate is unusable for comparison. Cited properly, LCOE explains why identical physical plants can have different electricity costs in different financing environments, which is the honest version of most "nuclear is cheap/expensive" arguments.

FOAK, NOAK and the factory argument

SMR economics rest on a trade. Smaller units give up scale economies: a 300 MW plant does not cost 30% of a 1,000 MW plant, because containment, licensing, staffing and sitework do not shrink linearly. In exchange, smaller units promise factory repetition: standardized modules, learning across units, shorter site schedules and cheaper capital because less money is at risk for less time. Whether repetition beats scale is an empirical question that first-of-a-kind (FOAK) projects cannot answer, since FOAK carries the one-time costs of tooling, qualification and design completion. Nth-of-a-kind (NOAK) estimates are therefore targets with a learning assumption inside them, and the honest way to read any SMR cost claim is to ask which unit number it describes and what learning rate it assumes. As of 2026 no Western SMR has an as-built commercial cost; every figure in circulation is a target, a contract, or a demonstration budget, and our project evidence checklist treats cost claims accordingly.

Reading a cost claim: the worksheet

For any project number, record: source and date; currency and price year; nominal or real; net or gross capacity; scope (EPC, owner's costs, contingency, first core in or out); financing included or overnight; and stage (target, contract, outcome). Two figures agree or disagree only after all eight fields match. This is the same source-dating discipline the site applies to fuel-cycle capacity claims, and it is why this guide quotes mechanisms and sensitivities rather than a table of headline $/kW figures that would mix eight-field combinations silently.

What construction costs mean for uranium

The fuel connection runs through commissioning dates. Capital cost and financing decide which announced projects reach construction; construction schedules decide when first cores are ordered, and first cores are bought before commercial operation; and the fuel bill itself is small next to capital, which is why fuel-price volatility never stops a build the way financing does (the fuel-cost breakdown). A demand model that counts announced gigawatts without a financing filter overstates the pipeline; the reactor tracker records project stage for exactly this reason, and corporate-level funding evidence sits in the fission funding tracker.

Frequently asked questions

What is nuclear overnight cost? The capital cost of a plant priced as if it were built overnight: engineering, procurement and construction with no financing or construction-period escalation. It isolates the engineering content of a design so different technologies can be compared before financing assumptions take over.

Why do nuclear cost estimates vary so much? Mostly because they measure different things: different price years, scopes (with or without owner's costs and contingency), net vs gross capacity, overnight vs financed totals, and targets vs as-built outcomes. Normalize those five attributes and much of the spread disappears; what remains is real project performance.

How much does financing add to nuclear construction cost? It scales with build time and the cost of capital. As a bounding illustration, $1 billion financed entirely for five years at 8% compounds to about $1.47 billion. Staged spending softens this, but long delays at high rates can add more to a project's cost than any physical overrun.

Are SMRs cheaper than large reactors? Per kilowatt, not automatically: small plants give up scale economies and must win them back through factory repetition and shorter schedules. That trade is unproven at commercial scale as of 2026; current SMR figures are targets or contracts, not as-built outcomes, and first-of-a-kind units carry one-time costs that targets for later units assume away.

Does the uranium price affect whether reactors get built? Barely. Fuel is a small share of nuclear's lifetime cost, so build decisions ride on capital cost, financing terms and schedule confidence. Uranium demand follows construction, not the other way around.

This article is for informational purposes only, not investment advice.