BRENNEN

Industries / Energy, water and waste / Power generation / Small modular reactors

Cost

This work estimates and tracks what a small reactor costs to build and run, and why the figures change between the first unit and later ones.

5

Milestones achieved, from 1986 to 2026.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1986

    Bigger plants, no savings

    Engineering models predicted that larger reactors would cost less per kilowatt. A 1986 Energy Information Administration study of 67 US plants found that a 1% increase in size cut real cost per kilowatt by about 0.6% if build time stayed the same, but each 1% of size added about 0.7% to build time, which raised costs by about 0.85%. Once build time was counted, the saving from size disappeared.

    Lessonsavings from size can be lost to longer, harder construction, the argument small reactors rest on.

    osti.gov

    Credit

    Energy Information Administration, US Department of Energy

  2. 2009

    Learning runs backwards in France

    France built its fleet with central decisions, a high degree of standardization and stable rules, the conditions expected to make each reactor cheaper than the last. Arnulf Grubler, working from public records, found that construction cost per kW rose more than threefold between the first and last reactor generations, while operating costs stayed flat; he called it negative learning. Others read the record differently: INL's 2024 review, citing the Nuclear Energy Institute, reports that South Korea's construction costs fell 63% from its first reactors over 1995 to 2011.

    Lessonbuilding more of a design does not lower costs by itself if each version grows larger and more complex.

    pure.iiasa.ac.atgain.inl.gov

    Credit

    Arnulf Grubler, IIASA (interim report 2009; Energy Policy, 2010)

  3. 2023

    A first SMR priced out

    The Carbon Free Power Project planned six 77 MWe NuScale modules at Idaho National Laboratory for Utah Associated Municipal Power Systems (UAMPS), a group of public power utilities. In January 2023 its target power price rose to $89 per MWh, from $58 by Utility Dive's account or $55 by the National Rural Electric Cooperative Association's, and IEEFA put the construction cost estimate at $9.3 billion, up from $5.3 billion. With too few utilities signed up to buy the output, NuScale and UAMPS ended the project on 8 November 2023.

    Lessona first plant's price can move far before construction, and buyers who signed up at the old price may leave.

    utilitydive.comcooperative.com

    Credit

    NuScale Power; UAMPS

  4. 2024

    Estimates pooled into ranges

    With no SMR yet built in the United States, Idaho National Laboratory, with Argonne, MIT and the Breakthrough Institute, pooled published cost estimates into ranges for energy planners. For a 300 MWe SMR built after the first demonstration, it put overnight cost (the cost to build, leaving out financing) at $5,500, $8,000 or $10,000 per kWe in 2022 dollars in its advanced, moderate and conservative cases, against $5,250 to $7,750 for a large reactor. It could use 35 of the 50 estimates it gathered and notes that "most of the data used in this report are estimates and not observed cost data."

    gain.inl.gov

    Credit

    Abdalla Abou-Jaoude and colleagues, Idaho National Laboratory

  5. 2026

    Darlington's first budget breaks

    In May 2025 Ontario approved OPG's C$20.9 billion budget for four BWRX-300s at Darlington, a figure that includes interest, inflation and contingency: C$6.1 billion for the first unit, C$1.6 billion for systems shared by all four, and falling to C$4.1 billion for the fourth. On 11 November 2026 OPG said Unit 1 would no longer meet its October 2030 date or the C$7.7 billion for the first unit and shared systems, citing GE Vernova Hitachi's unfinished engineering for a first-of-a-kind design. It gave no new figure.

    Lessona budget set while engineering is still under way moves when the engineering does.

    world-nuclear-news.orgworld-nuclear-news.org

    Credit

    Ontario Power Generation (disclosure); World Nuclear News (record)

Watch for

  1. Later units a third cheaper

    OPG plans for each Darlington unit after the first to cost less, from C$6.1 billion for the first to about C$5 billion and then about C$4 billion for the fourth, by applying lessons from the first, as on its refurbishment of Darlington's four older reactors, where it cut a year from the schedule between the first and last unit. After its November 2026 warning on Unit 1, OPG expects to "substantially complete" a reassessment of cost and schedule by mid-2027.

    Achieved whenthe Ontario Energy Board reviews actual costs for Units 1 and 2 and Unit 2 costs less than Unit 1.

    ans.orgworld-nuclear-news.org

    Claimed by

    Ontario Power Generation, June 2026 (unit costs) and November 2026 (reassessment)

Open

  1. 01

    Small reactors, estimate versus outcome

    An open ledger of every reactor under 300 MWe built or under construction worldwide, with its cost estimate at the start of construction and its final or latest cost, each linked to a source and converted to constant dollars.

    The brief
    Credit · unclaimed
  2. 02

    Learning rates inside price plans

    A public table of every small reactor plan that gives prices for a first unit and later ones, with the learning rate each plan implies (the percent cost drop each time the number of units built doubles), set beside measured rates from past reactor programs.

    The brief
    Credit · unclaimed

Open 01 · the brief

Small reactors, estimate versus outcome

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open ledger of every reactor under 300 MWe built or under construction worldwide, with its cost estimate at the start of construction and its final or latest cost, each linked to a source and converted to constant dollars.

Why it is still open

INL's 2024 review says that "most of the data used in this report are estimates and not observed cost data." Boston University's 2025 study of 662 energy projects puts the average nuclear cost overrun at 102.5% and finds more escalation above 1,561 MW, but its summary gives no figure for small reactors.

gain.inl.govbu.edu

Built with what exists

The IAEA's Power Reactor Information System for each reactor's size and dates; OSTI for old government reports; regulator and utility filings, such as the Ontario Energy Board's; company annual reports; an AI model that reads scanned PDFs, each figure checked by a person (see the Note "Seven more flagged"); a GDP price index; Python.

  1. First step

    List the reactors under 300 MWe in the IAEA database, pick ten with a public estimate at construction start, and find their final costs.

  2. Done looks like

    At least 30 reactors, each with a sourced estimate and outcome in constant dollars and the overrun computed, a second person rechecking 20% of rows, and a comparison with Boston University's nuclear-wide figure.

Open 02 · the brief

Learning rates inside price plans

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public table of every small reactor plan that gives prices for a first unit and later ones, with the learning rate each plan implies (the percent cost drop each time the number of units built doubles), set beside measured rates from past reactor programs.

Why it is still open

INL's 2024 review says that for SMRs "no observed data could be relied upon" and lists rates from 3% to 16% in the literature. OPG's Darlington plan runs from C$6.1 billion for the first unit to C$4.1 billion for the fourth, which by this file's arithmetic implies a drop of about 18% each time the number of units doubles, above the top of INL's range.

gain.inl.govworld-nuclear-news.org

Built with what exists

The estimate data set behind INL's review; OPG and Ontario Energy Board filings; owner and vendor filings for other projects; Python, with the code published.

  1. First step

    Compute Darlington's implied rate and those of the first-of-a-kind and later-unit pairs in INL's data set.

  2. Done looks like

    At least ten plans with sourced first-unit and later-unit prices, implied rates computed, compared with France's rising costs and South Korea's falling ones, and updated when Darlington Unit 2's actual cost is reported.