ArticlePDF Available

Historical construction costs of global nuclear power reactors


Abstract and Figures

The existing literature on the construction costs of nuclear power reactors has focused almost exclusively on trends in construction costs in only two countries, the United States and France, and during two decades, the 1970s and 1980s. These analyses, Koomey and Hultman (2007); Grubler (2010), and Escobar-Rangel and Lévêque (2015), study only 26% of reactors built globally between 1960 and 2010, providing an incomplete picture of the economic evolution of nuclear power construction. This study curates historical reactor-specific overnight construction cost (OCC) data that broaden the scope of study substantially, covering the full cost history for 349 reactors in the US, France, Canada, West Germany, Japan, India, and South Korea, encompassing 58% of all reactors built globally. We find that trends in costs have varied significantly in magnitude and in structure by era, country, and experience. In contrast to the rapid cost escalation that characterized nuclear construction in the United States, we find evidence of much milder cost escalation in many countries, including absolute cost declines in some countries and specific eras. Our new findings suggest that there is no inherent cost escalation trend associated with nuclear technology.
Content may be subject to copyright.
Historical construction costs of global nuclear power reactors
Jessica R. Lovering
, Arthur Yip
, Ted Nordhaus
The Breakthrough Institute, CA, USA
Department of Engineering and Public Policy, Carnegie Mellon University, PA, USA
Comprehensive analysis of nuclear power construction cost experience.
Coverage for early and recent reactors in seven countries.
International comparisons and re-evaluation of learning.
Cost trends vary by country and era; some experience cost stability or decline.
article info
Article history:
Received 15 September 2015
Received in revised form
12 January 2016
Accepted 13 January 2016
Available online 2 February 2016
Nuclear construction costs
Experience curves
International comparison
The existing literature on the construction costs of nuclear power reactors has focused almost exclusively
on trends in construction costs in only two countries, the United States and France, and during two
decades, the 1970s and 1980s. These analyses, Koomey and Hultman (2007); Grubler (2010), and Esco-
bar-Rangel and Lévêque (2015), study only 26% of reactors built globally between 1960 and 2010, pro-
viding an incomplete picture of the economic evolution of nuclear power construction. This study curates
historical reactor-specic overnight construction cost (OCC) data that broaden the scope of study sub-
stantially, covering the full cost history for 349 reactors in the US, France, Canada, West Germany, Japan,
India, and South Korea, encompassing 58% of all reactors built globally. We nd that trends in costs have
varied signicantly in magnitude and in structure by era, country, and experience. In contrast to the rapid
cost escalation that characterized nuclear construction in the United States, we nd evidence of much
milder cost escalation in many countries, including absolute cost declines in some countries and specic
eras. Our new ndings suggest that there is no inherent cost escalation trend associated with nuclear
&2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND
license (
1. Introduction
Studies by the Intergovernmental Panel on Climate Change and
the International Energy Agency have identied nuclear power as
a key technology in reducing carbon emissions (IPCC, 2014;IEA,
2014). Today, nuclear energy makes up one-third of global low-
carbon electricity, and countries with the lowest carbon intensities
depend heavily on low-carbon sources of baseload power: nuclear
and hydroelectric. Yet the high cost of nuclear power in developed
countries has slowed its deployment, as low-carbon nuclear power
cannot compete with cheaper fossil fuels, especially in deregulated
power markets. Additionally, cost estimates for future nuclear
energy are among the most important inputs to energy system
models and climate mitigation scenarios (Leibowicz et al., 2013;
Bosetti et al., 2015;Barron and McJeon, 2015).
Several analyses of historical nuclear cost trends have pointed
to escalating costs for nuclear power plants over time, raising
doubts about whether nuclear can become cost competitive (Bupp
and Derian, 1978; Hultman et al., 2007;Cooper, 2014). However,
past studies have been limited in their scope, focusing primarily
on cost trends in the 1970s and 1980s for the US (Komanoff, 1981;
Koomey and Hultman, 2007) and France (Grubler, 2010;Escobar-
Rangel and Lévêque, 2015). These studies represent 26% of the
total number of nuclear power reactors completed in the world
and only look at two of the 31 countries that generate electricity
from nuclear power today.
The US and France may not be representative of broad cost
trends, as they suffered rst-mover disadvantages of deploying an
evolving technology (Jamasb, 2007). More importantly, the US and
France built most of their reactors over 30 years ago. The last
Contents lists available at ScienceDirect
journal homepage:
Energy Policy
0301-4215/&2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (
Correspondence to: The Breakthrough Institute, 436 14th Street STE 820, Oak-
land, CA 94612, USA.
E-mail address: (J.R. Lovering).
Energy Policy 91 (2016) 371382
reactor to come online in the US began construction in 1978. The
limited scope of the existing literature on nuclear costs is further
limited by the industry wide disruption caused by the Three Mile
Island accident in 1979. Of the 100 US reactors included in pre-
vious studies, half were under construction and had not yet re-
ceived operating licenses when the accident occurred. Given the
event's potential effect on construction costs, there is a need to
study the cost experience of a wider sample of countries and eras
in nuclear power history.
In addition to the US and France, the UK, Germany, Japan, Ca-
nada, and the USSR were all building nuclear reactors during this
time period. When the US and Western European countries
stopped building nuclear power in the 1990s, several other
countries continued to build out their nuclear eets in East and
South Asia and Eastern Europe. In particular, large eets of stan-
dardized reactors were built in Japan, South Korea, India, and more
recently in China. While a handful of studies note the low cost of
reactors in these regions today (Du and Parsons, 2009;IEA, 2010),
there is little analysis of historic cost trends in these countries.
This study extends and reassesses the literature by collecting
and analyzing cost data from a broader set of countries and time
periods. We focus our analysis on the real Overnight Construction
Cost (OCC) of completed plants because it is both the dominant
component of lifetime costs for nuclear power, and the cost
component that varies most over time and between countries. The
metric OCC includes the costs of the direct engineering, procure-
ment, and construction (EPC) services that the vendors and the
architect-engineer team are contracted to provide, as well as the
indirect owners costs, which include land, site preparation, pro-
ject management, training, contingencies, and commissioning
costs. The OCC excludes nancing charges known as Interest
During Construction. Further details on OCC can be found in
Section 3.2.
We expand the scope of analyses to include the costs of 32 US
and eight French reactors built prior to 1970. Beyond the US and
France, we collect complete cost histories for Japan, South Korea,
West Germany, Canada, and India (153 reactors in total). To sum-
marize, our study provides costs for a full set of reactors in seven
countries, covering builds from 1954 through projects that had
been completed by the end of 2015, covering costs for 58% of all
power reactors ever built globally.
2. Literature review
Experience curves, progress ratios, and learning rates are all
methods of analysis that were originally used to compare in-
novation and advancement across aircraft manufacturing rms
(Wright, 1936), and have since been employed to analyze devel-
opment of a broad range of technologies including power plants
(Zimmerman, 1982;Joskow and Rose, 1985). Since nuclear power
plants are complex infrastructure projects not a product that
rolls off an assembly line a range of factors go into the nal cost.
To isolate learning effects for a specic reactor developer, many
studies have used regression models to isolate for a theoretical
learning-by-doing based on a manufacturer or architect-engineer
rm's progress.
Cantor and Hewlett (1988) summarized four such regression
studies that attempt to isolate the effects of learning, economies of
scale, and regulatory changes on Overnight Capital Costs of US
reactors. They found that individual rms experience learning, but
that the increased size of plants and increased regulation led to
longer lead times and higher overnight costs, thus offsetting any
learning-by-doing effect.
Kouvaritakis et al. (2000);Jamasb (2007), and Kahouli (2011)
also derived learning rates for nuclear construction costs for the
OECD and EU. They found that learning-by-searching, ie, im-
provement through R&D, can have an important effect. Berthélemy
and Escobar-Rangel (2015) performed a regression analysis to
isolate hypothetical cost drivers, including learning effects, using a
combined data set of French and US reactors. They nd that
standardization of reactor designs is key for decreasing lead times
and costs, and that innovation can actually lead to higher capital
costs and longer lead times.
While these studies calculate theoretical learning rates for
specic developers and construction rms, it is difcult to truly
isolate learning effects when so many other factors were changing
at the same time as rms potentially gained experience. Jamasb
(2007) demonstrated how incorporating multiple factors such as
technological improvements due to research and development
changed the learning-by-doing rate signicantly. Clarke et al.
(2006);Söderholm and Sundqvist (2007), and Pan and Köhler
(2007) warned against using learning curves beyond the scope of a
manufacturing rm, since there are many drivers of cost reduc-
tions that are unrelated to replications or experience. These dri-
vers include market demand, supply chain, labor relations, re-
search and development, and regulation.
Given these conation issues and in the absence of any causal
framework a simpler method is to look at historical cost trends
for reactors built within a specic country over time or by cu-
mulative deployed capacity; this metric is often referred to as an
experience curve. Such analysis can be likened to industry-wide or
country-wide learning and can shed light on the combined effect
of developer experience, learning-by-doing, and the accumulation
of other time-related cost drivers. Analyzing the historical ex-
perience in this way has been a common approach to help un-
derstand the prospects and challenges of nuclear power. Past
studies (Thomas, 1988;MacKerron, 1992;Koomey and Hultman,
2007;Escobar-Rangel and Lévêque, 2015) have documented dra-
matic cost escalation and have identied the presence of negative
learning-by-doing,suggesting an intrinsicor inevitable increase
in costs (Grubler, 2010). These results have played a role in in-
forming integrated assessment modellers and policy makers (Neij,
2008;Junginger et al., 2008;Harris et al., 2013;Azevedo et al.,
The phenomenon of cost escalation has been interpreted as a
lack of learning in the traditional sense of rm-level production,
but the studies have deployed a broader use of the term to de-
scribe a theoretical country-level, industry-wide learning. Experi-
ence curves may not be able to isolate rm-level learning, but they
can be useful in highlighting differences between the experiences
across countries or during different phases of reactor development
within a single country. Importantly, experience curves do not
provide a causal explanation of cost drivers for nuclear power (or
other energy technologies), but can help quantify historic trends
and lead to future case studies or econometric studies.
Despite these constraints, the single-factor learning curve
methodology has been commonly and broadly applied in studies
of nuclear cost trends, due to the availability of data and its ease of
use (Jamasb, 2007). Of particular note, Grubler (2010) analyzed the
historical costs of nuclear power for France and the US, and con-
cluded that nuclear power construction costs invariably exhibited
negative learningand forgetting by doing,citing an increase in
system complexity for nuclear power construction, which was
hypothesized by Lovins (1986),Bupp and Derian (1978), and Ko-
manoff (1981). Additionally, Grubler (2010), using Fig. 1, observed
arhythm of cost escalationfor both the US and France, de-
scribing 20 GW and 50 GW as threshold levelsat which cost
escalation acceleratedand skyrocketed.
Regardless of the methods used to analyze cost trends, the
existing literature mainly ignores cost data in several dominant
and emerging nuclear countries. The data analyzed for 99 US
J.R. Lovering et al. / Energy Policy 91 (2016) 371382372
reactors by Koomey and Hultman (2007) and 58 French reactors
(29 pairs) by Escobar-Rangel and Lévêque (2015) only represent
26% of the total number of nuclear power reactors completed in
the world (596 reactors in 34 countries). The US and France were
two of the nuclear pioneers, but until 2008, the last construction
start of a nuclear reactor was 1978 in the US and 1984 in France. In
the 24 years between 1984 and 2008, 96 reactors began con-
struction around the world. The exclusive study of the cost ex-
perience of the US and France neglects the experiences of other
pioneers such as Canada (25 completed reactors), the UK (45), and
USSR (39), early adopters such as Germany (36), Japan (62), and
India (21), and later adopters such as the Republic of Korea (24)
and China (28).
Even in the oft-studied US and France, most analyses have
excluded early eets of nuclear reactors, particularly demonstra-
tion reactors and commercial reactors built on turnkey contracts,
where reactors were built at a xed price. While there are good
reasons for excluding such reactors (Koomey and Hultman, 2007;
Escobar-Rangel and Lévêque, 2015), experience curve analyses for
other energy industries often include the early, small-scale de-
monstration plants (Ordowich et al., 2012;Rubin et al., 2007).
With a full cost history for a given energy technology, it is easier to
observe different eras when different drivers of cost trends may
have been dominant, such as R&D early on, economies of scale
later, economies of multiples, and nally industrial learning or
regulatory stability.
The state of the existing literature highlights the need for a
broader and more holistic analysis of the historical experience and
trends of nuclear costs. Rather than perform analysis in search of
learning-by-doing or economies of scale which may prematurely
misattribute cost trends based on a narrow data set we start
with an in-depth survey of nuclear construction costs across a
diverse set of countries. Once complete, further studies can ex-
plore cost drivers in each country.
3. Methodology
3.1. New data on reactor costs
This study provides a more holistic picture of nuclear costs by
lling important gaps in previous analyses, collecting costs for a
total of 349 reactors in seven countries (58% of all reactors built
globally). We use the IAEA Power Reactor Information System
database for a globally comprehensive listing of nuclear power
reactors completed as of 2015. We curate construction cost data
sets for 32 previously omitted reactors in the US, eight in France,
and new costs for 153 reactors in West Germany, Canada, Japan,
South Korea, and India. Our goal was to obtain complete cost trend
data for the countries with the largest nuclear reactor eets.
However, reliable and complete data could not be found for Russia
(39 nuclear power reactors, third largest nuclear energy producer),
China (28 reactors, fth largest producer), and for the second
generation of gas-cooled reactors in the UK (13 reactors built from
1967 through 1980).
Nuclear power encompasses a wide variety of reactor tech-
nologies, but over 80% of operating reactors around the world are
light-water reactors. For this study, we include cost data for light-
water reactors, as well as for heavy-water reactors and gas-cooled
reactors where data were available. While some cost estimates
were available for demonstrations of advanced reactors such as
fast breeder reactors, we excluded them from our cost trend
analysis if the demonstration was one-off and not followed by
commercial reactors of the same design.
3.1.1. Data from US and France
Past studies generally have cited a lack of data availability for
their partial coverage of the US and France (Koomey and Hultman,
2007;Komanoff, 1981), or have limited their scope of analysis to
operational second-generation reactors (Grubler, 2010). Because
this study intends to thoroughly explore the cost experience of
nuclear power, we present the costs of the rst-generation re-
actors for analysis and comparison.
We collected costs for 131 US reactors from several sources. For
99 reactors that began construction between 1967 and 1978, we
use the data from Koomey and Hultman (2007), which stem from
a database built by Komanoff (1981). The Koomey and Hultman
(2007) data are adjusted for minor methodological differences
regarding ination adjustments (see Appendix).
For the 14 reactors ordered in the US in the turnkey era be-
tween 1962 and 1968, we use estimates of the costs to the con-
tractors and builders prepared by United Engineers and Con-
structors in the Atomic Energy Commission (AEC) WASH-1345
report from 1974, as documented by Burness et al. (1980).Note
that these estimated costs are higher than the turnkey contract
price, because most of these plants were built at a loss to early
developers trying to gain market share (see Appendix).
For the 18 earlier demonstration
reactors that began con-
struction between 1954 and 1963, we use cost gures found in
IAEA (1963) and Loftness (1964), who compiled gures from
USAEC reports. The cost information for these turnkey and de-
monstration reactors used in this analysis differ from their re-
ported contract prices. The historical context (see Appendix)
suggests that the cost estimates are a more accurate representa-
tion of costs than the turnkey contract prices, and that they should
be seen as lower-bounds for the true costs.
For France, we use the Cour des Comptes (2012) construction
cost data for 66 nuclear power reactors, including seven gas-
cooled reactors and one PWR that did not receive attention from
Escobar-Rangel and Lévêque (2015);Boccard (2014),orGrubler
(2010). We exclude costs for France's two fast-breeder reactors,
since they were not followed by commercial designs, and we ex-
clude costs for two prototype gas-cooled reactors because data
were unavailable.
Fig. 1. Negative learning by doing in nuclear power, reprinted from Grubler (2010).
Red squares denote US costs and blue triangles denote French costs. Data from
Grubler (2010) were estimated before the release of reactor-specic costs from the
French Cour des Comptes in 2012. See Boccard (2014) and Escobar-Rangel and
Lévêque (2015) for discussion.
For this paper, demonstration reactors are limited to those that are ordered
by utility companies and connected to the grid for the purposes of power gen-
eration, and are not for research or experimental purposes.
J.R. Lovering et al. / Energy Policy 91 (2016) 371382 373
3.1.2. Data from other countries
For Canada, Yu and Bate (1982), representing the Atomic En-
ergy of Canada Ltd. (AECL), provided detailed cost data for 23 re-
actors. These data are corroborated and supplemented with data
from IAEA (1963) for the rst reactor at Rolphton, and from Cantor
(1985);McConnell et al. (1983);Thomas (1988), and the Pembina
Institute (2004) for nal costs of reactors completed after 1984, 24
out of 25 reactors in Canada. The cost of one reactor in Quebec was
For Germany, Kim (1991) obtained detailed cost data from plant
operators for 21 reactors in West Germany. Additional data for
earlier reactors were obtained from IAEA (1963) and El-Fouly
(1970). Costs were not available for the six reactors constructed in
East Germany and four experimental reactors: two high-tem-
perature gas-cooled reactors, one heavy-water gas-cooled reactor,
and one fast- reactor.
For Japan, the Institute for Energy Economics, Japan (IEEJ),
conducted a survey of utility companies in Japan to acquire cost
data for every Japanese reactor (IEEJ, 2012). We use IEEJ survey
results for 56 reactors, and corroborate the IEEJ data with partial
data from IAEA (1963),Marshall and Navarro (1991), and Du and
Parsons (2009), supplementing cost data for four more reactors to
cover all 60 reactors in Japan. We exclude costs for two experi-
mental reactors: the Fugen Advanced Thermal Reactor because no
data were available and the Monju Fast Breeder Reactor because it
was not followed by any commercial fast reactors.
For India, Bohra and Sharma (2006), representing the Nuclear
Power Company of India Ltd. (NPCIL), provided cost data for the
latest 16 reactors. Cost data for the earliest 14 reactors were also
presented in Ramana et al. (2005), who obtained data from the
Indian Department of Atomic Energy from 1996 and 2002 reports.
For the eight reactors covered by both studies, costs differ by
10% to þ20% after methodological adjustments.
For these eight
reactors, the cost information from Bohra and Sharma (2006) is
used, considering their recency, data access, and methodological
For Korea, a direct inquiry to the Korea Hydro and Nuclear
Power (KHNP) utility company collected data for all 26 reactors in
South Korea. These data are corroborated by information from the
KHNP (2014) annual report, the National Assembly Budget Ofce
(NABO) (2014), and Korea Power Exchange (KPX) (2013).
3.2. Metrics for comparison
We use the Overnight Construction Cost (OCC) specic to each
reactor as the metric of comparison in this study. The OCC includes
the costs of the direct engineering, procurement, and construction
(EPC) services that the vendors and the architect-engineer team
are contracted to provide, as well as the indirect owner's costs,
which include land, site preparation, project management, train-
ing, contingencies, and commissioning costs. For heavy-water re-
actors, the OCC includes the cost of the initial heavy-water in-
ventory. The OCC includes back-t costs but excludes retrotting
or capital expenditures after rst operation and the cost of the
initial fuel core.
The OCC represents the single largest component of the total
levelized cost of generating electricity with nuclear power, typi-
cally accounting for roughly 55%. In this study, we focus ex-
clusively on OCC because the other lifecycle costs approximately
15% for Interest During Construction (IDC), 15% for O&M and de-
commissioning provision, and 15% for fuel and provisions for used
fuel are more predictable and have had far less variation over
time and country (Dhaeseleer, 2013).
The overnightattribute refers to the construction cost of a
nuclear reactor as if the reactor construction process were com-
pleted instantaneously, without incurring the nancing charges
accrued before commercial operation, known as Interest During
Construction (IDC). The OCC metric is meant to isolate the cost
invariant to construction duration and interest rate, in order to
capture the cost intrinsic to the reactor technology.
For some
countries, costs were already reported in overnight terms, while in
others, we adjusted the data by removing the IDC. Then, the data
are adjusted for ination by their country-specic GDP deator to
the year 2010 from World Bank data. Finally, the OCC is normal-
ized to a per-kilowatt power rating based on the original designs
net power capacity, as reported in the IAEA PRIS database.
Although we exclude IDC from this analysis, it is worth noting
how large an effect IDC has on the total direct cost of a nuclear
plant. An anonymized survey of actual construction costs for US
plants found that on average IDC comprised 46% of the total up-
front cost (DOE, 1988). Davis (2011) found that IDC could range
from 12% of total construction costs with a 5% cost of capital and a
ve-year construction duration to 54% of total construction costs
for a 15% cost of capital and a ten-year construction duration.
Factors that affect IDC are much better understood, primarily
construction duration, discount rate, and the weighted average
cost of capital (Dhaeseleer, 2013).
In contrast to other studies that assess historical cost trends by
the reactor's date of commercial operation (Koomey and Hultman,
2007;Grubler, 2010), this study uses reactor construction start
dates from the IAEA PRIS database, dened as the rst foundation
concrete pour. Because construction durations have been ex-
ceptionally long, up to 1020 years at the extremes, the state of
technology and the reactor designs are not representative of the
date of eventual completion, but rather, more representative of the
date of the start of construction. Using construction start dates to
analyze the nuclear power experience allows for a focus on the
cost characteristics of the best available technologyat the time
of deployment, consistent with the technological learning litera-
ture (Junginger et al., 2008).
We present historical cost trends for each country to illustrate
the detailed and varied histories of its domestic nuclear power
industry. A country-level analysis of costs in local currency units
avoids issues related to currency valuation and international var-
iation in labor costs. Mott MacDonald (2011) estimated that labor,
either on-site or embodied in the supply chain, accounts for two-
thirds of the capital expenditure of nuclear power stations.
4. Results and discussion
4.1. US and France: new patterns do not t old stories
The US and France are the two countries subject to the most
analysis in the existing literature of nuclear cost trends. While
Koomey and Hultman (2007) and Escobar-Rangel and Lévêque
(2015) have shown sharply rising costs for the US, and smaller but
steadily rising costs for France, our more complete data set shows
that there is not a single story of rising costs, but a more complex
Costs for four earlier reactors were 1020% higher than the comparable ad-
justed gure from Ramana et al. (2005), and costs for four later reactors were 5
10% lower than the comparable adjusted gure in Ramana et al. (2005).
The overnight construction cost will inevitably include time-related costs
related to delays, such as additional price ination for nuclear-specic materials,
equipment, and labor, extension of equipment rentals, labor retention, and back-t
costs caused by regulatory changes during construction.
J.R. Lovering et al. / Energy Policy 91 (2016) 371382374
4.1.1. Cost trends in the United States
By capturing a full overnight construction cost history for the
US by construction start date, four distinct phases of nuclear
power construction become visible, shown in Fig. 2.
Between 1954 and 1968, starting with the rst reactor at
Shippingport, 18 demonstration reactors were ordered and com-
pleted. In this rst phase, overnight construction cost (OCC) de-
cline sharply, from a high of $6800/kW to a low of $1300/kW, an
81% drop, or an average annualized rate of decline of 14%. In this
period, reactor size increases from under 80 MW to 620 MW,
suggesting economies of scale were important. The second phase,
from 1964 to 1967, represents the era of turnkey contracts. The
OCC of these 14 reactors are in the range of $1000-1500/kW, a 33%
drop, or an average annualized rate of decline of 13%. In this per-
iod, reactor sizes increase to a range of 8001100 MW.
The cost experience of these earlier reactors contrasts sharply
with the picture in the existing literature. Koomey and Hultman
(2007) started with the rst non-turnkey reactor, Palisades, which
started construction in 1967 and cost $650/kW once completed.
Between 1967 and 1972, the 48 reactors that were completed
before the Three Mile Island accident in 1979 began construction.
Their OCC rise from a range of $600$900/kW to approximately
$1800$2500/kW. These reactors follow a trend of increasing costs
by 187%, or an annualized rate of 23%. Phung (1985) attributed
these pre-TMI cost increases to emerging safety requirements re-
sulting from pre-TMI incidents at Browns Ferry and Rancho Seco.
Two outliers, Diablo Canyon 1 and 2, cost about $4100/kW in
overnight construction cost, and were completed 17 and 15 years
later, in 1984 and 1985.
A break in construction starts is visible around 1971 and 1972,
which is likely attributable to a conuence of events affecting
nuclear power construction in the late 1960s and early 1970s.
These include the establishment of the Environmental Protection
Agency in 1971, and the AEC's gradual loss in public trust and its
eventual replacement by the Nuclear Regulatory Commission
(NRC) in 1975. Golay et al. (1977) determined that 88 reactors in
various stages of permitting, construction, and licensing were af-
fected by the 1971 Calvert Cliffs court decision resulting in revised
AEC regulations that included back-t requirements.
Finally, the last 51 completed reactors represent a set that be-
gan their construction between 1968 and 1978 and were under
construction at the time of the Three Mile Island accident in 1979.
For these reactors, OCC varies from $1800/kW to $11,000/kW.
Thirty-eight of these reactors fall within a mid-range of $3000/kW
to $6000/kW, with 11 between $1800 and $3000/kW and 10 be-
tween $6000 and $11,000/kW. From the OCC of about $2,000/kW
for reactors beginning construction in 1970, OCC increases another
50200%, or an annual increase of 515% between 1970 and 1978.
To better understand the impact of Three Mile Island on nu-
clear power construction, we plot reactor construction durations,
represented by the time between the construction start date and
the grid connection date, as an additional variable to OCC in Fig. 3.
When the full cost experience of US nuclear power is shown
with construction duration experience, we observe distinctive
trends that change after the Three Mile Island accident. As shown
in Fig. 3 in blue, reactors that received their operating licenses
before the TMI accident experience mild cost escalation. But for
reactors that were under construction during Three Mile Island
and eventually completed afterwards, shown in red, median costs
are 2.8 times higher than pre-TMI costs and median durations are
2.2 times higher than pre-TMI durations. Post-TMI, overnight costs
rise with construction duration, even though OCC excludes the
costs of interest during construction. This suggests that other
duration-related issues such as licensing, regulatory delays, or
back-t requirements are a signicant contributor to the rising
OCC trend. Phung (1985) observed retrot costs due to new safety
requirements before and after TMI. Rust and Rothwell (1995) ar-
gued this rise was due to unprecedented regulatory ux and un-
certainty post-accident. Hultman and Koomey (2013) disputed the
economic impact of the Three Mile Island accident on the US nu-
clear industry, but failed to observe its distinctive effects on
overnight construction costs. These results suggest that the Three
Mile Island accident in 1979 did uniquely affect the nuclear in-
dustry in terms of overnight construction cost.
4.1.2. Cost trends in France
The French nuclear power overnight construction cost
history follows distinct stages that we highlight in Fig. 4. The cost
history of the rst era of French nuclear power has not been
discussed in previous studies of French construction costs by
Grubler (2010) or Escobar-Rangel and Lévêque (2015). Prior to the
second-generation PWRs, France built a series of indigenously
designed gas-cooled reactors (GCRs). These reactors fall in cost
from 6500/kW to 1200/kW between 1957 and 1966, a decline of
82%, or 17% annualized. These GCRs scale from a size of 68 MW to
540 MW.
In 1962, a single imported Westinghouse 305 MW pressurized
water reactor, Chooz-A, began construction and cost
Fig. 2. Overnight Construction Cost of US Nuclear Power Reactors by Construction
Start Date.
Fig. 3. Overnight Construction Cost and Construction Duration of US Nuclear Re-
actors. Color indicates the commercial operation start date.
J.R. Lovering et al. / Energy Policy 91 (2016) 371382 375
approximately 2000/kW. Thomas (1988) gave credit to the par-
ticipation of Framatome and EdF in the early construction ex-
perience of this reactor, under a technology license from Wes-
tinghouse, for valuable lessons deployed in the second era of nu-
clear power in France.
From 1971 to 1991, when the French began rapidly expanding
their domestic nuclear industry, OCC rises slowly from 1000/kW
to 1500/kW2000/kW, representing a 50% to 100% increase, or a
2% to 4% annualized rate of escalation. Within this second phase,
the CP series of reactors increase in cost, while the costs of the
reactors in the P4 series are remarkably stable. The last two pairs
of reactors at Chooz-B and Civaux were the result of the N4 pro-
gram, which was intended to indigenize the reactor design and
move away from designs based on the Westinghouse license. The
costs at Chooz-B-1 and -2 deviate from the strong trend of stable
costs around 1400/kW.
While the cost escalation in France is trivial compared to the US
experience, it does require some explanation. Escobar-Rangel and
Lévêque (2015) conclude that rising labor costs (faster than ina-
tion), technological change due to increased regulation, and in-
creased complexity due to larger reactors led to higher
costs. Escobar-Rangel and Lévêque (2015) credited the vertical
integration of the utility and reactor developer, standardization of
reactor designs, and multi-siting of reactors, for keeping costs
An analysis of French nuclear construction costs and con-
struction durations together in Fig. 5 shows that the Chernobyl
accident in 1986 resulted in a small but observable increase in
costs and a steady increase in construction duration. In contrast to
the US experience with the Three Mile Island accident, the French
nuclear power construction cost and duration trends were much
less affected by the accident at Chernobyl. This is seen by com-
paring Figs. 3 and 5.
One challenge with the data is that in France, reactors were
built in pairs, resulting in an averaged construction cost for all of
the reactors at the same site. The OECD Nuclear Energy Agency
described the potential for cost savings for subsequent reactors at
the same site, from staged schedules, shared equipment and fa-
cilities, and construction experience (OECD NEA, 2000). However,
since costs for multiple-unit plants have often been reported in
aggregate, this level of learning has been obscured in the data (see
4.2. Other Western nuclear powers: data from Canada and West
Canada and Germany are two other Western countries that
early on had considerable nuclear power programs. The Canadian
cost history, shown in Fig. 6, is similar in shape to the French ex-
perience: sharply declining costs and then relatively mild cost
Canada built its rst nuclear power reactor in 1957, a 17 MWe
demonstration reactor, at an overnight cost of CAD$11,000/kW.
Canada's Deuterium Uranium (CANDU) reactor design, which uses
heavy water as a moderator, was scaled up to 200 MWe and then
500 MWe. The costs of these reactors, ordered from 1960 to 1974,
including the initial heavy water inventory, were between $2000
and $3000/kW, representing a cost decline of approximately 77%
(8% annualized) from the rst CANDU. Construction of larger
plants consisting of sets of four reactors started in 1971, and the
last construction start occurred in 1986. Twelve reactors even-
tually cost $25003000/kW, while six other reactors, the latter
four at Darlington, cost near $4000/kW, a 60% increase in costs, or
a 4% average annual increase. The mild cost escalation experienced
Fig. 4. Overnight Construction Cost of French Nuclear Reactors by Construction
Start Date.
Fig. 5. Overnight Construction Cost and Construction Duration of French nuclear
reactors. Color indicates the commercial operation start date.
Fig. 6. Overnight Construction Cost of Canadian Nuclear Reactors by Construction
Start Date. Canada's reactors stayed smaller in power rating and were almost al-
ways built in pairs or quadruplets.
J.R. Lovering et al. / Energy Policy 91 (2016) 371382376
in Canada could be due to consistency in builders and manu-
facturers, the smaller reactor sizes, or that reactors were almost
always built in tight pairs (close in time), as can be seen in Fig. 6.
The cost experience in West Germany follows a similar pattern
as the other Western countries, shown in Fig. 7. Germany ordered
its rst reactor from GE in 1958, and the Vak Kahl reactor was
reported to cost in the range of 2700/kW. Construction costs for
plants that began construction between 1965 and 1973 decline to
approximately 1000/kW, a 63% reduction (6% annualized). For
plants beginning construction between 1973 and 1983, there is a
trend of increasing cost from 1000 to 3000/kW, a 200% increase,
or a 12% average annual increase.
4.3. New cost histories: data from Japan, India, and South Korea
The cost experience of the US and France is important for
analysis because they have the rst and second largest eets of
nuclear reactors and were leaders in the early commercial nuclear
power industry. The cost histories of Canada and Germany provide
additional context as to the experience of other nuclear pioneers
in the Western world. However, these four countries share similar
rst-mover disadvantages, obstacles, and setbacks that come with
deploying an emerging, immature technology (Jamasb, 2007). For
a more complete global picture, we present new comprehensive
data from Japan, South Korea, and India, which now have the third,
fth, and seventh largest operating
reactor eets in the world.
More importantly, Japan, India, and South Korea continued
building nuclear power plants through the 1990s and 2000s when
the nuclear industry had stagnated in the United States and
In Japan, the cost trend repeats an L- or U-shape seen in the
four previous countries, but evolves differently afterward, as
shown in Fig. 8. In the rst phase of Japan's nuclear power history,
from 1960 to 1969, reactors were rst imported from American
and British companies, with costs of approximately ¥1,100,000/kW
and ¥600,000/kW for a 10 MW boiling water reactor and a
159 MW gas-cooled reactor, respectively. Reactor size increases
between 300 and 700 MW, and costs fall to ¥150,000/kW for re-
actors by 1971, representing an 82% decline (16% annualized). The
second phase, from 1970 to 1980, represents when Japanese
industries took over the construction and manufacturing of re-
actors. During this time, reactor size also grows to an average of
950 MW. The overnight construction cost increases from
¥150,000/kW to ¥300,000/kW, an increase of 100%, or an 8% an-
nual rate of escalation. In the third era of nuclear power con-
struction in Japan, from 1980 to 2007, costs remain between
¥250,000/kW and ¥400,0 00/kW, representing an annual change of
1% to 1%. This period experienced relatively stable costs over 27
The Indian cost history, shown in Fig. 9, differs from both
Western and East Asian experiences. India ordered a pair of boiling
water reactors from General Electric in 1964, at a price of `45,000/
kW, and a pressurized heavy-water reactor (PHWR) of the CANDU
design from Atomic Energy of Canada Ltd. (AECL) in 1965, at a
price of `65,000/kW. After testing its rst nuclear weapon in 1974
and their exclusion from the Non-Proliferation Treaty, India pro-
ceeded to design and deploy their own indigenous PHWR design
at 200 MW power capacities, which were much smaller than other
standard designs. The rst three of these indigenous reactors,
which began construction between 1968 and 1972, cost between
`35,000/kW and `45,000/kW, a decline of 38% (7% annualized)
Fig. 7. Overnight Construction Cost of West German Nuclear Reactors by Con-
struction Start Date. No data for East Germany and non-LWR/PHWR reactors.
Fig. 8. Overnight Construction Cost of Japanese Nuclear Reactors by Construction
Start Date. While Japan's costs increased from the early 1970s to the early 1980,
costs have remained relatively stable, albeit varied, over the last 30 years. Dis-
economies of scale are not as apparent as seen in the US and France.
Fig. 9. Overnight Construction Cost of Indian Nuclear Reactors by Construction
Start Date.
As of publication, most of Japan's reactor eet is temporarily closed awaiting
potential restarts after safety upgrades prompted by the Fukushima Daiichi acci-
dent in 2011. However, Japan is still considered to have 48 reactors operational.
J.R. Lovering et al. / Energy Policy 91 (2016) 371382 377
from the rst CANDU. However, the costs of the next eight reactors
ordered between 1976 and 1990 jump to `90,000/kW to `110,000/
kW, an increase of 150% (5% annualized). Following a ten-year
pause in reactor construction, six reactors of the Indian design and
two of the Russian VVER design began construction between 2000
and 2003, and cost approximately `90,000/kW, approximately 10%
lower than the cost of the previous era.
The data from South Korea, shown in Fig. 10, tell an entirely
new story. Notably, Korea entered the nuclear market much later
than US, France, Canada, Germany, or Japan. Korea's rst power
reactor, a 558 MW Westinghouse design, began construction in
1972, and cost approximately 4,000,000 KRW/kW. Korea skipped
the early, small-scale demonstration phase and went straight to
importing a large commercial reactor. In its rst phase of con-
struction, Korea continued to import several reactor designs from
American, French, and Canadian companies, a total of 9 between
1972 and 1993. These plants cost between 2,500,000 KRW/kW and
4,500,000 KRW/kW. Within this era of imported designs, costs fell
approximately 25% (2% annualized). In 1989, Korea began con-
struction on their rst domestically developed OPR-1000 design,
based on Westinghouse, Framatome, and Combustion Engineering
designs. This rst reactor cost 2,600,000 KRW/kW. Twelve reactors
of this standard design began construction between 1989 and
2008, and their costs declined in a stable manner, sitting between
2,000,000 KRW/kW and 2,500,000 KRW/kW, representing a 13%
cost decline (1% annualized). Overall, from the rst reactor in
Korea in 1971, costs fell by 50%, or an annual rate of decline of 2%
for the entire Korean nuclear construction history. This is in sharp
contrast to every other country for which we present cost data.
In Table 1, we provide a summary of the cost trends for each
country. Rather than calculating a single learning rate for each
country, we separate out specic eras and present the annualized
rate of change in overnight construction cost for these eras and the
total change in OCC in each era.
4.4. Global trends and experience curves
4.4.1. National experience curves of nuclear construction
A broader look at nuclear cost history allows us to analyze a
new set of country-level experience curves for nuclear power.
Rather than look at a trend in a single country, we have presented
complete cost histories for seven countries. Fig. 11 shows Over-
night Construction Costs scaled to the cost of the rst non-de-
monstration reactor in each country.
The most surprising feature is the large diversity in trends, with
the US and South Korea at the two extremes. Countries building
reactors more recently, particularly those with construction starts
after 1980, have different trend shapes than the early nuclear
pioneers. Rather than an invariable exhibition of negative learn-
ingand inevitableincreases in complexity intrinsic to nuclear
technology that lead to cost escalation (Grubler, 2010), it is clear
that there is not a singular cost trend for nuclear technology, but a
plurality of different country-specic experiences. A consistent
rhythmof cost escalation suggested by Grubler (2010) does not
match the historical record.
We see that the pre-commercial demonstrationcosts of glo-
bal nuclear reactors generally fell in the 1950's and 1960's. Ex-
cluding demonstration plants, reactor costs in the US, France, Ca-
nada, West Germany, and Japan experienced some cost declines in
the early years of commercialization, as shown by the costs below
1.0 in the 1960's and 1970's, before rising above 1.0.
One issue with experience curve analysis is the choice of sys-
tem boundary, which may exclude important parts of a learning
system, and could produce misleading results (Junginger et al.,
2008). Technologies are often globally developed, and nuclear
power is no exception, as evidenced by the diffusion of reactor
Fig. 10. Overnight Construction Cost of South Korean Nuclear Reactors by Con-
struction Start Date. South Korea began by importing large foreign reactor designs,
and then began building their own indigenous reactor in 1989. Also note that most
of South Korea's reactors were built in pairs.
Table 1
Summary of Cost Trends by Country.
Country Era (dened by time period in which reactors began construction) Annualized rate of change in OCC (%/yr) Total change in OCC by era (%)
USA 19541968, 18 demonstration reactors 14% 81%
19641967, 14 turnkey reactors 13% 33%
1967 1972, 48 reactors completed pre-TMI þ23% þ190%
19681978, 51 reactors completed post-TMI þ5toþ10 % þ50 to þ200%
France 19571966, 7 gas-cooled reactors 17% 82%
19711991, 59 light-water reactors þ2toþ4% þ50 to þ10 0%
Canada 1957 1974, 6 reactors 8% 77%
19711986, 18 reactors þ4% þ60%
West Germany 19581973, 8 reactors 6% 63%
1973 1983, 18 reactors þ12% þ200%
Japan 19601971, 11 imported reactors 15% 82%
1970-1980, 13 foreign designs þ8% þ100 %
19802007, 30 domestic reactors 1% to þ1% 17 % t o þ33%
India 1964-1972, 5 imported reactors 7% 38%
1971-1980, 8 domestic reactors þ5% þ150%
19902003, 6 domestic reactorsþ2 imported 1% 10%
South Korea 1972 1993, 9 foreign designs 2% 25%
19892008, 19 domestic reactors 1% 13 %
J.R. Lovering et al. / Energy Policy 91 (2016) 371382378
designs via exports and licenses. One way to explore this issue is to
analyze historical cost experience in a global context.
4.4.2. Nuclear construction costs in a global perspective
While the seven countries in this study present unique nuclear
construction histories, a global perspective can also produce ad-
ditional insight into the cost history of nuclear power. Fig. 12
shows the costs of nuclear reactors in the seven studied countries
in 2010$ USD equivalent. Compared to the picture commonly
presented in previous studies (see Fig. 1), Fig. 12 shows a more
complex set of trends.
The truncated cost history of the US can be seen as a global
outlier. Reactors in the US that began construction between 1971
and 1978 and were mid-construction during the Three Mile Island
accident experience a rapid increase in cost. No new reactors
started construction after 1978. This is in contrast to the reactors
in France that experience lower and stable costs. The cost ex-
perience of Canada and Germany follows patterns that are in be-
tween the US and France, with mild cost escalation for reactors
that began construction around 1970, and then atypically high
costs for the last few reactors beginning construction in the early
1980s. Like France, Canada and Germany also saw their last con-
struction starts in the 1980s.
The experiences of Japan, India, and Korea help ll in the pic-
ture past 1980. In Japan, costs also fall and escalate in a pattern
similar to Western countries, but with a lag of about ve years. The
costs of Japanese reactors beginning construction in the 1980s rise
above those in the 1970s, and become less consistent between
reactors, but do not appear to follow an escalation trend. In India,
costs increase from a low level and stabilize at the $2000/kW level.
Finally, in Korea, where nuclear power was adopted much later
than all six other countries, construction costs follow a steady
While this study is focused on presenting cost histories and not
analyzing potential drivers of cost trends, several possible ex-
planations are observable. In addition to the lower costs seen by
the later adopters of nuclear power, countries that emphasized
design standardization, such as in France and Korea, see more
stable costs, as qualitatively described by Choi et al. (2009) and
Lévêque (2014). Countries that consistently built reactors in pairs,
or larger sets at the same site, such as France, Canada, and Korea,
see lower costs than in the USA, Japan, and Germany.
4.4.3. Nuclear's cost experience curves in context
Other energy technologies have experienced similarly dramatic
rises and falls in cost. There is a large difference in learning curves
between small-scale modular energy technologies like solar pa-
nels and wind turbines and large energy infrastructure projects
like nuclear reactors and hydroelectric dams. In Fig. 13, the cost
experience curve of solar photovoltaics in Germany is compared
with the global nuclear construction cost experience based on
global cumulative deployment.
The construction cost history of nuclear power is signicantly
more varied than that of solar photovoltaics, but this is obscured
when it is presented as a single global learning curve such as in
Trancik (2006). While it is clear the construction cost of nuclear
power has experienced periods of rapid escalation in several
countries, there are also two periods in nuclear power construc-
tion history where costs declined sharply and at steep exponential
rates comparable to those experienced by solar photovoltaics: the
early development of nuclear power up to 100 GW, and the recent
experience of Korea. The massive declines in nuclear construction
Fig. 11. Comparison of Overnight Construction Cost Trends by Country Over Time
(indexed to cost of rst non-demonstration reactor).
Fig. 12. Overnight Construction Costs of Global Nuclear Reactors in USD2010. Costs
are adjusted by local GDP deator and to USD at 2010 market exchange rates.
Fig. 13. Historical Cost Experience Curve of Solar Photovoltaics compared with Cost
Experience Curves of Nuclear Power. Overnight construction cost is displayed as a
function of cumulative global installed capacity. The nuclear power costs are from
this study, and the solar photovoltaic costs are from Seel et al. (2014).
J.R. Lovering et al. / Energy Policy 91 (2016) 371382 379
costs in their early development suggest the critical role of cost
drivers other than learning-by-doing for nuclear power, such as
R&D, economies of unit scale, and economies of production scale.
The latest experience in South Korea, with its standardized design
and stable regulatory regime, suggests the possibility of learning-
by-doing in nuclear power.
The cost experiences of Japan and Korea at the latter end of
nuclear power construction history raise the possibility of spillover
and learning from the earlier experiences in other countries such
as the US and France. Thomas (1988) pointed out that learning in
nuclear power may require signicant operational experience in
addition to complete construction experience, before benets are
accrued in the form of new rationalized designs and lower con-
struction costs. This suggests that cost escalation with increased
experience is not an inevitable outcome.
On the other hand, rising capital costs for nuclear power have
been associated with improved safety and performance (Escobar-
Rangel and Lévêque, 2015).
Junginger et al. (2008) notes that the investment prices for
wind and solar technologies experienced negative learning post-
2002 and suggests the causes are increasing demand for these
technologies and rising raw material prices.
In the US, the capacity-adjusted price of a wind turbine rose by
100% from 2002 to 2008, an annual growth rate of 10%, caused bya
mix of diseconomies of scale, labor prices, steel prices, and cur-
rency movement (Bolinger and Wiser, 2012). Rather than predict
that wind power cost would continue to rise or that the technol-
ogy was inherently expensive, most analysis focused on under-
standing the drivers of these cost increases.
In Fig. 14, we show how the nuclear experience curves compare
to the US coal experience. In contrast to photovoltaic panels,
which are mass-produced in a factory, coal power plants are more
analogous to nuclear power plant construction. For both in-
dustries, plants faced increased environmental and safety regula-
tion that may have led to cost increases, along with increased
complexity of managing large construction projects. Between 1971
and 1978 in the US, the cost of a coal power plant increased at 8%
annually, in addition to construction-sector ination, primarily
due to increased air and water quality regulations (Komanoff,
1981). About 20% of this increase for coal plants could be attrib-
uted to new required equipment scrubbers and cooling towers
and the other 80% of cost increases were attributed to other
aspects of meeting tighter regulations: longer lead times for pro-
jects, more delays due to legal intervention, and higher labor costs
to evaluate environmental impacts (Joskow and Rose, 1985).
It is worth noting how nuclear power construction costs also
follow escalation trends similar to those for coal power in these
time periods and stage of development at the 100 GW level. Part of
the cost escalation of nuclear power may also reect changing
requirements in an era of extreme uncertainty for the technology,
and general trends in construction costs such as materials and
5. Conclusions and policy implications
This paper presents a new data set of historic experience curves
for overnight nuclear construction costs across seven countries.
From these data, we draw several conclusions that are in contrast
to the past literature. While several countries show increasing
costs over time with the US as the most extreme case other
countries show more stable costs in the longer term and cost
declines over specic periods in their technological history.
Moreover, one country, South Korea, experiences sustained con-
struction cost reductions throughout its nuclear power experience.
The variations in trends show that the pioneering experiences of
the US or even France are not necessarily the best or most relevant
examples of nuclear cost history.
These results show that there is no single or intrinsic learning
rate that we should expect for nuclear power technology, nor an
expected cost trend. How costs evolve over time appears to be
dependent on different regional, historical, and institutional fac-
tors at play. The large variance we see in cost trends over time and
across different countries even with similar nuclear reactor
technologies suggests that cost drivers other than learning-by-
doing have dominated the cost experience of nuclear power con-
struction. Factors such as utility structure, reactor size, regulatory
regime, and international collaboration may play a larger effect.
Therefore, drawing any strong conclusions about future nuclear
power costs based on one country's experience especially the US
experience in the 1970s and 1980s would be ill-advised.
5.1. Policy implications
Economic modelers and policy makers relying on projections of
future energy costs should be aware of the severe limitations and
frequent misapplication of theoretical learning rates in energy
technologies. Historical learning curves are context-dependent
and may not apply to future energy scenarios, where regulations,
values, and economies will be quite different from when and
where past nuclear power was deployed. The conation of learn-
ing curves and experience curves for the purposes of long-term
energy projections locks in implicit assumptions about the current
state of the domestic nuclear industrysuch as the state of tech-
nology, deployment rates, supply chain, and utility structure.
Koomey and Hultman (2007) and Grubler (2010) asked policy
makers to ground their assumptions of future costs with history
and to learn lessons from the past, but relied on a subset of the
global nuclear cost experience to make their case. There are many
lessons learned from these cases: the importance of reactor
standardization, multi-unit builds, and regulatory stability. But we
must also recognize that the US and France deployed their eet of
reactors in a unique stage of nuclear power history.
Rather than rely on learning curves to predict future costs,
decision makers should focus on pursuing and developing policies
that aim to drive the price of clean energy technologies down:
innovation policy, industrial policy, trade policy, and energy policy.
Where projections of future cost are necessary, they should be
Fig. 14. Historical Cost Experience Curve of Coal Power compared with Cost Ex-
perience Curves of Nuclear Power. Coal power plants have generally been cheaper
than nuclear power plants, but experienced a similar rise in costs more recently.
Coal costs are from McNerney et al. (2011).
J.R. Lovering et al. / Energy Policy 91 (2016) 371382380
based on relevant historical experience that matches the economic
drivers of their industry today. Assumptions regarding future costs
should reect the large variance in global and historic trends.
Rubin et al. (2007) note that the potential for costs [of several
GHG-mitigating technologies] to rise before they fall is an im-
portant nding affecting projections of future cost trendsand we
nd that this conclusion is applicable to the nuclear case.
5.2. Future work
Data for three other major nuclear powers- the UK, Russia, and
China- would add more breadth to our analysis of the historical
and institutional factors that inuence cost trends within coun-
tries. The costs of modern Russian and Chinese reactors would be
of particular interest, given their dominance in the nuclear con-
struction market today. In addition, there is a larger set of coun-
tries that may not have robust domestic nuclear industries, but
may have a different set of experience when importing and
building foreign reactor designs. Therefore, it would be useful to
gather data for countries with two to nine reactors, including
Belgium, Spain, Czech Republic, Taiwan, Switzerland, Slovakia,
Hungary, Finland, Pakistan, Argentina, Bulgaria, Romania, South
Africa, Mexico, and Brazil. These countries studied in aggregate
could provide insight on how small domestic nuclear programs
compare with the dominant countries, and how importing a small
handful of reactors compares with investing in a strong, domestic
While construction cost is the largest component of the cost of
nuclear power, other trends in factors such as operational and
maintenance costs, fuel, operational efciency, and capacity factor
have signicant inuence on costs. For example, Koomey and
Hultman (2007) showed that while construction costs ($/kW) of
the least and most expensive nuclear reactors in the US differed by
a factor of 12, the lowest and highest levelized cost of electricity
($/kWh) from these reactors only differed by a factor of 4. A future
study investigating trends in levelized cost of nuclear electricity
over time and across countries could provide further insight. Si-
milarly, with detailed performance data, future studies could at-
tempt to understand if higher-cost reactors are better performing
in the long-run, as suggested by Berthelemy (2012).
Our broader data set allows us to isolate more factors than
cumulative capacity where learning can occur. Given that learning
can be rm-level, industry-level, country-level, or global, future
studies could analyze the cost experience of specic reactor sets,
such as those built by the same architect-engineer, those with
their Nuclear Steam Supply System from the same manufacturer,
or those reactors in the same reactor model class or type. There
have been many regression studies that have attempted to isolate
the cause of cost increases in the past, but they have been con-
strained by their cost data.
J.L. designed the research and wrote the paper. A.Y. analyzed
the data and wrote the paper. T.N. designed the research and
wrote the paper.
The authors would like to thank Yuhji Matsuo of the Institute
for Energy Economics, Japan, and Sanghyun Hong of the University
of Adelaide for their assistance in data collection. The authors
appreciate the early review from Per Peterson, Harry Saunders,
Roger Pielke Jr., Loren King, Jesse Jenkins, Brian Sergi, David Hess,
and Ashley Finan. The authors received no specic funding for this
Appendix A. Supplementary material
Supplementary notes and data associated with this article can
be found in the online version at
Azevedo, I., Jaramillo, P., Rubin, E., Yeh, S., 2013. Modeling Technology Learning for
Electricity Supply Technologies, (May), 59. Report to Electric Power Research
Barron, R., McJeon, H., 2015. The differential impact of low-carbon technologies on
climate change mitigation cost under a range of socioeconomic and climate
policy scenarios. Energy Policy 80, 264274.
Berthélemy, M., 2012. What drives innovation in nuclear reactors technologies? An
empirical study based on patent counts. CERNA WORKING PAPER SERIES 2012-
Berthélemy, M., Escobar-Rangel, L., 2015. Nuclear reactorsconstruction costs : the
role of lead-time, standardization and technological progress. Energy Policy 82,
118 130.
Boccard, N., 2014. The cost of nuclear electricity: France after Fukushima. Energy
Policy 66, 450461.
Bohra, S.A., Sharma, P.D., 2006. Construction management of Indian pressurized
heavy water reactors. Nucl. Eng. Des. 236 (78), 836851.
Bolinger, M., Wiser, R., 2012. Understanding wind turbine price trends in the U.S.
over the past decade. Energy Policy 42, 628641.
Bosetti, V., Marangoni, G., Borgonovo, E., Diaz Anadon, L., et al., 2015. Sensitivity to
energy technology costs: a multi-model comparison analysis. Energy Policy 80,
Bupp, I.C., Derian, J., 1978. Light water: How the Nuclear Dream Dissolved. Basic
Books, Inc, New York City.
Burness, H.S., Montgomery, W.D., Quirk, J.P., 1980. The Turnkey Era in Nuclear
Power. Land Econ. 56, 2.
Cantor, R., 1985. An Analysis of Public Costs and Risks in the Canadian Nuclear
Industry. Duke University, United States.
Cantor, R., Hewlett, J., 1988. The Economics of Nuclear Power: Further Evidence on
Learning, Economies of Scale, and Regulatory Effects. Resour. Energy 10,
Choi, S., Jun, E., Hwang, I., Starz, A., Mazour, T., Chang, S., Burkart, A.R., 2009.
Fourteen lessons learned from the successful nuclear power program of the
Republic of Korea. Energy Policy 37 (12), 54945508.
Clarke, L., Weyant, J., Birky, A., 2006. On the sources of technological change: As-
sessing the evidence. Energy Economics 28 (5-6), 579595.
Cooper, M., 2014. Unavoidable Economics of Nuclear Power. Corporate Knights,
Cour des Comptes, 2012. Les coûts de la lière électronucléaire (The costs of the
nuclear power sector of France).
Dhaeseleer, W. D., 2013. Synthesis on the Economics of Nuclear Energy.
Davis, L., 2011. Prospects for US Nuclear Power After Fukushima. Energy Inst. Haas
Work. Pap. Ser.
DOE. 1988 Technical Reference Book for the Energy Economic Data Base Program
EEDB Phase IX,, DOE/NE-0092.
Du, Y., Parsons, J. E., 2009. Update on the cost of nuclear power. Center for Energy
and Environmental Policy Research Working Paper.
El-Fouly, M. F., 1970. Cost survey of nuclear power stations in industrialized and
developing countries. In: International Atomic Energy Agency Symposium on
nuclear energy costs and economic development.
Escobar-Rangel, L., Lévêque, F., 2015. Revisiting the cost escalation curse of nuclear
power New lessons from the French experience. Econ. Energy Environ.
Policy 4, 2.
Golay, M.M., Saragossi, I.I., Willefert, J.-M., 1977. Comparative Analysis of United
States and French Nuclear Power Plant Siting and Construction Regulatory
Policies and Their Economic Consequences.
Grubler, A., 2010. The costs of the French nuclear scale-up: a case of negative
learning by doing. Energy Policy 38 (9), 51745188.
Harris, G., Heptonstall, P., Gross, R., Handley, D., 2013. Cost estimates for nuclear
power in the UK. Energy Policy 62 (August), 431442.
Hultman, N.E., Koomey, J., Kammen, D.M., 2007. What History can teach us about
J.R. Lovering et al. / Energy Policy 91 (2016) 371382 381
the future costs of U.S. nuclear power. Environ. Sci. Technol. 41 (7), 20882093.
Hultman, N., Koomey, J., 2013. Three Mile Island: the driver of US nuclear powers
decline? Bull. At. Sci. 69 (3), 6370.
IAEA, 1963. Nuclear Power Costs. In: IAEA General Conference.
IEA, 2010. Projected Costs of Generating Electricity 2010. OECD Publishing, France
IEA, 2014. The way forward: ve key actions to achieve a low-carbon energy sector.
IEEJ, 2012. Summary and Evaluation of Cost Calculation for Nuclear Power Gen-
eration by the Cost Estimation and Review Committee. Executive summary
Background to the power generation cost estimation, May, pp. 112.
IPCC, 2014. Contribution of Working Group III to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change.
Jamasb, T., 2007. Technical change theory and learning curves: patterns of progress
in Electricity Generation Technologies. Energy J. 28 (3), 5172. http://dx.doi.
Joskow, P.L., Rose, N.L., 1985. The effects of technological change, experience, and
environmental regulation on the construction cost of coal-burning generating
units. RAND J. Econ. 16 (1), 1.
Junginger, H., Lako, P., Lensink, S., van Sark, W., Weiss, M., 2008. Technological
learning in the energy sector, Universiteit Utrecht.
Kahouli, S., 2011. Effects of technological learning and uranium price on nuclear
cost: preliminary insights from a multiple factors learning curve and uranium
market modeling. Energy Econ. 33 (5), 840852.
KHNP, 2014. . Annual Report.
Kim, J.-G., 1991. Economics of Nuclear and Coal in Germany. University of Essen,
Komanoff, C., 1981. Power Plant Cost Escalation: Nuclear and Coal Capital Costs,
Regulation, and Economics. Van Nostrand Reinhold Company, Inc, New York.
Koomey, J.G., Hultman, N.E., 2007. A reactor-level analysis of busbar costs for US
nuclear plants, 19702005. Energy Policy 35 (11), 56305642.
Korea Power Exchange (KPX), 2013. 현황 20122013.7.
Kouvaritakis, N., Soria, A., Isoard, S., 200 0. Modelling energy technology dynamics:
methodology for adaptive expectations models with learning by doing and
learning by searching. Int. J. Glob. Energy Issues 14, 104115.
Leibowicz, B.D., Roumpani, M., Larsen, P.H., 2013. Carbon Emissions Caps and the
Impact of a Radical Change in Nuclear Electricity Costs. Int. J. Energy Econ.
Policy 3(1), 6074.
Lévêque, F., 2014. The Economics and Uncertainties of Nuclear Power.
Loftness, R., 1964. Nuclear Power Plants. D. Van Nostrand Company Inc, United
Lovins, A.B., 1986. The origins of the nuclear asco.. In: J., Byrne, D., Rich (Eds.),
Transaction Books, New Brunswick.
MacKerron, G., 1992. Nuclear costs: why do they keep rising? Energy Policy 20 (7),
Marshall, J.M., Navarro, P., 1991. Costs of nuclear power plant construction: theory
and new evidence. RAND J. Econ. 22 (1), 148.
McConnell, L., Woodhead, L., Fanjoy, G., 1983. CANDU operating experience. In:
Nuclear Power Production, IAEA CN 42/68.
McNerney, J., Doyne Farmer, J., Trancik, J.E., 2011. Historical costs of coal-red
electricity and implications for the future. Energy Policy 39 (6), 30423054.
Mott MacDonald, 2011. Costs of low-carbon generation technologies. Study of the
UK Committe on Climate Change, May.
NABO, 2014. 의쟁과과2014. 3.
Neij, L., 2008. Cost development of future technologies for power generationA
study based on experience curves and complementary bottom-up assessments.
Energy Policy 36 (6), 22002211.
Ordowich, C., et al., 2012. Applying learning curves to modeling future coal and gas
power generation technologies. Energy Fuels 26, 753766.
Pan, H., Köhler, J., 2007. Technological change in energy systems: learning curves,
logistic curves and inputoutput coefcients. Ecol. Econ. 63 (4), 749758. http:
Pembina Institute, 2004. Appendix 2 Ontario's Nuclear Generating Facilities: A
History and Estimate of Unit Lifetimes and Refurbishment Costs.
Phung, D., 1985. Economics of nuclear power: past record, present trends and fu-
ture prospects. Energy 10 (8), 917934.
Ramana, M.V., DSa, A., Reddy, A., 2005. Nuclear energy economics in India. Energy
Sustain. IX (2), 3548, Retrieved from 〈〈〉〉.
Rubin, E.S., Yeh, S., Antes, M., Berkenpas, M., Davison, J., 2007. Use of experience
curves to estimate the future cost of power plants with CO
capture. Int. J.
Greenh. Gas Control 1, 188197.
Rust, J., Rothwell, G., 1995. Optimal response to a shift in regulatory regime: the
case of the US nuclear power industry. J. Appl. Econom. 10 S http://onlineli
Seel, J., Barbose, G., Wiser, R., 2014. An analysis of residential PV system price dif-
ferences between the United States and Germany. Energy Policy 69, 216226.
Söderholm, P., Sundqvist, T., 2007. Empirical challenges in the use of learning
curves for assessing the economic prospects of renewable energy technologies.
Renew. Energy 32 (15), 25592578.
Thomas, S., 1988. The Realities of Nuclear Power: International Economic and
Regulatory Experience. Cambridge University Press, Cambridge, UK.
Trancik, J.E., 2006. Scale and innovation in the energy sector: a focus on photo-
voltaics and nuclear ssion. Environ. Res. Lett. 1, 014009.
Wright, T.P., 1936. Factors affecting the cost of airplanes. J. Aeronaut. Sci. 3 (4),
122128 .
Yu, A.M., Bate, D.L., 1982. Trends in the capital costs of CANDU generating stations.
In: Proceedings of the IAEA International Conference on Nuclear Power
Zimmerman, M., 1982. Learning effects and the commercialization of new energy
technologies: the case of nuclear power. Bell J. Econ. 13, 297310.
J.R. Lovering et al. / Energy Policy 91 (2016) 371382382
... For instance, the U.S. has since the early 1990s imported most of its uranium fuel from Kazakhstan, Russia, Canada, Australia and elsewhere (EIA, 2021a). As for the cost of nuclear power, its construction costs have rapidly escalated since the 1960s in the U.S. and Germany, and to a lesser extent in France, Canada and Japan 1 (Cohn, 1997;Lovering et al., 2016). ...
... In recent years, however, these estimates have changed. For example, new nuclear plants can occasionally be built in five to six years (or even less), and a growing number of plants are having their operating licenses renewed and extended so they can run for up to 80 years (WNA, 2020b; Lovering et al., 2016;DOE, 2021). At the same time, however, there is a growing number of nuclear reactors being decommissioned, as many of the first generation of nuclear facilities that have reached the end of their operating licenses cannot be costeffectively upgraded to remain in service. ...
... Construction costs of new nuclear builds actually stabilized or even slightly declined since the 1980s in Japan, as well as South Korea and India(Lovering et al., 2016). ...
Full-text available
Nuclear power provided just 4.3% of global energy use in 2019. However, due to its low life-cycle greenhouse gas emissions, many have called for it to play a much more prominent role in the electricity mix. At the same time, nuclear power is one of the most controversial energy sources, owing to its risk of catastrophic accident and nuclear weapons proliferation, high cost, and unsolved problem of radioactive waste disposal. This chapter takes a neutral position on the future of nuclear power. Both a top down scenario modeling approach and a bottom up assessment of national nuclear expansion plans will be discussed to determine the likely future of nuclear power. Based on both approaches, it is found that nuclear power is highly unlikely to expand very much and contribute to a sustainable energy transition, at least not before 2040. In the most likely scenario, the majority of future nuclear power generating capacity will be due to life extensions of existing plants rather than new builds. Moreover, new builds are currently planned in less than 20 countries, mainly China, India and South Korea. Thus, because of the significant lead time required to introduce large-scale changes in the electric utility sector, I conclude that nuclear power will not be part of a sustainable energy transition, at least through the middle of the 21 st Century. Forthcoming in: Asif, Muhammad, ed. 2024. Handbook of Energy and Environment in the 21 st Century.
... These conclusions are supported by data from the construction of fleets of nuclear power stations in the USA and France in the 1980s, as reported by Lovering et al [2] in 2016 in their work on historical construction costs of global nuclear power reactors. The US built some fifty reactors using many different vendors, constructors and designs. ...
... The results summarised in Figure 1 show that for projects built in the 1970s and 1980s, the French power stations cost on average 60% less than those built in the US and were delivered as much as three years earlier. [2] The UK Energy Technology Institute (with CleanTech Catalyst & Lucid Strategy) in the Nuclear Cost Driver report [3], has highlighted more recent examples of good practice in nuclear power stations built in Japan and Korea that used standard designs and lean construction ideas. Their improved cost and schedule outcomes were similar to France in 1980s. ...
Full-text available
Lean methods of design and of construction are key to reducing the cost of nuclear power, as laid out in this article.
... The projects take many years to complete so there are few opportunities for the teams to learn from project to project and many opportunities for projects to be impacted by unforeseen events. [3] These conclusions are supported by data from the construction of fleets of nuclear power stations in the USA and France in the 1980s (Lovering [3]). The US built some fifty reactors using many different vendors, constructors and designs. ...
... The projects take many years to complete so there are few opportunities for the teams to learn from project to project and many opportunities for projects to be impacted by unforeseen events. [3] These conclusions are supported by data from the construction of fleets of nuclear power stations in the USA and France in the 1980s (Lovering [3]). The US built some fifty reactors using many different vendors, constructors and designs. ...
Conference Paper
Full-text available
Nuclear power is returning to the energy agenda in Europe and elsewhere across the globe because of the challenges of Climate Change and a refreshed understanding of the need for energy security. Nuclear power has a cost problem which is largely driven by the high cost of construction and long build schedules. This paper addresses the key ways of making nuclear power affordable by: taking a programme approach, adopting modern production methods that are commonplace elsewhere but not applied in the nuclear industry and by creating production systems that profit from series build as against one-off projects and drive efficiency throughout the supply chain.
... Another interesting observation is the major differences in nuclear construction costs around the world with the South Korean approach outperforming most other countries. Indeed, according to Lovering, Yip, and Nordhaus (2016), there is nothing inherent in the technology that predicates the cost escalation seen in some countries since the 80s. This is discussed more in detail in Section 2, but it indicates that how nuclear power projects are performed is critical for the costs and hence the topic of this paper. ...
... They do hold true because as Lovering, Yip, and Nordhaus (2016) show; there are no intrinsic cost escalations in nuclear power as many authors suggest. The South Korean numbers in Figure 2 are equally interesting. ...
... With the rapid growth in the social economy, traditional thermal power generation fails to meet societal needs and results in irreversible environmental pollution. The large demand for energy in various countries makes the vigorous development of nuclear energy inevitable [1]. Additionally, promoting the development of nuclear energy is one of the essential measures for China to adjust its energy structure and achieve the goal of being "dual carbon" [2]. ...
Full-text available
Accurately predicting severe accident data in nuclear power plants is of utmost importance for ensuring their safety and reliability. However, existing methods often lack interpretability, thereby limiting their utility in decision making. In this paper, we present an interpretable framework, called GRUS, for forecasting severe accident data in nuclear power plants. Our approach combines the GRU model with SHAP analysis, enabling accurate predictions and offering valuable insights into the underlying mechanisms. To begin, we preprocess the data and extract temporal features. Subsequently, we employ the GRU model to generate preliminary predictions. To enhance the interpretability of our framework, we leverage SHAP analysis to assess the contributions of different features and develop a deeper understanding of their impact on the predictions. Finally, we retrain the GRU model using the selected dataset. Through extensive experimentation utilizing breach data from MSLB accidents and LOCAs, we demonstrate the superior performance of our GRUS framework compared to the mainstream GRU, LSTM, and ARIMAX models. Our framework effectively forecasts trends in core parameters during severe accidents, thereby bolstering decision-making capabilities and enabling more effective emergency response strategies in nuclear power plants.
Full-text available
Com o objetivo de analisar, através de revisão bibliográfica, os aspectos ambientais da utilização da energia nuclear como matriz energética e apresentar a opinião do autor em base destes, o trabalho foi elaborado contendo o conceito de energia nuclear e o funcionamento de uma usina, bem como aspectos ambientais, segurança e monitoramento, licenciamento ambiental, rejeito e um comparativo com outras unidades geradoras de energia. Em vista do desafio de diminuir as emissões atmosféricas globais e ao mesmo tempo assegurar o atendimento ao crescimento da demanda energética mundial, a energia nuclear apresenta potencial para diversificar a matriz energética global. A utilização isolada de energias tradicionais já desenvolvidas, como petróleo, carvão, gás natural, hidrelétrica, biomassa e energia eólica impacta o planeta pela emissão de gases de efeito estufa, contaminantes, ocupação de grandes áreas e demais impactos ambientais originados pelo seu funcionamento. Como o desafio global é atender ao objetivo de desenvolvimento sustentável n° 7, estabelecido pela Organização das Nações Unidas (ONU), que busca reduzir as emissões atmosféricas e garantir acesso à energia renovável e sustentável, a energia nuclear apresenta-se como uma ferramenta limpa de diversificação da matriz energética global. Porém, não deve ser vista como a única solução para os problemas energéticos de um país, devendo fazer parte de um programa diverso de produção energética junto com a produção e utilização de outras fontes renováveis de energia, como a eólica e a biomassa.
Recent research studies have investigated the use of high-strength materials in nuclear power plants to enhance the constructability of their massive squat-reinforced concrete shear walls. Despite the advantages of these high-strength materials, the dynamic response of such walls has not yet been fully investigated at the different damage states. To address this, the current study focuses on developing fragility functions for squat-reinforced concrete shear walls with high-strength materials to evaluate their seismic response compared to their counterparts with normal-strength materials. In this respect, a numerical OpenSees model, validated using previous experimental programs that have been conducted on reinforced concrete shear walls (i.e., with different aspect ratios, vertical and horizontal web reinforcement ratios, yield/ultimate strengths of reinforcement, and concrete compressive strengths), is used in the current study. Incremental dynamic analyses, using the 44 far-field ground motion records recommended by the FEMA P695 methodology, are then performed to develop fragility functions for reinforced concrete shear walls with normal- and high-strength materials at different damage states. These damage states are characterized by several performance indicators including cracking and crushing of concrete, residual displacements due to sliding, and reinforcement buckling/fracturing, following the FEMA P-58 guidelines. Finally, design recommendations are presented to enhance the seismic performance of squat-reinforced concrete shear walls when high-strength materials are adopted in nuclear construction practice.KeywordsSeismic responseNuclear-reinforced concreteShear walls
In view of the worldwide requirement for electricity, the limited possibilities offered by renewables, and the comparatively favorable costs of nuclear power, many countries regard nuclear power as the option for the future. The costs of nuclear power are analyzed and presented for the nuclear power plants of Leibstadt and Gösgen, both of them located in Switzerland. The specific roles of nuclear fuel and of decommissioning and waste management are considered, as are the total costs arising to society compared with those caused by other electricity production systems. The ongoing construction of 2 new nuclear reactors in Europe also provides reliable information about the costs of new future reactors. It is good to see that this favorable cost level is achieved despite safety measures and safety inspections, despite the fact that all costs of the government and public authorities are borne by the operators, despite provisions made for decommissioning and waste management.
Is nuclear power a thing of the past or a technology for the future? Has it become too expensive and dangerous, or is it still competitive and sufficiently safe? Should emerging countries invest in it? Can we trust calculations of the probability of a major nuclear accident? In the face of divergent claims and contradictory facts, this book provides an in-depth and balanced economic analysis of the main controversies surrounding nuclear power. Without taking sides, it helps readers gain a better understanding of the uncertainties surrounding the costs, hazards, regulation and politics of nuclear power. Written several years on from the Fukushima Daiichi nuclear disaster of 2011, this is an important resource for students, researchers, energy professionals and concerned citizens wanting to engage with the continuing debate on the future of nuclear power and its place in international energy policy.
This paper provides an econometric analysis of nuclear reactor construction costs in France and the United States based on overnight costs data. We build a simultaneous system of equations for overnight costs and construction time (lead-time) to control for endogeneity, using change in expected electricity demand as instrument. We argue that the construction of nuclear reactors can benefit from standardization gains through two channels. First, short term coordination benefits can arise when the diversity of nuclear reactors' designs under construction is low. Second, long term benefits can occur due to learning spillovers from past constructions of similar reactors. We find that construction costs benefit directly from learning spillovers but that these spillovers are only significant for nuclear models built by the same Architect–Engineer. In addition, we show that the standardization of nuclear reactors under construction has an indirect and positive effect on construction costs through a reduction in lead-time, the latter being one of the main drivers of construction costs. Conversely, we also explore the possibility of learning by searching and find that, contrary to other energy technologies, innovation leads to construction costs increases.