Capital Cost Estimation for Advanced Nuclear Power Plants

Stewart W.B.1, Shirvan K.1,*

1
Department of Nuclear Science and Engineering, Massachusetts Institute of Technology, USA

* Corresponding Author. Email Address: [email protected] (K. Shirvan)

Abstract
The first-of-a-kind (FOAK) nuclear plants built in the last 20 years are 2X over budget and schedule in the
US and some European countries. One of the nuclear industry’s proposed remedies is the small modular
reactor (SMR). SMR designs leverage five factors to be more economically competitive than large reactors:
1) multiple units; 2) increased factory production and learning; 3) reduced construction schedules; 4) plant
design simplification and 5) unit timing. There are currently no studies that quantitatively account for
these factors and compare different near term SMRs with Gen III+ large plants. This work presents a
nuclear plant cost estimating methodology using a detailed bottom up approach for over 200 structures,
systems, and components. The results compare relative cost for two large pressurized water reactors, one
with active safety and one with passive safety, to two SMR designs, one with multiple reactor power
modules and one with a single reactor module. Passive safety systems showed noticeable savings at both
the large and small scale reactors. The power uprating of a SMR by 20% resulted in ~15% savings in the
overnight unit capital cost. Overall, if built by an inexperience vendor and work force, the two SMRs’
overnight cost were higher than large reactors since significant on-site labor still remains while losing
economy of scale. However, the single-unit SMR had significantly less total person-hours of onsite labor,
and if built by an experienced workforce, its overnight construction cost showed potential to be
competitive and avoid cost-overrun risks associated with megaprojects.

Highlights:
• Bottom-up cost estimation of large and small nuclear plants are made.
• Typical SMRs require at least similar unit on-site labor hours vs. large reactors.
• Introduction of passive safety and power uprate are effective at reducing cost.
• FOAK overnight unit cost of SMRs can be noticeably higher than large reactors.
• Experienced vendor and work force could make SMRs competitive with large plants.

Keywords: Nuclear Energy, Cost Estimation, SMR, Megaprojects, Modular Construction

Word Count: 7,701

List of Abbreviation
BE - Best Experience
EEDB - Economic Energy Data Base
FOAK - First-Of-A-Kind
I2S-LWR - Integral Inherently Safe Light Water Reactor
IAEA - International Atomic Energy Agency
iPWR – Integral PWR
kWe - Kilo-Watts Electric
LASR - Large Active Safety Reactor
LPSR - Large Passive Safety Reactor
ME - Median Experience
MIT - Massachusetts Institute of Technology
MMNC - Multi-module Natural Circulation
MWe - Mega-Watts Electric
NC-SMR - Natural circulation SMR
NOAK - Nth-Of-A-Kind
NSSS - Nuclear Steam Supply System
O&M – Operation and Maintenance
PWR – Pressurized Water Reactor
PWR12 - Westinghouse Pressurized Water Reactor
RMS - Root Mean Square
ROK – Republic Of Korea
RPV - Reactor Pressure Vessel
SMR - Small Modular Reactor
SSCs - Structures Systems and Components

1 Introduction
Cost has been the primary issue facing the nuclear industry for over twenty years. When explaining the
relatively limited expansion of nuclear power, the recent MIT Future of Nuclear report stated, “the
fundamental problem is cost” [1]. An assessment by the Nuclear Energy Agency of FOAK nuclear plants
that began construction in the last fifteen years showed an average cost escalation and construction
delays of 2X the original estimate [2]. Lovering et al. showed that many countries experienced
reductions in specific capital costs in their early days of adoption through the economy of scale [3].
More recently, however, on average large nuclear plants have become untenable megaprojects
suffering from mismanagement and overly complex designs. The notable exceptions to this experience
were the first four Advanced Boiling Water Reactors built in Japan in the 1990s and 2000s which were all
completed in under four years and under-budget.

The proposed remedy is the small modular reactor (SMR) which aims to simplify and standardize nuclear
plants away from the megaproject scale seen today. Carelli et al. outlined a set of six factors for
estimating the cost competitiveness of SMRs: economy of scale, multiple units, learning, construction
schedule, unit timing, and plant design [4]. The loss of economy of scale escalates specific capital costs
for SMRs, but the other five factors may make up the difference. Kuznetsov pointed to similar factors:
reducing complexity, sizing the reactor for transportability, off-site refueling, multi-module sites, and
higher thermal efficiencies [5]. However, some studies question the cost reduction potential [6] and
technological advantages of SMRs [7].

Moreover, it is not trivial to quantify the magnitude of these factors for a specific nuclear plant design.
Carelli et al. estimated the value of their six factors for a four pack of the 335 MWe IRIS reactor concept:
economy of scale – 1.7, multiple units – 0.86, learning curve – 0.92, construction schedule & timing –
0.94 cumulative, and plant design – 0.83 [4]. In net, the generic SMR had a 5% higher capital cost than a
generic large reactor. Meanwhile, in the recent International Atomic Energy Agency (IAEA) handbook,
SMART SMR was listed at $10,000/kWe, 4x higher than the experience of APR1400 in Republic of Korea
(ROK) [8]. SMART, however, has only 100 MWe of capacity, so it pays a higher economy of scale penalty
than other SMR designs.

In this study, we reframed the six factors from Carelli et al. Modularization was separated from plant
design, as it is something that can be applied to systems within a plant design. Unit timing was rolled
into construction schedule, as also done by Carelli et al., because it is the shortened construction
schedule that allows for units to be delivered on pace with demand growth. The multiple unit factor
was folded into the plant design because some designs are multi-unit by design such as NuScale. Finally,
megaproject risk was added because it is a significant cost barrier for the industry today. Table 1
summarizes these factors, their components, and their impact as used in this study. These factors are all
highly correlated – they each can impact the cost of the same structure, system, or component in
different ways. This creates a challenge when trying to generalize their impact across multiple SMR
designs and highlights the need for a case-by-case quantification.
Table 1. Factors driving cost, the components used to estimate them, and how each factor impacts cost or value. *Construction
schedule was not considered as part of this study.

Cost Factor Components Impact and Value

Economy of scale Structures, components, labor Baseline cost
Learning Factory, site labor, and site material learning rates Future cost
Cost of building a factory, percentage of work moved Baseline cost, Construction
Modularization
offsite, offsite labor efficiency gain schedule
Active/passive safety, plant simplification (e.g. integral
Plant design PWR), new architectures (submerged containment Baseline cost
vessel)
Overnight cost, construction method (e.g. steel plate
Megaproject risk composites), total site labor hours, offsite work, Cost and schedule overrun
regulatory risk
Interest during construction,
Construction schedule* Economy of scale, modularization, plant design, learning
capital at risk

Mignacca et al., in a review of SMR economics, noted that “there is a lack of standardized approach in
the evaluation and economic performance of SMRs” [9]. There are several design specific cost estimates
but their methodologies vary. An AP1000 estimate by Black & Veatch was a point estimate, not a
detailed assessment of the structures, systems, and components[10]. A NuScale estimate by Scherrer
was closed and confidential [11], and an estimate by Black et al. cited proprietary vendor data [12] UxC
reported vendor cost estimates from NuScale, mPower, SMR-160 [13]. Zhang et al. showed that a two-
unit high temperature gas reactor SMR had a 5% higher cost than a single unit with a larger power
capacity due to plant simplicity and economy of multiples [15]. Maronati et al. estimated the cost of a
large reactor, the Integral Inherently Safe Light Water Reactor (I2S-LWR) using cost data based on a 1144
MWe pressurized water reactor in the Economic Energy Data Base (EEDB) [16]. They used commodity
quantities and design-specific details to scale the costs from the EEDB to I2S-LWR, and they expanded
the EEDB to account for new technology such as micro-channel heat exchangers and seismic isolation.
Most recently, Lloyd et al. conducted a study of nuclear costs showing that a high degree of
modularization could reduce capital costs for a 300 MWe SMR 45%, giving it a lower cost than a large
reactor [17].
The IAEA and Ganda et al. each published a methodology for nuclear plant cost estimation. The IAEA
feature cost scaling method uses nine plant design characteristics, such as thermal power, efficiency,
power density, number of fuel elements, etc., to estimate the capital cost [18]. However, this method
cannot estimate the cost of unconventional designs such as the NuScale’s containment vessel, so it is
limited to designs similar to the reference plant. Similar to Maronati et al., Ganda et al. estimated the
cost of the ABR1000 sodium cooled reactor by scaling reference nuclear plant cost data from the EEDB
[19] using over 30 design features including volumes and surface area of structures, primary pump flow
rates, pressure vessel mass, waste production rates and others [20]. Ganda et al. included labor and
material cost uncertainty in an update to their method [21]. Neither model estimates the impact of
learning rates, modularization, or indirect costs. The OPEN100 estimated the cost of their open-source
100 MWe PWR by linearly scaling the highest level EEDB costs and applying cost reductions from other
industries [22]. Champlin estimated the cost of applying advanced construction techniques such as steel
plate composites to containments and shield buildings [23].

In this work, we present a nuclear capital cost estimation methodology that is based on the EEDB [19], is
applicable to all light water reactor designs, and establishes a consistent set of assumptions to quantify
deriving cost factors for the selected designs. Then, we use the tool to estimate the capital costs of large
LWRs, based on publicly available information for AP1000 and APR1400 and small modular reactors,
based on public information available for NuScale and SMR-160 plant designs. This work distinguishes
itself from its predecessors in its detailed bottom-up approach using best engineering based judgment
and practices to capture the intrinsic features of SMR designs, including factory production, learning
rates, passive safety, simplified electrical and piping while quantifying the lack of economy of scale in a
holistic framework.

2 Methodology
Using assumed component cost scaling relationships collected from a range of sources [21] [24] [25], we
use the EEDB as a baseline to estimate costs. The EEDB is a detailed, component-by-component cost
breakdown of over 1400 cost items for a four loop 1200 MWe Westinghouse pressurized water reactor
(PWR12). For each cost item, the EEDB provided the factory cost, site labor cost, site labor hours,
material quantity, and material cost. Costs are organized into a code of accounts system, with one-digit
accounts summing their respective two-digit accounts and so forth up to ten digits of detail. Table 3 is a
summary of the one- and two-digit accounts. The database contains costs for median experience plant
builds (PWR12-ME) and better experience plant builds (PWR12-BE), and the PWR12-ME capital cost was
more than double the PWR12-BE capital cost. In this analysis, the PWR12-ME was taken to be
representative of FOAK cost, and therefore, the FOAK estimates in this study include an implicit
contingency (defined narrowly as allocated risk in the predicted overnight construction cost) based on
the median experience plants built. In this way, contingency costs were distributed across the relevant
cost accounts as indicated by the difference between the PWR12-BE and PWR12-ME three-digit
accounts. This is an important note as typical publicly published cost data by vendors are NOAK cost
estimates. In the following sections we present: methods for inflating the EEDB from 1987 to 2018 for
scaling each component cost, for accounting for modularization and learning to reach NOAK cost, for
design specific adjustments, and for estimating indirect costs.

The EEDB combined all the factory nuclear steam supply system (NSSS) costs into a single account, 220.
This account included the costs of the reactor pressure vessel, steam generator, primary coolant piping,
pressurizer, and control rods. Ganda performed an extensive cost analysis dividing the single factory
costs among these specific components [20]. This work adopted Ganda’s breakdown of NSSS factory
costs.

To bound the range of SSCs necessary to model within the cost tool, this study analyzed four different
nuclear plant designs beyond PWR12. The first was a large (>1000 MWe) Gen III+ PWR with passive
safety systems similar to the Westinghouse AP1000 design. The second was a large Gen III+ PWR with
active safety systems similar to the KEPCO APR1400. The third was a multi-module, small modular,
natural circulation, integral PWR similar to NuScale. The fourth was small modular, natural circulation,
integral PWR similar to the Holtec SMR-160. It should be noted that the Holtec SMR-160 design is the
only one that has not gone through design certification and its design specifications will likely evolve
more relative to the other reference designs.
Table 2. This study considered four nuclear plants beyond PWR12: a Gen III+ passively safe PWR – LPSR, a Gen III+ actively safe
PWR LASR, a multi-module small modular reactor – MMNC, and a single module small modular reactor – NC-SMR.
Large Passive Large Active Multi-module Natural
Name: PWR12 Safety PWR Safety PWR Natural Circulation circulation SMR
(LPSR) (LASR) iPWR (MMNC) (NC-SMR)
Westinghouse 4- Westinghouse KEPCO APR1400 NuScale Holtec
Similar to:
Loop PWR AP1000 SMR-160
Net Power (MWe) 1144 1117 1400 12 x 57* (685 total) 160
*Note in section 3.3 we will explore the recently announced 77 MWe reactor module

The methodology of our model was a six-step process graphically explained in Figure 1. The EEDB was
reduced to 204 SSCs known to be cost drivers and design differentiators. The cost estimation tool: 1)
scaled each cost based on the details of the plant, 2) made adjustments for novel components, 3)
reduced costs for known simplifications, 4) modularized the applicable SSCs, 5) applied learning rates,
and 6) estimated the indirect costs.

Figure 1. Graphical representation of the steps in the cost estimation process and the corresponding section of the article that
discusses the details behind the step.

2.1 Inflation and cost escalation

The DOE published the most recent version of the EEDB in 1987, so the factory, labor, and material costs
need to be inflated and escalated to the 2018 cost basis chosen for this work. Each labor and material
cost were inflated using the appropriate index from the Bureau of Labor Statistics [26]. For example, the
pipe fitter index applied to labor costs for piping, the millwright index applied to labor costs for
equipment installation, the wire and cable index applied to material costs for electrical work, etc.
Factory costs were escalated with the most similar material cost index. This approach properly
accounted for varying inflation rates and avoided the five year limit on inflating heavy industry plant
costs [27].

Above inflation, Ganda et al. and the Electric Power Research Institute applied nuclear escalation rates
of 30% and 70% respectively [20] [28]. The nuclear escalation rates account for changes in regulation
and losses in nuclear project construction management expertise. These escalation rates applied
uniformly to all capital costs, so we did not apply them in this work. Therefore, given other uncertainties,
the results in this study should be considered relative to one another and not in an absolute sense. Actual
costs of a future project may vary with changes to environmental factors left unquantified here.
Regulation may change and either increase or reduce actual project cost of a FOAK plant. The location of
a project has a similar effect, as the environmental factors differ among countries. Table 3 provides the
one- and two-digit account summary results for this inflation. It should be noted that the assumed
inflation is only applicable to US project cost estimates.
Table 3. One- and two-digit accounts for the PWR12-ME inflated to 2018 ($/kWe)
Account Account Description Factory Cost Labor Cost Material Cost Total Cost
A.1 Total Base Cost 2,760 1,349 591 4,700
A.2 Total Direct Costs 997 1,006 295 2,299
A.21 Structures and Improvements 45 391 176 612
A.22 Reactor Plant Equipment 440 204 38 682
A.23 Turbine Plant Equipment 348 148 22 519
A.24 Electric Plant Equipment 63 131 38 233
A.25 Miscellaneous Plant Equipment 39 86 13 138
A.26 Main Condenser Heat Reject System 61 46 8 115
A.9 Total Indirect Costs 1,763 343 296 2,402
A.91 Construction Services 203 306 258 766
A.92 Engineering & Home Office Service 841 0 0 841
A.93 Field Supervision & Field Office Service 719 38 38 794

2.2 Direct cost estimation methods

The developed cost estimate tool utilizes five methods for scaling and estimating 204 separate direct
SSC costs from the EEDB PWR12-ME for a new plant. A full list of these costs can be found in the
supplementary information, the scaling method used, and the parameter values can be found in the
supplementary information. As an indication of the level of detail, the median value of these 204 SSC
costs was $1.6M for PWR12-ME. This first method came from Towler et al. for systems and equipment
costs and follows Equation 1:

Equation 1

𝐶 = 𝐴 ∗ [𝐵 + 𝐷 ∗ 𝑃𝑛 ]
Where C is the cost; A is the nuclear cost escalation factor; B, D, and n are constants from Towler et al.
inflated to 2018 using the Chemical Engineering Plant Cost Index, and P is a parameter describing the
component [24]. This method applied to twelve costs including the reactor pressure vessel (RPV),
primary pumps, steam generator, pressurizer, cranes, and the NuScale containment vessel. The
parameter, P, was either the vessel mass, steam generator heat transfer area, pump flow rate, or crane
capacity. The nuclear cost escalation factor came from escalating the estimate from Towler to match the
reported value for PWR12-ME. The difference between Towler and PWR12-ME ranged from 13-20X.

The second method was applied to 131 SSC costs and used the parameter, P, and exponent, n,
recommended by Towler in an exponentially scaled relationship to the EEDB cost:

Equation 2

𝑃𝑛𝑒𝑤 𝑛
𝐶 = 𝐶𝐸𝐸𝐷𝐵 ∗ ( )
𝑃𝐸𝐸𝐷𝐵
This comparative method mostly applied to structures and buildings, so the parameter, P, was the
volume of the foundation, surface area of the containment, or other structures-based values. This
method also applied to service air, water, and steam systems that scaled on building volumes, control
rod drive systems that scaled on the number of control rods, and waste processing systems that scaled
on the primary coolant flow rate.

The third method used Equation 2 but either the plant level thermal or electrical power (in the case of
multi-unit sites). 35 SSC costs scaled with this method including turbine equipment, electrical
equipment, and the safety grade HVAC systems in the reactor and auxiliary buildings. This method also
applied in cases where the SSC parameter was unknown for less developed plant designs. Using the
plant level thermal power allowed for cost savings in multi-unit designs like NuScale where several
reactors will share systems and equipment.

The fourth method was a fixed cost at the PWR12-ME cost. This applied to costs such as the security
building, fire house, technical support center, some monitoring systems, and waste water treatment
systems. The final method was simply a direct cost input, so that the users could apply their own known
cost. This method was not used in any of the estimates in this study.

2.3 Design specific adjustments

2.3.1 Containments
The AP1000 has a fundamentally different containment than the PWR12: it uses a standalone steel
structure surrounded by a reinforced concrete shield building instead of a steel lined reinforced
concrete shell. The EEDB estimated the cost for a standalone steel containment structure for the
APWR6, a 600 MWe advanced-PWR, and the cost for a steel lined reinforced concrete shell for the
PWR6, a 600 MWe PWR. The standalone steel structure cost was 2X the steel lined concrete shell cost
per unit area due to higher price of procuring and installing steel vs. equivalent steel-lined concrete
structures. As such, this analysis accounts for this escalation.

The NuScale containment is more similar to a reactor pressure vessel than a site-built structure. As a
result, its cost estimation used Equation 1 for the reactor pressure vessel, but scaled with the mass of
the containment. The NuScale containment is stainless steel whereas the PWR12-ME RPV was stainless
steel cladded carbon steel, and Towler et al. reported a 2.3X cost escalation for stainless steel vessels
over carbon steel vessels. This is the assumed methodology used for this study. It is noted that if the
AP1000 stainless steel containment cost is scaled based on its weight and directly applied to NuScale
containment, then the cost would be ~10% lower that the assumed methodology.
2.3.2 Design simplifications
The AP1000, and many subsequent designs, applied substantial simplifications to the reactor plant
equipment piping and valve systems. Some of these savings came from improved design, and some
came from passive safety features that reduced the number of redundant systems. The EEDB made an
estimate of the savings for these simplifications, and we used the APWR6/PWR6 ratio to approximate
the savings. Note that the APWR6 was an early conceptual design not identical to the AP600 that
ultimately became the AP1000, but these estimated cost savings were taken to be representative of
what simplifications were possible from Gen III designs. Table 4 summarizes the relative cost reductions
across Account 22 which reports the plant simplifications we applied to the plant models in this study. In
contrast to the LPSR, the cost estimate for the LASR did not include cost reductions for account 223 due
to non-existence of passive safety systems. However, the cost reductions for the other accounts in
Table 4 did apply to LASR based on the available information on APR-1400.
Table 4. APWR6/PWR6 cost ratio for reactor plant equipment design simplification
Account & 222.12: Reactor 223: Safeguards 224: Radwaste 225: Fuel Handling 226: Other Reactor
Description Coolant Piping system Processing & Storage Equipment
Ratio 0.25 0.71 0.76 0.52 0.50

The AP1000 had many structural simplifications in Account 21. We used published drawings of the
AP1000 plant to estimate the dimensions of each structure for our estimate of the LPSR. The costs were
then scaled based on those dimensions and the dimensions of the PWR12 plant. Many buildings were
eliminated in the AP1000, for example: emergency feedwater pump building, main steam and feedwater
pipe enclosures, and others, and the cost of these buildings were eliminated for the AP1000 model. The
APR1400 did not have as many structural simplifications as the AP1000, and this was reflected in the
estimate for the LASR.

Similarly, passive safety and digital I&C should result in dramatic reductions in Account 24: Electrical
Plant Equipment. The implemented passive safety eliminated the need for Class 1E emergency diesel
generators and certain redundancies, and digital I&C led to simplified wiring installations. The EEDB
made an estimate of these savings for both of these cost reductions – again using the ratio of
APWR6/PWR6. Table 5 summarizes the relative cost reductions across account 24.
Table 5. APWR6/PWR6 cost ratio for electrical plant equipment simplifications from digital I&C and passive safety.
Account & 241: 242: Station Service 243: 244: Protective 245: Electrical Structures &
Description Switchgear Equipment Switchboards Equipment Wiring Containter
Ratio 0.55 0.27 0.91 1.01 0.58

NuScale had a further set of cost reductions: no primary coolant pump or piping, integrated pressurizer,
and integrated steam generator. The PWR12 steam generator mass was 300 tons total, half of which
was the vessel. Ganda scaled steam generator costs by mass [20], so eliminating the steam generator
vessel, as in the case of an integral PWR, reduced the steam generator cost by 50%. Similarly, for the
pressurizer, as noted by Chen et al. [29], the mass of the pressurizer vessel is the dominant cost relative
to the spray nozzle, relief valve, and heater, so the model assumed a 90% cost reduction. The Electric
Power Research Institute has been developing advanced fabrication methods for SMR RPVs that aim to
reduce RPV costs by 40% [30]. Our model assumed that this technology will exist at the time of
commercialization for NuScale and SMR-160. These assumed cost reductions applied to both the MMNC
and the NC-SMR cost estimates and are summarized in Table 6. The reactor building pool cost for the
MMNC scaled using its surface area and the surface area of the spent fuel pool in PWR12-ME.
Table 6. Cost reductions by code of account for integral PWRs
Account & 222.11 Primary 222.12 Primary 222.13 Steam 221.1 Reactor
222.14 Pressurizer
Description coolant pump system piping generators Pressure Vessel
Cost Reduction 100% 100% 50% 90% 40%

2.4 Learning
By producing a greater number of smaller reactors both factory producers and onsite construction
teams learn how to manufacture, construct and install equipment and buildings in a progressively more
cost-efficient way. Factory work can be continually refined with no theoretical limit on the cost
reduction, so the standard learning rate is modeled as:
Equation 3

𝐶𝑁
= 𝑁 log2 (1−𝑟%)
𝐶0
Where CN is the cost of the Nth component, C0 is the cost of the first component, and r is the learning
rate. In Equation 3, every time production doubles, the cost has reduced by the fraction r. Given the lack
of accessible information on learning rates for nuclear plant components, using cost data aggregated by
Duffey across gas turbines, wind turbines, small airplanes, and other systems, a least-squares fit of
Equation 3 yielded a 16% learning rate [31]. This cost learning curve applied to the factory costs for new
modularized components in our analysis such as the integral reactor primary system of NuScale.

Historical nuclear plant data does not have similarly high and boundless learning. Figure 2a shows cost
data from Lovering et al. for the French P4 and the Korean OPR-1000, both of which had a clear
asymptotic trend in reductions after the 20% in total cost reduction [3]. Lovering et al. stated regulatory
stability (which was the case for France and ROK) would be key to inferring real learning rates, hence
selecting just these two series. Moving from site to site, new workers execute the construction tasks,
and while some learning has been retained by the vendor, there is a limit to the value of that learning.
Therefore, we hypothesized there would be different learning rates among the factory, labor, and
material cost categories.

To divide the learning rate among the factory, labor, and material costs, we used the ratio of direct costs
for the PWR12-BE to the PWR12-ME shown in Table 7. Only direct costs are shown here because we
address indirect cost estimation in a following section. Table 7 shows Labor hours and cost experienced
the greatest cost reduction, followed by Material costs, then Factory costs. Factory cost learning was
likely limited as many of these components had already reached substantial production quantities, and
thus N in Equation 3 was already very high. Therefore, in our modeling, N started at 100 for Factory
costs. Although the data from Duffey indicated no bound for Factory cost reductions, we constrained
the Labor and Material cost reductions according to the PWR12-ME to PWR12-BE savings.

Figure 2b shows a fit of Equation 3 minimizing the RMS error with different learning rates for factory,
labor, and material costs to the average of the OPR-1000 and P4 series data. The Factory cost learning
rate was 16% from Duffey, and the fit yielded 13.1% and 7.1% learning rates for the Labor and Material
costs respectively. As Labor and Material costs are highly correlated, the Labor and Material rates were
constrained to reach their thresholds at the same plant number. Table 7 provides a summary of the
resulting model used in the remainder of the analysis. These learning rates were inferred from plants
that were mostly built at different sites, so the learning rate may be higher for multi-plant sites.
Aggregating the effect across the three learning rates, and assuming no new equipment (N0=100 for
factory costs), the average learning rate of our model was 7%. These rates were consistent with what
others in the field have estimated. Lyons et al. reported a range from 2 to 15% [32], and a University of
Chicago study of nuclear economics reported a range from 3 to 10% [33].

Our model also accounted for intra-plant learning. For example, in the case of the NuScale plant, there
are 12 reactor power modules, so it is unlikely the factory, labor, and material costs will be consistent
from fabricating and installing the first to the last module. As a consequence, the learning model in
Table 7 accounted for intra-plant multiples. Nevertheless, due to lack of available representative data,
the assumed learning is driven from plant level data and applied to individual SSCs. This dilutes the
bottom-up nature of our tool and in future work, bottom-up learning models for on-site construction
should be developed upon availability of data.

1.05 1.05
OPR-1000 P4, OPR-1000 Average
1.00 1.00
P4 Series EEDB Based Aggregate Model
0.95 0.95
Cost/FOAK

0.90 0.90
Cost/FOAK

0.85 0.85

0.80 0.80

0.75 0.75

0.70 0.70

0.65 0.65
0 5 10 15 20 0 5 10 15 20
Reactor Plant Number Reactor Plant Number
(a) (b)
Figure 2. a) Cost data for the French P4 and Korean OPR-1000 series reactors from Lovering et al. [3]; b) Average of the P4 and
OPR-1000 costs and the EEDB based learning model aggregating factory, labor, and material cost learning rates and thresholds.
The parameters for this model are in Table 7.

Table 7. Ratio of PWR12-BE to PWR-ME direct costs showing the total savings from FOAK to NOAK
Factory Cost Labor Hours Labor Cost Material Cost Total Cost
𝑃𝑊𝑅12 𝐵𝐸
97% 55% 55% 73% 76%
𝑃𝑊𝑅12 𝑀𝐸
N0 100 1 1 1 -
Learning rate, r 16% 13.1% 13.1% 7.1% -
CN/C0 Limit 0 55% 55% 73% -

2.5 Modularity
Construction time and plant design were two of the IAEA six cost factors, and modularization of SSCs in
SMR designs addresses these two factors. SMR designers intend to aggregate subsystems to form
modules, and modules can be manufactured in a factory, shipped to the site, and installed as a whole.
With this construction scheme, modularity reduces the onsite construction labor required for
installation. We quantified the impact of this process and integrated it within the EEDB cost scaling
framework.

The global consulting firm McKinsey reported that modular construction does not yield significant cost
savings, but it does reduce construction time [34]. McKinsey estimated that modular construction
reduces onsite labor 25-65% by shifting 25% of it offsite and leveraging productivity gains in a factory.
The reported site labor savings were clear, but the total cost savings, including construction of the
factory to produce the modules, yielded cost changes from -20% to +10%. In this analysis, for
modularized accounts, 50% of labor costs move to factory costs at twice the productivity, reducing the
site labor costs and increasing the factory equipment costs. These assumptions were very similar to
those recommended by the Economic Modeling Working Group for advanced nuclear reactors [25] and
those used by Lyons et al. [32]. In future if data on cost savings from modular construction on individual
SSC is available, more granular modular construction model (i.e., bottom-up) could be applied.

For NuScale, the following systems were modular: turbine generator plant equipment, reactor
equipment, the main heat transport system, reactor instrumentation and control, and the containment
vessel. For SMR-160, the same systems were modularized except the containment vessel.

2.6 Indirect costs

The EEDB breaks down indirect costs into three categories: Construction Services, Engineering & Home
Office Services, and Field Supervision & Field Office Services. We extracted correlations between direct
and indirect costs using the data for the PWR12-BE and PWR12-ME. Similar to the direct costs, the EEDB
divided the indirect costs into Factory Equipment Cost, Site Labor Hours, Site Labor Cost, and Site
Material Cost. Table 8 shows scaling relationships deduced from the reported PWR12-ME and PWR12-
BE data that proved to be accurate to +/-10%. In practice, the direct site labor cost, labor hours, or
material cost of the new reactor concept was multiplied by the scaling value to estimate the indirect
cost. The indirect material costs were escalated with the ratio of average workers to account for tools
and equipment needed for additional workers. The indirect factory costs accounted for home and field
offices, taxes, and insurance, so they were escalated with the ratio of construction time as estimated by
the vendor.
Table 8. Indirect cost modeling assumptions and relations based on PWR12-ME and PWR12-BE.
Base Scaling
Base Scaling Relation Escalation Relation
Value
𝑃𝑊𝑅12 𝐵𝐸 𝐼𝑛𝑑𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝐿𝑎𝑏𝑜𝑟 𝐶𝑜𝑠𝑡
Site Labor Cost 36% -
𝑃𝑊𝑅12 𝐵𝐸 𝐷𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝐿𝑎𝑏𝑜𝑟 𝐶𝑜𝑠𝑡
Site Labor 𝑃𝑊𝑅12 𝐵𝐸 𝐼𝑛𝑑𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝐿𝑎𝑏𝑜𝑟 𝐻𝑜𝑢𝑟𝑠
36% -
Hours 𝑃𝑊𝑅12 𝐵𝐸 𝐷𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝐿𝑎𝑏𝑜𝑟 𝐻𝑜𝑢𝑟𝑠
Site Material 𝑃𝑊𝑅12 𝐵𝐸 𝐼𝑛𝑑𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝑀𝑎𝑡𝑒𝑟𝑖𝑎𝑙 𝐶𝑜𝑠𝑡 𝑁𝑒𝑤 𝑃𝑙𝑎𝑛𝑡 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 # 𝑊𝑜𝑟𝑘𝑒𝑟𝑠
79%
Cost 𝑃𝑊𝑅12 𝐵𝐸 𝐷𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝑀𝑎𝑡𝑒𝑟𝑖𝑎𝑙 𝐶𝑜𝑠𝑡 𝑃𝑊𝑅12 𝐵𝐸 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 # 𝑊𝑜𝑟𝑘𝑒𝑟𝑠
Factory Equip. 𝑃𝑊𝑅12 𝐵𝐸 𝐼𝑛𝑑𝑖𝑟𝑒𝑐𝑡 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝐶𝑜𝑠𝑡 𝑁𝑒𝑤 𝑃𝑙𝑎𝑛𝑡 𝐶𝑜𝑛𝑠𝑡𝑟𝑢𝑐𝑡𝑖𝑜𝑛 𝑇𝑖𝑚𝑒
132%
Cost 𝑃𝑊𝑅12 𝐵𝐸 𝐷𝑖𝑟𝑒𝑐𝑡 𝑆𝑖𝑡𝑒 𝐿𝑎𝑏𝑜𝑟 𝐶𝑜𝑠𝑡 𝑃𝑊𝑅12 𝐵𝐸 𝐶𝑜𝑛𝑠𝑡𝑟𝑢𝑐𝑡𝑖𝑜𝑛 𝑇𝑖𝑚𝑒
3 Results
3.1 Benchmarking the model and cost estimates
In this subsection, the data on total concrete volume and reported cost estimates from the referenced
plants (if available) are compared with the developed generic models in this study. For the cost
estimates, the vendor and utility estimates that are based on proprietary data on expected overnight
cost of the plants are included. In some cases, if the project did not realize unforeseen delays, the final
realized cost is also compared. It should be noted that it is generally unclear what is included in the
noted cost estimates (e.g. total project cost vs. overnight cost) and the benchmark should be treated as
indicative that the model appropriately captures trends but not as model validation.

In scaling the 204 direct SSC costs, the cost estimation tool calculated the total safety grade concrete to
make pieces of the cost estimate, and these data are captured in Table 9. The large reactor building in
the MMNC plant requires the most safety grade concrete while the NC-SMR required the least.
Significant design effort was put into reducing the size and count of safety grade buildings for the LPSR,
and as a result, the LPSR had the lowest concrete per MWe capacity required. To benchmark these
estimates, Table 9 compares the calculated safety grade concrete to estimates of total concrete from
the recent MIT Future of Nuclear study. Assuming safety grade concrete is 60% of total concrete, the
estimates from our mode are within 10% of the estimates from the MIT Future of Nuclear study.
Table 9. Safety grade concrete for each plant as calculated inside the model compared to an estimate of the total concrete
(safety and non-safety) from [1].
PWR12-ME LPSR LASR MMNC NC-SMR
Safety Grade Concrete (m3) 84,017 46,892 195,402 109,354 25,117
Safety Grade Concrete (m3/MWe) 73 42 140 160 157
Total Concrete (m3/MWe) 122 71 237 229 184
Reference Total Concrete (m3/MWe) [1] 110 60 - 260 -
(PWR12) (AP1000) (NuScale)

The FOAK costs for the reactor concepts are shown in Figure 3a. The FOAK cost estimate for the LPSR
was $4,328/kWe which was within 1% of the original cost estimate for the AP1000s at the Vogtle site in
the US [2]. However, the realized cost of the Vogtle plant was estimated to be $8,600/kWe [2]. Merrow,
in a RAND Corporation report that regressed an extensive database of megaprojects, estimated cost
escalations of 59% if a project uses a new construction method and 78% if a project encounters
regulatory challenges [35]. For example, Westinghouse used steel plate composite for constructing the
AP1000 shield buildings at the Vogtle site which was the first use of this construction method for nuclear
projects in the US. Second, Westinghouse encountered design updates to accommodate regulatory
changes around requirements such as for airplane impact to the containment. Both of these factors
were likely drivers of the 100% cost escalation observed at the Vogtle site not accounted for in the
current cost model. The AP1000 plants at Sanmen 1 & 2 had lower capital costs due to the lower cost of
labor and material in China compared to the US. In China, labor rates are 5-18% of US labor rates [1],
and construction commodity prices are up to 25% lower [36]. Assuming that 30-50% of factory costs are
labor costs for nuclear components [29], with these lower labor and commodity costs in China, the cost
model estimates a range of $1,700-2,300/kWe for Sanmen 1 & 2 compared to the initial announced
budget of $2,044/kWe [2]. Also, the cost escalation at Sanmen was less than the escalation at Vogtle,
likely because Sanmen did not use steel plate composites or experience as many regulatory hurdles.
In Table 10, the LASR cost estimates were averaged over the NOAK to fourth-OAK plants because the
available costs for the Barakah site in the UAE are from a contract price for four units. The LASR cost
estimate was 22% higher than the realized cost for the four-unit APR1400 plant at Barakah in the UAE.
As in the case of Sanmen, however, the lower realized cost is likely the result of lower labor and material
costs in the UAE. Dorfman reported KEPCO’s bid was about 30% lower than the next cheapest bid [37]
indicating they may have underbid the Barakah site contract to help establish its first international
nuclear construction business. Finally, the APR1400 had an advantage over the AP-1000 or EPR in that it
was a design evolution from the OPR-1000, and KEPCO had essentially no gap in nuclear construction
between the last OPR-1000 and the first APR1400. This is also true for the mentioned positive
experience for the Advanced Boiling Water Reactors in Japan.

The MMNC FOAK overnight cost estimate was $6,554/kWe which was 37% higher than the PWR12-ME
and 51% higher than the AP1000. At a meeting for the Carbon Free Power Project for the Utah
Associated Municipal Power Systems, a cost estimate of $6B was disclosed [38]. However, this cost is for
the total project cost and the overnight cost is only a component of this estimate. Meanwhile, World
Nuclear News reported that the NOAK NuScale 720 MWe plant cost is $3,600/kWe which was within 7%
of our 10th-OAK MMNC cost estimate of $3,856/kWe. Black et al. estimated the overnight capital cost of
the NuScale plant to be $3,466 which was also near our 10th-OAK estimate, but they did not report
whether their estimate was FOAK or NOAK. In a relative context, the direct costs between the PWR12
and NuScale were essentially equivalent in their study, whereas we estimated NuScale direct costs to be
60% higher than the PWR12 direct costs. The primary drivers of these differences were Account 21:
Structures and Improvements and Account 23: Turbine Plant Equipment. Account 21 costs for NuScale
were more than double than those for PWR12 in our model, consistent with the NuScale plant requiring
the most amount of concrete of any nuclear plant concept [1]. The significant low of economy of scale
for Account 23 drove NuScale costs much higher than for the PWR12 even accounting for on-site, intra-
plant learning rates. In summary, while the main motivation behind the development of the cost
estimating tool is to derive relative insights into SMR vs. Large LWR cost drivers, the tool still
demonstrates its ability to reliably estimate overnight cost of nuclear projects.
Table 10. Capital cost benchmarking between the estimates produced in this work and publicly available dota
Cost Estimate Similar reference cost estimates Similar realized costs
($/kWe) ($/kWe) ($/kWe)
Large Passive Safety 4,328 4,300 [2] 8,600 [2]
PWR (LPSR) (FOAK - US) (AP1000 - Vogtle) (Vogtle)
1,700-2,200 2,044 [2] 3,154 [2]
(FOAK – China) (AP1000 – Sanmen) (Sanmen)
Large Active Safety 5,337 4,358 [39]
PWR (LASR) (First 4-unit average) (APR1400 4-unit average - Barakah)
Multi-module SMR 3,856 3,600 [40]
(MMNC) (10th-OAK) (NuScale, NOAK)

3.2 FOAK and 10th-OAK Cost Estimates

Figure 4 breaks down the direct costs for the five reactors into the six accounts from Table 3. The
Structures and Improvements account trends well with the safety grade concrete estimates from Table
9 except for a few cases. The NC-SMR uses less safety grade concrete per unit capacity than the MMNC,
and it is only marginally higher than the LASR. This was likely the result of the loss of economy of scale
for the small 160 MWe plant – shrinking the building does not allow for the wall and foundation
thicknesses to shrink because local effects drive the structural design constraints. The LPSR design uses
40% less safety grade concrete than the PWR12-ME, but the Structures and Improvements account was
only 5% less. This was the result of the expensive passively safe standalone steel containment structure.
This type of containment was significantly more expensive than the standard steel lined concrete
structure.

Within the Reactor Plant Equipment account, the impact of different design decisions can be seen. In
the LPSR, the use of passive safety and a simplified design reduced costs 14% from the PWR12-ME, but
the LASR reduced the cost of this account 5% from the PWR12-ME with a larger economy of scale. The
MMNC had an escalated Reactor Plant Equipment cost due the large number of small-scale vessels,
steam generators, and other reactor equipment. The MMNC did benefit from intra-plant learning for the
12 modules. Absent this effect, the MMNC Reactor Plant Equipment costs would have been 21% higher.
Further, the estimate cost reduction by e-beam welding the vessel as proposed by EPRI saved
$279/kWe.

Within Turbine Plant Equipment and Electrical Plant Equipment, there was strong favorability to the
large plants over the NC-SMR. In the MMNC plant, each reactor module has its own turbine, so these
turbine costs should be high due to the loss of economy of scale. However, intra-plant learning and
modularization led to 35% cost savings for MMNC Turbine Plant Equipment. The passively safe designs,
(LPSR, MMNC, and NC-SMR) all benefited from a 43% cost reduction in Electrical Plant Equipment. The
turbine plant equipment cost conclusions should not be treated as universal because a vendor can
choose to greatly simplify the turbine equipment by balancing system complexities and thermodynamic
efficiencies. For example, the number of feedwater heaters can be reduced resulting in lower system
costs but also lower efficiencies. The analysis here assumed a similar level of complexity was maintained
as the turbine size was rescaled except in the cases where the turbine plant was modularized. In these
cases, the 50% of site labor costs were shifted to the factory cost consistent with our modularization
methodology, and these modules were given higher learning rates.

Figure 2b shows learning progression for the overnight costs from the FOAK to the 10th-OAK plant for
each of the five reactors. The MMNC and NC-SMR, with significant modularization across many
categories experienced an average of a 36% cost reduction. The PWR12-ME, LPSR, and LASR with large
structures and heavy dependence on onsite labor experienced a cost reduction of 31% In absolute
numbers, PWR12-ME, LPSR, and LASR experienced roughly $1,500/kWe in cost savings across the first
ten plants. In contrast, the more heavily modularized plants, MMNC and NC-SMR, experienced roughly
$2,500/kWe in cost reductions. However, these larger cost reductions did not ultimately result in less
expensive plants than either the LPSR or PWR12-ME.
 FOAK Overnight Cost ($/kWe) 8,000
8,000 PWR12
PWR12 7,000 LPSR

Overnight Cost ($/kWe)

7,000 LASR
LPSR
LASR MMNC
6,000 6,000
MMNC NC-SMR
5,000 NC-SMR 5,000
4,000
4,000
3,000

2,000 3,000
1,000
2,000
- 1 2 3 4 5 6 7 8 9 10
Total Cost Direct Costs Indirect Costs Reactor Plant Number
(a) (b)
Figure 3. (a): Total, direct, and indirect cost estimates for the FOAK; (b): Total overnight cost estimates from FOAK to 10th-OAK
showing the rate of cost reduction.

FOAK Overnight Cost ($/kWe)

1,600
PWR12
1,400
LPSR
1,200
LASR
1,000 MMNC
800 NC-SMR
600
400
200
-
Structures and Reactor Plant Turbine Plant Electric Plant Miscellaneous Main Condenser
Improvements Equipment Equipment Equipment Plant Equipment Heat Reject
System
Figure 4. Direct cost estimates broken down into six main accounts.

Passive safety systems in the MMNC plant saved a total of $398/kWe: $125/kWe in electrical plant
equipment, $68/kWe in safeguards systems, and $205/kWe in associated indirect costs. In comparison,
just the liner for the reactor building pool cost $299/kWe, half direct and half indirect cost. So, the
passive safety feature of the MMNC saved in net $100/kWe not including the added cost to the reactor
building to support the mass of the pool, the cost of the new refueling scheme, or the construction risk
associated with such a large civil structure.

The NC-SMR specific costs were the highest among the five plant designs primarily due to the loss of the
economy of scale. The exception was that the reactor plant equipment costs were lower for the NC-SMR
than for the MMNC because the vessel style containment of the MMNC shifted this cost from structures
and improvements to the reactor plant equipment. Despite the high specific cost, the absolute cost of
the NC-SMR was only 25% of the other plants, making it a dramatically smaller construction project.

The LPSR, similar to the AP1000, used steel plate composites for the shield building around the
containment. Champlin provided a method for estimating the cost change with steel plate composite
construction [23]. Steel plate composites replace traditional rebar and formwork for concrete
superstructures. The material cost was 1.5X traditional rebar, but it eliminated formwork cost. There
was an added operating engineer cost for using a crane to install the plate composites. For high density
rebar, pressure containing structures, steel plate composites should reduce costs, but the LPSR shield
building is not a pressure containing structure. As a result, steel plate composites reduced labor costs
$4/kWe but raised material costs $9/kWe, for a net increase of $5/kWe or about $5M.

3.3 Variations on the MMNC Design

The previous results showed how the cost estimating tool was useful for comparing costs across very
different nuclear plant architectures. The cost estimating tool, however, can also estimate the value of
different design consideration within a particular architecture. For example, the licensed version of the
NuScale plant is a twelve module plant with 160 MWt units, but NuScale has been open about uprating
to a twelve module plant with 192 MWt modules, and more recently, NuScale discussed a six and twelve
module plants with 246 MWt modules [41]. The aim of each of these iterations is to meet the capacity
and cost needs of a particular customer. Here, we evaluated the cost impacts of these proposals on our
similar MMNC plant.

For these case studies, the number of modules affected the number of vessels, steam generators,
control rods, turbines etc., but the size of the reactor building and pool were constant. The number of
reactors and their thermal power affected the coolant flow rate through the core which in turn affected
the cost of the nuclear service water system, liquid waste system, and gas waste system. The plants with
lower number of modules accrued learning by doing savings more slowly. For the original MMNC plant,
the turbine building was in two parts with six turbines each, so for the six module plants, one turbine
building was eliminated. The overnight costs were estimated for four cases shown in Figure 5 and Figure
6.

These four cases showed an interesting balance of the economy of scale and total project cost. The first
uprate from 12 x 160 MWt to 12 x 192 MWt, lowered the cost-per-unit-capacity cost 14% while only
increasing the total plant cost 4%. Downsizing from 12 x 192 MWt to 6 x 192 MWt lowered the total
plant cost 29% but raised the cost-per-unit-capacity 43%. Scaling the reactor building and pool down to
match the six module plant yielded an additional 6% in cost reductions.

Similar to the first uprate, going from 6 x 192 MWt to 6 x 246 MWt lowered the cost-per-unit-capacity
19% while only increasing the total plant cost by 4%. In the first two cases, the Structures &
Improvements cost was identical, and the difference seen in Figure 6 was due to the plant capacity
change. Within the Reactor Plant Equipment cost account, the twelve 160 MWt module plant and six
246 MWt module plants had similar costs – showing an equilibrium point of learning (twelve module
plant) and power uprate (six module plant). These tradeoffs can help reactor designers and utilities
make more informed decisions, and this opens the door for automated design optimization for lowest
cost and/or best market capacity fit.
 10,000
12 x 160 MWt
9,000
12 x 192 MWt
8,000
6 x 192 MWt
Overnight Cost ($/kWe)

7,000 6 x 246 MWt

6,000
5,000
4,000
3,000
2,000
1,000
-
Total Base Total Direct Total Indirect
Cost Costs Costs

Figure 5.FOAK overnight costs normalized to the plant capacity for four iterations of the MMNC plant;

2,500
12 x 160 MWt
Overnight Cost ($/kWe)

2,000 12 x 192 MWt

6 x 192 MWt
1,500
6 x 246 MWt

1,000

500

-
Structures and Reactor Plant Turbine Plant Electric Plant Miscellaneous Main
Improvements Equipment Equipment Equipment Plant Condenser Heat
Equipment Reject System

Figure 6. FOAK overnight costs normalized to the plant capacity for four iterations of the MMNC plant for six cost categories.

4 Discussion
The remaining factor in the IAEA’s list, is “Unit Timing” which accounts for the ability of smaller nuclear
plants to cumulatively meet demand and lower the total capital at risk during construction. While Figure
2b showed the savings from the first unit to the tenth unit, that representation did not account for the
capacity of each unit. Thus, Figure 7 charts the cumulative costs versus cumulative installed capacity. In
Figure 6, the result of the cumulative cost analysis showed the LPSR was the lowest cost option,
consistent with the previous results, unless less than 1 GWe of total capacity is required. The smaller
size and learning rates allow for five NC-SMR plants to be built for the cost of the first LPSR plant.
Similarly, four NC-SMR plants can be built for less than the cost of the first MMNC plant. Even though
the FOAK NC-SMR plant was 20% higher cost per unit capacity than the MMNC, the cumulative costs
track in line in Figure 7. The LASR plant had a similar trend, where two MMNC plants or nine NC-SMRs
cost the same as one LASR. Further, the slow learning rate of the onsite labor heavy LASR design allowed
for both the NC-SMR and MMNC to have lower equivalent capacity cumulative costs after the second
LASR plant.

Carelli et al. showed smaller units such as the NC-SMR have lower total capital invested in any one year
and bring in revenue streams more quickly due to their shorter deployment schedules. As a result, they
will accumulate less interest during construction which will lower total capital costs, and they may
attract lower cost capital via lower interest rates.

16,000
PWR12
14,000 LPSR
LASR
12,000
Cumulative Cost ($Ms)

MMNC
10,000 NC-SMR

8,000

6,000

4,000

2,000

-
0 500 1000 1500 2000 2500 3000
Cumulative Capacity (MWe)
Figure 7. Cumulative costs for consecutive deployments of each reactor type. Each point is a new plant.

The cumulative capacity and cost trends also have risk-informed decision making implications. Although
difficult to quantify, megaprojects, which most nuclear projects are, are at a high risk of cost overruns
and delays. Many have studied this field, and there is disagreement as to the extent that project
characteristics or stakeholder's perspective drives the cost overruns associated with megaprojects [44].
However, in evaluating the megaproject-cost overrun risk associated with a particular nuclear design,
we focused on the design centric project characteristics because at this stage of analysis, these are
observable. Other factors such as optimism bias, stakeholder conflict, organizational complexity, etc. are
all strong cost overrun drivers [44] but cannot be evaluated at the conceptual design phase of a nuclear
plant. Further, some project characteristics are inherent to any nuclear project: interdependent systems,
multi-disciplinary teams, high degree of regulation. We do not discuss these aspects as they are not
differentiable among nuclear projects. A survey of the literature ( [45] [46] [47] [48] [35]), led to four key
factors that were correlated with cost escalations and construction delays: projects with total installed
cost greater than $1.5B, use of a new construction method, a long construction duration, a low use of
offsite labor. We translated these factors into a qualitative assessment of the five reactor designs
considered in this paper in Table 11.

Except for the NC-SMR, all designs had FOAK costs greater than $1.5B. Both the MMNC and LPSR
proposed using steel plate composites which is still a new construction method in the nuclear industry.
Onsite direct labor hours were high for all designs except the NC-SMR. 5.2 million person-hours of work
was still a large size project, but it was less than one third any of the other plants. The LASR had the
highest number of onsite labor hours required with over 35 million hours. In practice this design
architecture can work in low labor cost or high labor efficiency environments such as China and Korea. In
the US and the EU, however, high labor costs make this prohibitively expensive. Both MMNC and NC-
SMR used modularization to shift 29% and 22% of onsite labor to offsite facilities, respectively. After
considering modularization, however, MMNC and NC-SMR both still required substantially more onsite
labor per plant capacity (hours/kWe) than the PWR12 or LPSR. This result revealed SMR plants are not
necessarily more factory produced than traditional large PWRs.

The NC-SMR was the superior design from a cumulative cost and megaproject-risk perspective for
markets with less than 3 GWe of demand. However, if the plant is entering a large market with a high
probability of more than 3 GWe installed capacity, the LPSR, despite the megaproject risks, potentially
becomes the better option.
Table 11. Factors correlated to megaproject cost overruns and construction delays [46] [47] [48] [35]. Note: For NC-SMR Nth-of-
a-Kind (20th order) numbers are also included in brackets.
Factor PWR12 LPSR LASR MMNC NC-SMR
FOAK >$1.5B ✓ ✓ ✓ ✓ -
New construction method - ✓ - ✓ -
Long construction duration
✓ ✓ ✓✓ ✓
(FOAK direct onsite
(21.9 Million) (17.7 Million) (36.3 Million) (17 Million) (5.2 [2.9] Million)
person-hours)
(19.1) (15.8) (25.9) (24.8) (33 [18])
(Hours/kWe)
Low % Modularized
✓ ✓ ✓
(% Total labor moved
(29%) (22% [22%])
offsite)

5 Conclusions and Future Work

This work presented a bottom-up advanced nuclear power plant cost estimation tool. The tool has value
as an independent analysis of vendor quotes useful for policymakers, utilities, and designers. As a
demonstration of the methodology, we estimated the costs for the first ten plants from FOAK to NOAK
for five reactor designs: a Gen III PWR – PWR12, a large passive safety PWR – LPSR, a large active safety
reactor – LASR, a multi-module natural convection SMR – MMNC, and a natural convection SMR – NC-
SMR. The LPSR was the lowest cost per capacity plant but was at high risk for cost overruns and delays
as a megaproject. The NC-SMR was the most expensive FOAK per unit capacity but had equal or better
cumulative costs per cumulative capacity to the LASR and MMNC plants, and the lowest risk for cost
overruns. This demonstrated that learning-by-doing is the best cost reducing strategy for new nuclear
projects, and smaller plants, that are highly modularized, can most effectively leverage this strategy.
Modularization alone is not a sufficient cost reducing strategy, but it augments the impact of learning-
by-doing.

All of the cost differences estimated here can be eclipsed by cost overruns and delays. In considering the
megaproject cost overrun drivers, the best option for lowest cost and risk is to optimize the design using
traditional construction methods where there is ample supply of nuclear specific skilled labor such as
the LASR (e.g. APR1400) experience. This strategy avoids the potential cost overruns and delays
associated with new construction techniques and is likely to achieve NOAK costs more quickly.

While the methodology presented here was fairly comprehensive, there were many assumptions and
uncertainties. The learning and modularization models that were applied on the individual SSCs were
derived based on high level data and exploring a bottom-up methodology in their treatment is
worthwhile. Future work should also include uncertainty quantification of the learning, modularization,
and indirect cost estimation models. Maronati et al. assessed the cost of unknown unknowns in nuclear
plant construction for the generic PWR12 design [49]. Future work should attempt to quantify the
megaproject risk and unknown unknowns for specific nuclear plant designs, so that the analyses that
were performed to create Figure 2b and Figure 7 can be corrected to account for these risks. This
proposed analysis will need to include some assessment of the uncertainty associated with these risks as
well. Finally, detailed O&M cost estimation and cost reduction strategies to overcome lack of economy
of scale should not be ignored and will be included in the overall costing methodology. In this future
state, the model can be used as an active part of the design process for optimization and optimization
under uncertainty.

Acknowledgments
This work was primary support by Electricité de France (EDF) and partially supported by FORTUM. The
authors thank the EDF project advisory board including Prof. Jacopo Buongiorno (MIT), Mr. Eric Ingersoll
(LucidCatalyst and Terrapraxis), Dr. Jeremy Gregory (MIT) and Prof. Keith Ridgway (University of
Strathclyde). The review comments by Prof. Bojan Petrovic and Dr. Romney Duffey were also
appreciated.

Credit author statement

W. Robb Stewart: Methodology, Software, Formal Analysis, Visualization, Investigation,
Writing – Original Draft
Koroush Shirvan: Conceptualization, Methodology, Investigation, Supervision, Funding
acquisition, Writing – Original Draft

References

[1] J. Buongiorno and e. al., "The Future of Nuclear Energy in a Carbon Constrained World," MIT
Energy Initiative, Cambridge, 2018.

[2] Nuclear Energy Agency, "Unlocking Reductions in the Construction Costs of Nuclear: A Practical
Guide for Stakeholders," OECD, 2020.

[3] J. R. Lovering, A. Yip and T. Nordhaus, "Historical construction costs of global nuclear power
reactors," vol. 91, 2016.

[4] M. D. Carelli, P. Garrone, G. Locatelli, M. Mancini, C. Mycoff, P. Trucco and M. E. Ricotti, "Economic
features of integral, modular, small-to-medium size reactors," Progress in Nuclear Energy, vol. 52,
no. 4, pp. 403-414, 5 2010.
[5] V. Kuznetsov, "Options for small and medium sized reactors (SMRs) to overcome loss of
economies of scale and incorporate increased proliferation resistance and energy security,"
Progress in Nuclear Energy, vol. 50, no. 2-6, pp. 242-250, 2008.

[6] M. V. Ramana, "Small Modular and Advanced Nuclear Reactors: A Reality Check," IEEE Access, vol.
9, pp. 42090-42099, 2021.

[7] B. K. Sovacool and M. V. Ramana, "Back to the Future: Small Modular Reactors, Nuclear Fantasies,
and Symbolic Convergence," Science, Technology, and Human Values, vol. 40, no. 1, pp. 96-125, 7
2014.

[8] International Atomic Energy Agency Department of Nuclear Energy, "Advances in Small Modular
Reactor Technology Developments," Austria, 2020.

[9] B. Mignacca and G. Locatelli, "Economics and finance of Small Modular Reactors: A systematic
review and research agenda," Renewable and Sustainable Energy Reviews, vol. 118, 2020.

[10] Black & Veatch, "Cost and Performance Data for Power Generation Technologies," National
Renewable Energy Laboratory, 2012.

[11] P. Scherrer, "Estimating the cost of Small Modular Reactors," Grenoble Phelma, Grenoble, France,
2017.

[12] G. Black, F. Aydogan and C. Koerner, "Economic viability of light water small modular nuclear
reactors: General methodolgy and vendor data," vol. 103, 2019.

[13] Ux Consulting, "SMR Market Outlook," Roswell, GA, 2013.

[14] R. Rosner, S. Goldberg and J. S. Hezir, "Small Modular Reactors-Key to Future Nuclear Power
Generation in the," 2011.

[15] Z. Zhang and Y. Sun, "Economic potential of modular reactor nuclear power plants based on the
Chinese HTR-PM project," Nuclear Engineering and Design, vol. 237, no. 23, pp. 2265-2274, 12
2007.

[16] G. Maronati, B. Petrovic and P. Ferroni, "Assessing I2S-LWR economic competitiveness using
systematic differential capital cost evaluation methodology," Annals of Nuclear Energy, vol. 145,
2020.

[17] C. A. Lloyd, T. Roulstone and R. E. Lyons, "Transport, constructability, and economic advantages of
SMR modularization," Progress in Nuclear Energy, vol. 134, 4 2021.

[18] International Atomic Energy Agency, "Approaches for Assessing the Economic Competitiveness of
Small and Medium Sized Reactors," IAEA Nuclear Energy Series, Vienna, 2013.

[19] U.S. DOE, "Phase IX Update (1987) Report for the Energy Economic Data Base Program," U.S.
Department of Energy, 1988.
[20] F. Ganda, T. Taiwo and T. Kim, "Report on the Update of Fuel Cycle Cost Algorithms," Argonne
National Laboratory, Chicago, 2018.

[21] F. Ganda, E. Hoffman, T. Taiwo, T. Kim and J. Hansen, "Report on the ACCERT Cost Algorithms
Tool," Argonne National Lab, Chicago, 2019.

[22] Energy Impact Center, "OPEN100 Economics," [Online]. Available: https://www.open-

100.com/economics. [Accessed 13 11 2020].

[23] P. Champlin, "Techno-Economic Evaluation of Cross-Cutting Technologies for Cost Reduction in

Nuclear Power Plants," MIT. Masters Thesis, 2018.

[24] G. Towler and R. Sinnott, "Chapter 7 - Capital Cost Estimating," in Chemical Engineering Design
(Second Edition), Butterworth-Heinemann, 2013, pp. 307-354.

[25] The Economic Modeling Working Group of the Generation IV International Forum, "Cost
Estimating Guidelines for Generation IV Nuclear Energy Systems," OECD Nuclear Energy Agency,
2007.

[26] BLS, "Occupational Employment Statistics," U.S. Department of Labor, Bureau of Labor Statistics,
Washington, D.C., 2019.

[27] W. Vatavuk, "Updating the CE Plant Cost Index," Chemical Engineering, pp. 62-70, January 2002.

[28] EPRI, "Workshop about Economics-Based R&D for Nuclear Power Construction," 2019.

[29] X. Chen, A. Kotlyarevsky, A. Kumiega, J. Terry, B. Wu, S. Goldberg and E. A. Hoffman, "Small
Modular Nuclear Reactors: Parametric Modeling of Integrated Reactor Vessel Manufacturing
Within a Factory Environment," U.S. Department of Energy , 2013.

[30] D. Gandy and M. Albert, "SMR Pressure Vessel Manufacturing and Fabrication," Electric Power
Research Institute, 2019.

[31] R. Duffey, "Size and Cost Optimization of Nuclear Reactors in Energy Markets: The Need for New
Approaches and Advances," Ottawa, 2018.

[32] R. Lyons and T. Roulstone, "Production Learning in a Small Modular Reactor Supply Chain," The
International Congress on Advances in Nuclear Power Plants, 2018.

[33] G. S. Tolley and D. W. Jones, "The Economic Future of Nuclear Power," The University of Chicago,
2004.

[34] N. Bertram, S. Fuchs, J. Mischke, R. Palter, G. Strube and J. Woetzel, "Modular construction: From
projects to products," McKinsey & Company: Capital Projects & Infrastructure, 2019.

[35] E. Merrow, "Understanding the Outcomes of Megaprojects," RAND Corporation, 1984.

[36] SteelBenchmarker, "Global HRB Prices Sustain Prolonged Rally," www.steelbenchmarker.com,
2020.

[37] P. Dorfman, "Gulf Nuclear Ambition: New Reactors in United Arab Emirates," The Nuclear
Consulting Group, 2019.

[38] Utah Associated Municipal Power Systems (UAMPS), "Development Approach to the Carbon Free
Power Project," 21 July 2020. [Online]. Available:
https://www.losalamosnm.us/UserFiles/Servers/Server_6435726/File/Government/Departments/
Public%20Utilities/CFPP/20200721-
Presentations/CFPP%20Town%20Hall%20Development%20Approach%20_Mason%20Baker_7-21-
20.pdf.

[39] Nuclear Engineering International, "Kepco and Enec set up joint venture for Barakah NPP," 25
October 2016. [Online]. Available: https://www.neimagazine.com/news/newskepco-and-enec-set-
up-joint-venture-for-barakah-npp-5647366/. [Accessed 18 November 2020].

[40] World Nuclear News, "NuScale announces SMR power uprate," 11 November 2020. [Online].
Available: https://www.world-nuclear-news.org/Articles/NuScale-announces-SMR-power-
uprate#:~:text=The%20increase%20in%20generating%20capacity,about%20USD2850%2C%20the
%20company%20said. [Accessed 9 12 2020].

[41] NuScale Power, "NuScale Power Announces an Additional 25 Percent Increase in NuScale Power
Module™ Output; Additional Power Plant Solutions," 2020. [Online]. Available:
https://newsroom.nuscalepower.com/press-releases/news-details/2020/NuScale-Power-
Announces-an-Additional-25-Percent-Increase-in-NuScale-Power-Module-Output-Additional-
Power-Plant-Solutions/default.aspx.

[42] C. Lewis, R. MacSweeney, M. Kirschel, W. Josten, T. Roulstone and G. Locatelli, "Small modular
reactors Can building nuclear power become more cost-effective?," Ernst & Young Global Limited,
London, 2016.

[43] G. Black and S. Peterson, "Economic Impact Report: Construction and Operation of a Small
Modular Reactor Electric Power Generation Facility at the Idaho National Laboratory Site, Butte
County, Idaho," Regional Economic Development for East Idaho, 2019.

[44] G. Locatelli, "Why are Megaprojects, Including Nuclear Power Plants, Delivered Overbudget and
Late? Reasons and Remedies," 2018.

[45] G. Locatelli, M. Mikic, M. Kovacevic, N. Brookes and N. Ivanisevic, "The Successful Delivery of
Megaprojects: A Novel Research Method:," https://doi.org/10.1177/875697281704800506, vol.
48, no. 5, pp. 78-94, 10 2017.

[46] A. Fayek, S. Revay, D. Rowan and D. Mousseau, "Assessing Performance Trends on Industrial
Construction Mega Projects," Cost Engineering, vol. 48, pp. 16-21, 2006.
[47] A. Chanmeka, S. Thomas, C. Caldas and S. Mulva, "Assessing key factors impacting the
performance of oil and gas projects in Alberta," Can. Journal of Civl Engineering, vol. 39, pp. 259-
270, 2012.

[48] B. Flyvbjerg, "What You Shoul Know About Megaprojects and Why: An Overview," Project
Management Journal, pp. 6-19, 2014.

[49] G. Maronati and B. Petrovic, "Making Construction Cost Estimate of Nuclear Power Plants
Credible: Assessing Impact of Unknown Unknowns," Nuclear Technology, 2020.

[50] C. Lloyd, C. Middleton and T. Roulstone, "The impact of modularisation strategies on Small
Modular Reactor Costs," The International Congress on Advances in Nuclear Power Plants, 2018.