Skip to content

A Small Bet with Big Stakes

Two economists make a bet on how much it costs to build a nuclear power plant.

Back in November 2011, I made a bet with economist Geoff Rothwell about the future of nuclear power.  By making a bet we weren’t trying to trivialize the issue. To the contrary.  We had different predictions, and we saw this as a way of standing behind our beliefs, albeit in a modest way.

Our bet is about how much it costs to build a nuclear power plant. This might not seem very important, but I have grown to believe that this is the single most important number when it comes to determining the prospects for nuclear power, the only large-scale source of baseload electricity generation that does not emit carbon dioxide.

Geoff and I were at a conference, and I referred to an MIT study that estimated a cost of construction of $4,200 per kilowatt of capacity. Geoff thought this was too high.  So he challenged me to a $20 bet.  We agreed that we would use the two new reactors currently under construction at the Vogtle Plant in Waynesboro, Georgia as our measure. The new reactors will have a combined capacity of 2200 megawatts. So at $4,200 per kilowatt this is $9.2 billion in total construction cost. I went easy on Geoff and agreed to exclude financing costs and focus only on the “overnight” cost of construction.

The Vogtle project is extremely interesting because it is the first new nuclear construction in the United States in three decades. The Southern Company received a construction license from the U.S. Nuclear Regulatory Commission (NRC) in February 2012 and began construction officially in March 2013, though “pre-construction” has already been going on for several years.

It is not clear yet who will win the bet. There have been some delays, but construction is moving forward. Photos from January 2014 from the Southern Company’s website (below) document continued progress on the containment vessels and cooling towers. Click here for the complete set of photos.

Vogtle 1

Vogtle 2

According to a recent article by Reuters, the projected cost of the reactors is now $14 billion. This includes financing costs, however, so it is not clear how this compares to $4,200 per kilowatt. The new reactors are scheduled to come online in 2017 or 2018 so more-detailed budget information will become available soon.

Geoff is the Principal Economist at OECD Nuclear Energy Agency and has worked on nuclear energy his entire career so this bet would seem to be a risky move. I am not a nuclear engineer nor am I an expert on construction budgets. But I have looked closely at the history of nuclear construction projects in the United States, and the empirical evidence is not encouraging.

Construction Costs

Source: Lucas Davis, “Prospects for Nuclear Power”, JEP, 2012.

The figure above plots overnight construction costs for U.S. nuclear reactors by year of completion. While one might have hoped to see evidence of “learning-by-doing” the data actually show a pronounced increase in construction costs over time. Reactors that were ordered during the 1950s were built quickly and at relatively low cost, while reactors ordered during the 1970s ended up taking much longer and costing more.

Part of the story is a rapidly evolving regulatory process. A joke in the industry was that a reactor vessel could not be shipped until the total weight of all required paperwork had equaled the weight of the reactor vessel itself. The NRC has new procedures intended to streamline the regulatory process, but it remains to be seen how well these reforms will work in practice.

But even ignoring potential regulatory delays, nuclear power plants are still enormously challenging projects. The sheer scale of these facilities means that most components must be specially designed and constructed, often with few potential suppliers worldwide. These components are then assembled on site, and structures are built to house the assembled components. All stages of design, construction, assembly, and testing require highly-skilled, highly-specialized engineers and managers.

So it will be very interesting to continue to follow the experience at Vogtle.  From the point-of-view of the planet, I hope I am wrong, and that the reactors end up costing much less than expected. To address climate change we need as many technological alternatives as possible, and if nuclear power could be done cheaply it would transform energy markets. But we also need to be honest about the costs and these large construction costs make it difficult to make an economic argument for nuclear power.

Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas.

Suggested citation: Davis, Lucas. “A Small Bet with Big Stakes” Energy Institute Blog, UC Berkeley, February 18, 2014,



Lucas Davis View All

Lucas Davis is the Jeffrey A. Jacobs Distinguished Professor in Business and Technology at the Haas School of Business at the University of California, Berkeley. He is a Faculty Affiliate at the Energy Institute at Haas, a coeditor at the American Economic Journal: Economic Policy, and a Research Associate at the National Bureau of Economic Research. He received a BA from Amherst College and a PhD in Economics from the University of Wisconsin. His research focuses on energy and environmental markets, and in particular, on electricity and natural gas regulation, pricing in competitive and non-competitive markets, and the economic and business impacts of environmental policy.

34 thoughts on “A Small Bet with Big Stakes Leave a comment

  1. I love this “nukes are expensive” myth. Our 1.5 trillion tons of CO2 loosed on the planet in 150+ years is “expensive”, more expensive than humanity will easily pay. Here’s my favorite example for anti-nukes — take their favorite ‘expensive’ plant, say the Finnish one, now $11B over its $4B budget. When started up in a few years, it’ll be cranking out ~1500 Benjamins each hour as revenue at $0.10/kW Hr. There are 8760 hours/year. Do the math. If it substitutes for German ‘renewable’ juice at $0.30/kWHr, it’ll pay itself off in 4 years, then run for decades, making billions more than it cost to build and operate. If it got carbon credits, well game over, ‘renewables’ (except local solar PV/hot-water).

    Writers, media and too many researchers don’t yet seem to get what it means to have swamped the planet’s natural carbon cycle by a factor of ~1500. In other words, if the cycle keeps trucking along at sequestering 300 million tons of C in seafloor limestone each year, our 500 billion tons of fossil c might actually be handled in 1500 years, being recycled to seafloor limestone by oceanic life forms.

    But that’s only if a) we stop burning all C today; and b) if ocean pH is protected from further dropping below 8.0 (we’re at 8.1 average now). The CO2 in air is only ~25% dissolved in seas so far. pH is dropping so fast that before 2050, extinctions will occur that will halt that ~300 million tons of natural C sequestration. Oops. How expensive will it be to lose ~20% of all human food protein on earth?

    The idea in the ’60s was to eliminate combustion power by about 2000:
    That meant deploying 1GWe of non-emitting power each week by 1980 — each week. Oops.

    The expense we now have will rightly cause our descendants to spit on our graves, if those can be found.

    But we must get working, regardless. “Nukes are expensive” — right.
    As the old farm saying goes: “There’s no substitute for human stupidity.”

    Where are those folks who knew what to do >50 years ago? Y’know, folks like those who got us to the Moon in less than a decade. or who built advanced reactors in less time?

    Dr. A. Cannara
    650 400 3071

    • Wholesale electricity prices in Scandinavia and Germany have both been much smaller than the $0.1 or $0.3 per kwh figures you quote. In the past 5 years, Scandinavian prices have fallen from a high of 0.053 EUR/kwh in 2010 to 0.029 EUR/kwh in 2014 on average in recent years[1]. Forward prices for Germany next year are currently at 0.031 EUR/kwh[2]. At the current exchange rate of about 1.15, we’re looking at prices in the ballpark of $0.035 per kwh–not $0.3 or even $0.1.

      Account for that, and your breakeven period goes from 4 years to 40—and that’s if you forget to account for operating costs. which contrary to popular belief are substantial. Sure, fuel is cheap, but O&M and incremental capital investments are substantial. And don’t forget about waste storage costs. EIA estimates that Fuel+O&M costs for new nuclear amount to about $0.024 per kwh[3]. If that’s the only cost (e.g., no interest payments) and European electricity prices stay at their current level around $0.035, then a plant is making only $0.01 per kwh. Now your breakeven period isn’t 4 years, it’s closer to 100 years.


  2. Assoc. Prof. Davis is more likely to win the bet because any modern construction ought to take caution against more eventualities including accidents as happened in Fukushima. Hitherto Fukushima, designers probably did not think about Tsunamis affecting nuclear installations.

    Lower costs in developing countries like China are not examples to learn from unless it can be established that they are not cutting corners and are considering every safety concern in advance countries like the USA.

    One question borders my mind: Assoc. Prof. Davis mentioned in paragraph 2 that ” nuclear is the only large-scale source of baseload electricity generation that does not emit carbon dioxide.” My question is: Is Hydro also not suitable for large scale baseload electricity generation?

  3. And Areva is experiencing massive overruns on nukes in Finland and France. The cost overruns on building and more recently refurbishment of CANDU units in Ontario and New Brunswick have also been stunning.

  4. I’d have to reiterate the concern about leaving out finance costs. My understanding was that it was finance costs that caused the increasing cost trend for new construction that you mentioned, and killed the US nuclear industry. Comparing your chart with the history of the prime lending rate suggests a causation — High interest rates would also affect the supply chain, not just the plants themselves.

    With low interest rates now, plus CWIP and loan guarantees and Price Anderson and the nuclear PTC ($22 per MWh for the first 8 years = ~$2.7 billion), you’d think Vogtle would be a lot cheaper (including finance costs). Or perhaps all the subsidies are driving up construction cost, since contractors know that Southern can still absorb higher costs.

    Worth exploring. Thanks for the post.

    • No. Finance costs contributed to the escalation in costs for many plants, particularly those that took decades to build. But the direct costs rose dramatically, as well, as is obvious if you look at the escalation in costs of the plants whose owners recovered CWIP in rate base. Scope grew throughout these projects, as the builders found out just how complex they were and the owners and NRC found new problems with the designs.

  5. It is unfortunate that it is so costly to build a nuclear power plant. I do not understand why more effort has not been made to develop standardized, modular, factory-made nuclear plants that can be more easily deployed and do not represent such a massive single investment.
    Of even greater concern though is the potential closure of existing nuclear plants threatened by Exelon and others. As more and more nukes approach their initial life expectancy, the cost of keeping them operating is higher than their expected revenue. A nationwide GHG cap and trade program or a carbon tax would make GHG emission-free nuclear generated energy more competitive, even compared to natural gas plants. With the astronomic cost of building new nukes, it would appear to be prudent to find ways to keep existing nukes operating. Nuclear plants in the US generate about 800 million MWh per year and emit virtually no CO2. Each 1,000 MW base-loaded unit retired and replaced with natural gas or coal increases total CO2 emissions by 4 to 8 million tonnes per year.

    • 1,200 MW, or even 600 MW nuclear units are too large to build in a factory. Much smaller units could be more extensively prefabricated, but might be just as expensive, without the economies of scale. Building anything is expensive if failure is not an option.

      Nuclear units are also very expensive to keep in service, in terms of operating costs and retrofit costs. That is particularly true as units age. If the plants could be operated economically, Dominion, SCE, and Duke would not have retired Kewaunee, San Onofre 2 & 3, and Crystal River, and Entergy would not have announced the retirement of Vermont Yankee.

      • The world aircraft manufactories build about 8 units the size and complexity of a SMR each and every DAY. That suggests that we could be installing ~1GW capacity/d around the world. Demand drives the price in that high demand requires efficient manufactory to meet.

    • France is interesting because beginning in the 1970s it adopted a single design for all of its reactors. This high degree of standardization led to average construction costs below the U.S. average, though still not cheap by any means. Newly available data has made possible some interesting new studies on this topic, including looking for learning-by-doing. The analyses differ on the magnitude, but it seems pretty clear that, if anything, there was actually cost escalation, similar qualitatively to the United States.

      Click to access CostEscalationPaper.pdf

      • No, it didn’t. Look at the wikipedia list of commercial reactors and examine all the French units in detail. There are quite a number of designs. They were LESS “one off”, but not standardized either.

  6. Thanks for this post. What I am curious about is whether and how advances in the fields of construction and supply chain management since the prior build cycle, as well as design standardization, will affect the cost of nuclear power. Also, do we need better metrics of cost relative to performance along desired criteria?

  7. Leaving out the financing costs makes this analysis closer to fantasy. With natural gas and renewable energy, their deployment can be operational within one year to eighteen months – not a decade or more like nuclear power. Leaving nuclear waste and decommissioning costs out of this equation also skews the results. And finally, mining uranium, building generation plants for a decade, and waste storage for hundreds+ plus years — does not in anyone’s imagination make nuclear power carbon free by any means. So I laud Professor Davis on starting the analysis, but I chide him a bit for not making it more “real world”. Scott Sklar, Adjunct Professor, The George Washington University

    • Financing costs are economic costs, and I agree, should be included in any comparison of generating technologies. Indeed, any attempt to calculate the levelized costs of electricity from different sources necessarily must take a stand on the cost-of-capital and the time profile of construction costs. Our focus on overnight costs was more pragmatic. We wanted to make the bet about construction costs, not interest rates. Most analyses of nuclear construction costs have focused on overnight costs for this same reason.

  8. Lucas — The more relevant bet is on the overnight costs of new nuclear in China and the rest of the developing world, where most of world’s future demand growth is. There the data is far more encouraging, with reported overnight costs in the $2,000-3,500 range. Who knows, we might even learn something from the Asian build?

    • Great point. I’ve heard a rule-of-thumb for nuclear power plants that 60% of construction cost is labor, so it makes sense that costs could vary widely across countries. Closely examining the cost data from recent builds in China is a hugely important priority for future work.

      • We learned that despite a worst case accident of THREE reactors that didn’t even have the most basic of safety upgrades and had a monumentally dumb extreme load scenario, nothing much happened in the grand scheme of things.

        Reactor construction has INCREASED ~67% since Fukushima. Seems the world learned the right lesson after all.

    • Interesting to note that China itself is making an each way bet on nuclear – concurrently building Chinese Generation II design reactors yet also US and European Generation III reactors as well. The additional safety and other enhancements of Generation III do have some capital cost implications that are relevant here.

  9. Lucas,

    You may have already won the bet. I suspect the $14 billion construction cost estimate that Southern Company refers to on their website is the plant’s overnight cost because the utility is authorized to include the financing costs, i.e., “Construction Work in Progress” (CWIP), in its rate base. Through the CWIP vehicle Southern Company is contemporaneously recovering the financing costs from their customers through retail rates.

    • Yes this is correct that Southern Company is already beginning to pass on these costs to electric customers. It is just hard without seeing details to know exactly how they came up with the $14 billion number. Another wrinkle is the $8.3 billion loan guarantee Southern Company has received under the 2005 Energy Policy Act which presumably lowers their borrowing cost substantially.

      • The loan guarantee basically says that if the government decides to screw up the business plans of the Southern Company, they will pay for it. It is only appropriate since it is dumb government policy that causes most of the problems!

      • According to Georgia Power’s 11th Semi-Annual Construction Monitoring Report (dated August 2014), the estimated value of the loan guarantee, as measured by savings on AFUCD, will sum to roughly $250 million in 2018 dollars. In other words, this amount of money will not need to be added to the ratebase and later recovered (with interest) over the life of the plant. I don’t know what their inflation projection is, but based on one I have, it’s equal to about $208 million in 2007 dollars.

        Georgia Power’s estimate only refers to its share of the plant; if you gross it up to account for the other owners, the total is about $454 million in 2007 dollars. If you divide by the net electrical capacity of the plant (2234 MW), then you arrive at a result of an implied subsidy of 20 cents per watt to the total, final capital cost.

        Per my comment elsewhere in this thread, the total capital cost has now been revised to about $16.5 billion due to construction delays. (Time is money!) This equates to ~$7.4/W. 20 cents per watt divided by $7.4 per watt = 2.75%. So, the loan guarantees pencil out to something equivalent to a 2.75% investment tax credit. (c.f. the 30% investment tax credit for solar)

        Of course, the value of a subsidized interest rate will grow with the construction delays as more AFUDC accumulates, so this is likely an underestimate, but somewhere in the ballpark of the true magnitude.

    • Sorry, Robert, but your suspicion is incorrect. $14 billion includes both overnight and financing costs inflation-adjusted back to 2007 dollars, as estimated when the plant was originally certified, anyways. Because of the delays and outlays arising from addressing the problems with the modules delivered by CB&I (more info here:, the final in-service cost is looking more like $16.5 billion.

      For the original and latest forecasts of both overnight and financing charges for Vogtle 3 & 4, go here:

      Readers should take care to note that all dollars are reported in 2007 dollars, and only Georgia Power’s share (45.7%) of the project is reported.

      Lucas, I take it your bet with Gregory was based on MIT’s 2009 update, and then was adjusted for inflation to 2011 dollars? For convenience, I will continue the rest of my post in 2007 dollars, in which case your bet is based on a figure of $4000/kW, as reported by both MIT (2009) and the Vogtle Construction Monitoring Reports.

      The originally certified overnight cost for Vogtle was $4,328/kW. This would imply that Gregory lost the bet when the Georgia Public Service Commission approved the certificate of public necessity and convenience in March of 2009 . Perhaps Gregory should have done a little more research before he accepted the bet. Whatever the case, the latest forecast of overnight cost from Georgia Power is $4,974/kW.

%d bloggers like this: