You are here

The high and rising cost of nuclear reactors

Nuclear Monitor Issue: 
Dr. Mark Cooper

A- The range of current cost estimates
Table III-1 presents more than sixty estimates of nuclear reactor costs from over three dozen entities that have been published since 2001, when the nuclear industry first claimed a nuclear renaissance was imminent. The table shows the overnight, all-in, and busbar costs, where they are available, and attempts to impose order on the projections by stating costs in constant 2008 dollars, using the GDP deflator to restate the costs. When the dollar vintage was not specified in the study, it was assumed to be the year of the study. Figure III-1 shows the overnight costs for both the completed plants and the projections for future plants. The estimates are roughly equally divided between government consultants, utilities, government entities, utilities, and Wall Street/independent analysts, plus a small number of academic institutions. Many of the estimates are not very well explained or documented, while a few are analyzed in great detail.

The projected costs have quickly escalated over the past decade. The low estimates from vendors, academics, and government agencies have approximately doubled. However, they remain below the estimates from many of the utilities and well below the estimates from Wall Street and independent analysts. Several aspects of the cost estimates are worthy of note.

  • First, there has been a sharp increase in projected costs in a short period of time.
  • Second, the early government and academic costs were quite low.
  • Third, the recent utility cost estimates have doubled or tripled the first estimates but still tend to be lower than the estimates from Wall Street and the independent analysts.
  • Fourth, the governmental entities tend to use the average of other analyses, particularly the utilities.
  • Finally, the independent analysts tend to be the highest.

Even adjusting for inflation and stating all of the estimates in constant 2008 dollars, the projections are all over the map. However, it turns out that it is not very difficult to reconcile the estimates. A small number of variables account for the differences.

What these differences in estimates correlating with the type of institution making the estimate indicate is difficult to say. Utilities, especially in the early phase of the regulatory process, have an interest in understating costs, as long as the estimates are nonbinding. Low-balling the costs helps to get the power plant approved. In theory, Wall Street analysts are objective, but the recent crisis in the financial sector has called that into question. Wall Street analysts and rating agencies may have agendas related to their efforts to win clients.

B. Construction costs
Of the three dozen estimates included in Table III-1, several have publicly available and detailed documentation that enables us to isolate the key causes of differences in cost estimates. Most of the studies do not. Rather, they create high and low cost cases that assume different values for a number of variables simultaneously. These “high and low what if” scenarios may seem to bracket the range of possibilities, but if there is no reason to believe that the elements of the high or the low scenario should go together, the exercise may not be informative. It would be better to identify the individual impact of each cost element and project costs on a probabilistic basis.

B1. Overnight and busbar costs
Overnight costs are the single most important cost element. Overnight costs exhibit a strong direct relationship to busbar costs. Some of the studies provide a basis for describing the impact of overnight costs on busbar costs holding other elements constant. Figure III-3 graphs the results of four such studies. Each of the studies included in Figure III-3 provided a narrow range of overnight costs with which the effect of overnight costs on busbar costs can be estimated, holding all other things constant. Those projections have been extended over a wider range of overnight costs estimates to assess the magnitude of the effect of overnight costs on busbar costs across the studies.

The MIT model suggests that for every $1,000 of increased overnight costs, the busbar costs go up by 1.8 cents in the utility finance model and 2.4 cents in the merchant finance model. Moving from overnight costs of about $2,000 to about $7,000 raises the estimated busbar costs around 8 cents/kWh in the utility model and about 12 cents in the merchant model. In the Harding study, busbar costs go up about 2.4 cents per kWh for every $1,000 increase in overnight costs. In the University of Chicago study, the increase in busbar costs per $1,000 in overnight costs was 3.0 cents per kWh.

2. Financial models
There are two key elements that affect the extent to which financial costs magnify overnight cost differences. The higher the rate of return and cost of debt, the higher the financial costs. The larger the share of equity as compared to debt, the higher the financial cost.

Much of the impact of financial cost models can be encapsulated in the difference between utility and independent company finance. Some argue that independent power producers will build plants on a speculative basis.(*1) Others argue that only utilities will build them, and only with clear guidance to public utility commissions about needs and cost recovery.(*2) To date, the latter appears to be closer to the mark. Joskow and others do not believe that merchant nuclear reactors are very likely to be built, which is contrary to the assumption in the MIT analysis, so they applied a utility finance model to the MIT cost estimate. The Joskow numbers are shown in Figure III-3. With a lower cost of capital in utility finance version of the MIT analysis, nuclear reactors have lower capital costs and produce lower priced electricity.

The MIT model suggests that at $2,000 for overnight costs the difference between a utility and a merchant financial model is about 1.5 cents per kWh. The California Energy Commission Cost of Generation Model puts this figure at about 1.4 cents at an overnight cost of $2,950 per kW. As the overnight costs increase, the impact of the financial model is magnified. Thus, at $7,000 for overnight costs, the difference between the merchant and utility models in busbar costs is almost 5 cents per kWh.

C. Operating costs
Another cost element that can easily be factored into the framework of this analysis is the operating and maintenance costs. While construction and capital costs tend to attract the most attention, operating costs are significant. The MIT study used a low operating cost (including fuel) that it admitted was optimistic.(*3) Others have estimated operating costs (including fuel) to be much higher (See Figure III-4). The difference is between about 1.5 cents per kWh to almost 3 cents per kWh. The Keystone base case for operation and maintenance costs (including fuel) was 2.1 cents higher than the MIT base case. Adding this operation and maintenance cost difference to the overnight costs in the MIT study, based on the utility finance model (which was the approach taken in the Keystone study), we largely resolve the difference between the projected busbar costs as shown in Figure III-5.

Source: MIT, "The Future of Nuclear Power", 2003; Keystone Center, "Nuclear Power Joint Fact-Finding", June 2007; Severance, Craig A., "Business Risks and Costs of New Nuclear Power", January 2, 2009; Harding, Jim, "Economics of Nuclear Reactors and Alternatives", February 2009.

D. Escalators
This analytic exercise is just arithmetic until it is tied to real world causes. The MIT study started with low overnight costs (as hypothesized by the earlier Department of Energy funded studies) and then hypothesized ways overnight costs might decline.(*4) Many of the later studies derive their estimates by applying escalators to the early studies. In many of the studies since 2001, a wide range of overnight costs is presented as scenarios because there is uncertainty about construction costs, and construction costs have been rising.

The choice of an escalation rate for costs is an effort to properly inject reality into the model. Many of the discussions of escalation refer to the Cambridge Energy Research Associated (CERA) index of power plant construction costs. Harding points out that the CERA index for nuclear plant escalation has been as high as 14 percent per year.(*5) Harding identifies four levels of escalation of costs: zero, 4%, 8%, and 14%. Harding’s early analysis used the 4% figure and his later analysis argues that the 8% figure is closer to reality.(*6) He points out that the heavy construction cost index calculated by American Electric Power has been increasing at a rate of 10.5% per year. Thus, his conclusion that the 8% figure is a better basis for estimating overnight costs is moderate. In the Harding mid-scenario, the 8% escalation puts the overnight costs at $7,100 and the busbar costs at 17.3 cents per kWh. In the Harding high scenario, the 8% escalator yields overnight costs of $8,000 per kWh and busbar costs of 19.0 cents. Harding’s high model with high escalation puts the cost in the range of 21.2 to 23.5 cents. The MIT model with utility costs and Harding O&M costs predict the same busbar costs as specific overnight costs.

An update to the MIT study underscores how important these escalators can be.(*7) It cites the CERA index showing an increase in nuclear construction costs of 22.5% per year between 2002 and 2007, the years for which it estimated costs. However, it escalated costs at 15% per year to arrive at a cost of $4,000 in 2007 dollars, which results in a cost in the low end of recent estimates from utilities. If it had used the higher observed escalation rate for 2002-2007, it would have arrived at a figure that was about $1,500 per kW higher, or more than one-third higher.(*8)

Similarly, Severance uses an 8.8% figure for escalation, which puts the overnight costs at $7,400 in his most likely case and the busbar costs at 25 cents per kWh. The Severance analysis yields high busbar costs because it includes two other costs not included in other analysis. Severance adds 2 cents for property taxes and 2 cents for decommissioning costs, which are higher costs than used by others. Excluding these, Severance’s costs of 21 to 25 cents are close to Harding’s high-end estimates (21.2 cents to 23.5 cents).

There are two different escalations that are being estimated in these studies. First is the increase in costs that is projected because of past escalation. Since many of the studies launch from the earlier low-ball estimates, they must deal with the increase in cost estimates that have already taken place. As the various cost indices suggest, that increase has already been substantial. Whether costs will continue to escalate in the future is a separate question.

The estimates by Florida Power and Light (FPL) illustrate this distinction. The non-binding cost estimate was derived by escalating and modifying the earlier cost estimate from TVA for its proposed Bellefonte reactors.(*9) Moving from a 2004 estimate to a 2007 estimate, the projected cost of the plant doubled in real terms, suggesting an extremely high rate of escalation of 25% per year.(*10) Looking forward, however, FPL projects only a 2.5% real rate of escalation to arrive at a mid-point overnight cost estimate of just under $3,600 per kW in 2007 dollars.(*11) FPL acknowledges that Moody’s has questioned the low figures being used by utilities.(*12) If FPL used the rate of escalation of 8% for the next decade, its estimate would be well over $6,000, close to the number used by Moody’s.

Ironically, much of the analysis in the early 21st century sought to explain how very low capital and busbar costs might come about, since the historical experience suggested much higher costs. More recent analysis has attempted to explain why the earlier cost estimates were too low and how quickly costs had escalated and could escalate in the future. The current estimates of construction costs, which are much higher than the early estimates, should not have been a surprise.

They are perfectly consistent with the historical trend, as shown in Figure III-6.

There is a twist in the escalation of costs. The current recession has lowered material costs and reversed the dramatic upward trend in costs, but the CERA index shows only a moderate decline in the cost index.(*13) The index is down by less than 10%. However, utilities, whose cost estimates in 2007-2008 failed to reflect the full impact of prior cost escalation, are suddenly offering assurances that the slack markets caused by the recession will moderate future cost increases.(*14) They are admitting much higher numbers in their current statements than were used to launch their efforts to gain approval of the plants, but then attempting to cushion the impact with the assurance that declining commodity costs will lower costs. Although some have pointed out that commodity costs are a small part of total costs,(*15) the utility approach renders nuclear construction cost almost as volatile as fossil fuel prices, leaving one to wonder what will happen when the recession ends or if a flurry of orders puts pressure on prices.

E. Capacity factors and plant life
The methods used above to reconcile the differences between the various estimates have all relied on the base or mid-case estimates. We use these estimates for the comparative analysis because the studies’ authors tend to run their scenarios as modifications of the base case. These base cases tend to use the high capacity factors and long facility lives that are observed at present, which is the end stage of the cohort of reactors.

Capacity factors are an important assumption. Capacity factors of 90% that are observed today took two decades to achieve. It may be a mistake to assume that new reactors will achieve those high capacity factors from day one. In so far as the reactors and technologies are new and unique, there may be a substantial learning process before such high levels of reliability are achieved. The average capacity factor for reactors that have been operating in the U.S. is about 79%. The average for the reactor brought on line in the ten years between 1989 and 1999 is 88%.

Although capacity factors and reactor operating lifetimes do not have as dramatic an impact as the construction and capital costs, they are important. In the MIT study, with the base case assumption of a 40-year life for the reactor, decreasing the capacity factor from the base case assumption of 85% to 75% increases the busbar cost from 7.7 cents (2008 $) to 8.6 cents. Assuming the 85% base case capacity factor, lowering the lifespan of the reactor from 40 years to 25 years increases the cost from 7.7 cents to 8.6 cents. The worst case considered by MIT (75% capacity/25-year life) had a busbar cost of 9 cents, compared to the base case of 7.7 cents. The Keystone study varied both lifespan and capacity factor together. Moving from the base case of 40-year life and 90% capacity to the worst case, 30-year life and 75% capacity, raised the busbar cost from 9.7 cents to 11.4 cents. The busbar costs are higher in the Keystone study in large part because the overnight costs were assumed to be higher, as shown above in Table III-1.

This review can be used to suggest the impact of various key variables that affect the cost of nuclear reactors, although a range of projected costs will not be specified until the history of the industry is reviewed in more detail in the next section. Here the relative importance of each of the key factors in the general context of a move from overnight costs of $2,000 to overnight costs of $7,000 can be explored. The analysis must start with the range of overnight costs because the impact of the financial and plant characteristic assumptions varies depending on those costs. Starting from the MIT utility model, adding $5,000 of overnight costs would add about 9.6 cents per kWh to the estimate. In the merchant model it would add 1.5 to 3 cents per kWh. Assumptions about plant life and capacity factors could add another 1.7 to 3.4 cents per kWh. O&M costs are independent of the other costs, but the difference between the studies runs in the range of 2 cents. Given these large differences in cost projections, it is easy to reconcile the low 5.2 cents per kWh estimate of a utility finance model based on the MIT 2003 overnight costs to the high estimate of 16 cents per kWh, based on the CEC utility finance model. Starting at 6.1 cents in Joskow’s application of the utility model to the MIT base case, adding 9.6 cents for an additional $5,000/kW of overnight costs and 2.1 cents for operation and maintenance costs would yield an estimate of 17.8 cents, just above Harding’s estimate of 17.3 cents.