Why is nuclear power so expensive? (Patreon)

[This is a transcript with links to references.]

There are a lot of bad arguments against nuclear power. For example, that it’d just be too simple a solution to climate change, we need to make the transition difficult and painful. That’s not a good argument because it’s not actually that simple to prevent further global warming, even with nuclear power. But there are two good arguments against nuclear power: it’s too expensive, and the power plants take too long to build. But are they true? And if so, why is it that nuclear power is so slow and so costly? That’s what we’ll talk about today.

Before we talk about cost and speed, let’s quickly go through other common arguments against nuclear power, just so you know why I don’t think they’re good arguments.

The most common argument against nuclear is that nuclear waste is such a big problem. But in truth, there is very little nuclear waste. Because nuclear power has such a high energy density, you need little fuel to create energy with it, and consequently there’s little waste. Most of the waste is only mildly radioactive and that what’s left can be recycled or buried. I collected the numbers about nuclear waste in a previous episode.

American commercial reactors have only generated about 90 thousand metric tons of spent fuel since the 1950s. That could be fitted on a single football field at a depth of less than 10 yards. Doing this’d have the added benefit of mutating football players to grow more legs, which’d make the sport considerably more interesting but I digress.

I get often asked “Yes but would you want to live next to a nuclear waste deposit?” to which the honest answer is it wouldn’t bother me at all.

Another bad argument is accidents. Yes, there have been two major nuclear accidents, but if you look at the numbers of fatalities and further detrimental health effects, they’re nowhere near as high as the number of people who have died in plane crashes, and are basically nothing compared to the fatalities we can expect from climate change, which might well be hundreds of millions of people. Again, I collected the numbers in a previous episode.

Clearly you don’t want accidents, and not just because they kinda make bad publicity. But nothing comes without risks, and the risk of further increasing carbon dioxide levels is much higher than that of nuclear accidents. There is no way a nuclear power plant can go so terribly wrong that an accident would kill more people than letting carbon dioxide levels increase.

Another bad argument is: I can’t see it, but it can kill me, we can’t allow that. Well, you know what, you can’t see most pollution in the air, water, or soil so why aren’t you running around screaming all day that civilization must be discontinued immediately. Seems like somewhat of an overreaction?

yeah, the good thing about radioactive pollution is that in contrast to most other types of pollution, it’s easy to measure. Worried about radioactive pollution? Go and buy a Geiger counter. Want to know how many microplastics you’re breathing, or what’s with those forever chemicals in your tap water? Well, good luck finding someone to analyse your samples.

This is not a complete list of bad arguments, I’m sure you have come across some more. But the accusation that nuclear power plants take a long time to build and are just too expensive to make commercial sense are good arguments, so let’s look at those two.

We’ll start with the long construction time. I’ll be talking about the time from the beginning of the construction to the end, not including the planning stage. For a typical fossil fuel plant, it takes something between 3 and 4 years. If it’s a small gas plant, they can do it in about 2 years.

But how long does it take for a nuclear power plant? Well according to this guy who recently left a comment on my facebook page, it takes tens of years and then it’ll be ten times as expensive. Well maybe, just maybe, you shouldn’t trust random guys on facebook.

To be fair, he’s a bit on the extreme end. The numbers that I have seen being thrown around on social media is that it takes a decade to build a nuclear power plant. A typical example is this article by Naomi Oreskes in Scientific American from last year. It’s titled “New Nuclear Power Plants Are Unlikely to Stop the Climate Crisis”. In this article, she claims that “Globally, the average construction time is about 10 years” and “Some notorious examples have taken much longer”.  

Ok, let’s look at the source that she provides for this. It’s conference proceedings from 2011, with two authors. The paper says:

“The average construction time of nuclear power plants between 1976 and 2009 was ninety-two months or 7 point 7 years with a maximum of 10 years between 1996 and 2000.”

Ah right. She picked the average from those four years where the construction time was particularly long for whatever reasons, and then decided to not tell you about the past 20 years. Where it’s been lower.

Ok, so you get anti-nuclear activists claiming it takes 10 years to build a nuclear power plant, but how long does it really take? Before we can talk about this, I need to remind you of a subtlety of statistics. That’s the difference between the mean and the median of a distribution.

The mean of the distribution is the same as the average. You get it by summing up the values of data points and then dividing by their number. The *median is the point in the distribution that separates the lower 50 percent from the higher 50 percent.

Now if you have a symmetric distribution, like this standard normal distribution, then the mean and median are identical. However, if you have a distribution that’s skewed to one side, then the mean value will move towards the longer tail. It’s not that one of those statistical measures is right and the other wrong, they just tell you different things. As you see from this illustration, if you have a skewed distribution, the mean will give you a misleading impression of what’s typical because it pays more attention to the non-typical tail.

Confusing the two happens all the time. For example, if you were recently somewhat surprised to read that the “standard American household is now a millionaire” and wondered why none of your friends are millionaires, that was referring to the *mean household net worth. But in the United States the mean net worth isn’t very representative of a typical household, because you have a very heavily weighted tail of a few super-rich people. The *median American household net worth gives you a better impression. It’s something like 190 thousand dollars. . So that’s why your friends aren’t all millionaires, it's not the tap water.

Now remember that Oreskes wrote something about “notorious examples” where it took a very long time to build a power plant. Same problem as with the billionaires’ net worth. The most notorious example of a nuclear power plant that took a long time to build is the second unit of the Watt bar station in the United States. It took a whopping 43 years from the start of construction to the beginning of commercial operations. Though the construction was paused in between for 22 years, so this number is itself misleading.

That was a lot of warnings of how to interpret the data, now let’s see how they look. These figures were made by Hannah Ritchie who does amazing work for our world in data. As you can see, the average – that’s the mean value – of the construction times has somewhat gone up and down over the decades, but in the past 30 years or so, it’s been between 6 and 7 years. This is kinda compatible with the numbers from the proceedings article which Naomi misquoted.

This was the mean value, and as you expect, the median is lower than that. You can see this here. This figure shows the percentage of reactors that were finished within a certain time. You can see that it has this very long tail. The mean is 7 point 5, the median 6 point 3.

That’s the *global average, but there are big differences between countries. In Japan, the median time was merely 52 months, that’s about 4 and a half years, so actually not that different from fossil fuel plants. Japan also holds the record for the fastest-built nuclear power plant in the past decades. It took 39 months, so a little more than 3 years.

In the United States, the median is 91 months, that’s about 7 and a half years. But in the 1950s, the Americans built nuclear power plants much faster, in one case in just 21 months.

To summarize this part: It’s true that it takes longer to build a nuclear power plant than to build a fossil fuel plant, and renewable energy power plants can be built even faster. But the difference isn’t as big as some environmentalists claim. It doesn’t take decades; it rarely takes even one full decade. Typically it takes something like 6-7 years. The Japanese have done it routinely within 5 years. Contrary to rumours, the Japanese are also still using nuclear power and are even planning to build several new nuclear power plants. 

Let’s then talk about costs. The cost for building a new nuclear power plant is typically 5-10 billion US dollars. On the one hand, that seems like a lot of money. On the other hand, it’s still less than Elon Musk paid for twitter. Without context it’s hard to tell if that’s much or not.

So here’s the context. First, it really makes no sense to talk about the total costs of a power plant, because some of them are larger and some smaller. It’s like talking about the cost of wine without saying whether it’s the price for a glass or a barrel.  

Therefore, when it comes to power plants, one usually either talks about the cost per energy capacity, or the cost per power that the plant eventually delivers. The units are then either dollars per Watthours or dollars per Watt. If you’re a particle physicist then maybe you’d prefer something like Swiss Francs per Teraelectronvolt Fermi over c, but let’s not go there. It’s complicated enough already because, let me guess, you’ve forgotten what the difference is between energy and power. That’s totally okay. If I wouldn’t have to talk about it on YouTube, I’d probably also have forgotten.

The difference between a unit of energy like Watthours and a unit of power like Watts is that the energy tells you the total you get, whereas the power tells you what you get per time. Suppose you pour your wine out of a bottle. Then the content of the bottle is kind of like the energy capacity of the power plant, whereas the amount of wine that you pour per second is kind of like the power.

This means if you want to know the cost normalized by the size of the power plant, you’ll look at something like the cost per bottle of wine, so for the plant that’s the cost per energy capacity. To get this number you want to average over the entire lifetime of the plant and take into account the entire costs, both for construction and operation. The result is known as the levelized cost of energy, or levelized cost of electricity.

Okay, now that we know what we’re talking about let’s look at the numbers. The levelized cost of energy for nuclear power has been going up in the past decades. It’s not controversial, pretty much all sources agree on this. These are unsubsidised costs. Nuclear power is now three times as expensive as solar and more than twice as expensive as gas.

So we see that, yes, nuclear power is more expensive than other sources of energy, but it’s not *hugely more expensive. It’s not a factor 100, it’s not even a factor 10, it’s a factor 2 to 3. As a theoretical physicist I’d say that’s of order one, so basically the same really. You probably shouldn’t leave your asset management to a theoretical physicist.

Okay, so nuclear power is expensive, but what makes it expensive? The biggest contributor to the cost is the construction, whereas the operating costs are quite moderate. To get a sense of the construction cost, it makes sense to first leave aside the cost that accumulates from interest rates during the construction – we’ll come back to this later. For now, we’ll pretend that the plant is built overnight. It’s called the “overnight construction cost”.

A look at this overnight construction cost reveals the problem. It’s been going up for decades and it’s been the same in almost all countries. This figure shows the overnight construction cost for nuclear power plants as a function of the total installed capacity, globally. So from left to right you have more nuclear power plants on the planet.

You see that all those curves have a V-shape. They go down, then they start going up. It’s the same in all countries, though in the United States the rapid increase in cost is particularly pronounced. Only exception is South Korea, and since no one knows how that could possibly be, experts have doubted that the numbers are correct.

This turnaround happened somewhere between 32 and 64 Gigawatt. We can see from this figure that this was some time in the 1960s, so presumably the hippies are to blame.

More seriously, let’s not lump this all together because while the curves look similar in shape, the total costs still have large differences between countries. A 2018 report by researchers from MIT found that in East Asia, the construction cost for a nuclear power plant was about three to 4 thousand US dollars per kilowatt, while in Europe and Northern America, the cost was almost twice as high with 8,000 USD per kilowatt.

Now remember that the difference between the cost of nuclear and fossil fuel plants was something like a factor 2 to 3, and you’ll see why nuclear makes sense in some places and less in others.   

Ok, we’re getting closer, but just why is it so expensive to build nuclear power plants in some places of the world?

In a 2020 paper, a team of American researchers say the issue is that in Europe and North America, labour productivity has generally decreased greatly, not just for nuclear power plants. They say that in recently built plants labour productivity has been up to 13 times lower than in the 1950s. On-site productivity has sharply declined and this, they say, is the major reason that the costs for constructing the whole thing have more than doubled from 1976 to 2017.

Alright, but that doesn’t explain why nuclear power plants are so particularly expensive. What are those guys doing? Are they watching too much YouTube? Binge drinking? Is it the hippies after all?

Well, no, it’s because of an ever-increasing number of regulations and safety requirements for building nuclear power plants. Those have not only greatly increased the amount of paperwork and the time to fill in this paperwork, but they also necessitate testing and retesting parts of the equipment which takes time and that looks like productivity is declining.

The impact of these regulations is quite astonishing. A 2017 paper estimated that extra quality requirements for nuclear power plants, which they call a “nuclear premium,” make up 23 percent of the cost of concrete, and 41 percent of the cost of steel, which together are the main building parts of the power plant.

Over time, the problem with labour productivity has only become worse. Because the more regulations there are, and the fewer nuclear plants are being built, and the more difficult it becomes to find skilled workers. This makes it more expensive to build them, and so fewer are being built, etc etc. This is why the Koreans and Japanese still build nuclear power plants faster if they build them in other countries. It’s not just the regulations, it’s also their experience and good organization and management.

And here’s the kicker. The issue with the regulations becomes considerably worse if you take into account that in reality power plants are not built overnight.

You see, most people don’t have 10 billion dollars lying around and can just build a nuclear power plant like that. They have to take on a loan. And on the loan, there’s an interest. And the longer they need to build the power plant, the longer it’ll take until they can start paying back the loan.

Of course it’s generally expensive to build a power plant, no matter what type, but the longer it takes to start operations, the higher the additional cost from the interest rate. That is, a big part of the reason why nuclear power is so expensive in the North-Western part of the world is that it takes so long to build.

A 2004 study commissioned by the US Department of Energy found that the interest payments during construction can be as much as 30 percent of the total capital cost, and it can get close to 40 percent if the plant takes 7 years to build.

That’s right, 40 percent of the cost is interest on the loan! And that was in 2004.

And then there’s the insurance costs. You see, while nuclear accidents have had a low number of fatalities, the total damages have been enormous because the contamination can spread so far. Decontaminating all this area, to the extent that it’s possible at all, is extremely expensive. And that makes insurances very uncomfortable.

According to the US Nuclear Regulatory Commission, owners of nuclear power plants pay an annual premium of 450 million dollars in private insurance for offsite liability coverage for each reactor site, plus another 131 million per reactor as a second-tier self-insurance.

And here I was thinking my car insurance is expensive.

However, if you crunch the numbers, this insurance doesn’t have much of an impact on the cost of nuclear energy. According to a review from 2013  it adds roughly 10 cent to 3 Euro per MegaWatthour which is in the range of 1 percent tops.

So the bottom-line is this: Nuclear power plants are expensive mainly because they take a long time to build, and they take a long time to build because of all kinds of extra regulations. In Europe and North America, the construction is extra slow because labour productivity is low, and the problem is exacerbated by lack of experience and bad management.

Now fulfilling regulations for nuclear power plants is about the absolutely last thing you want to rush, so it might seem like those costs are pretty much inevitable. But remember that the Japanese manage to do it. This shows that it is totally possible to speed up the checking and testing and proceedings of the paperwork, if there’s a will to do it.

Ah, yes, you might say, but look what happened in Fukushima. That’s what happens if you build too fast like those Japanese have done. Well actually I’d say what Fukushima has shown is how remarkably safe nuclear power plants are.

Remember that the Fukushima accident happened in the aftermath of an earthquake of magnitude 9. That’s pretty much as strong as earthquakes get. The earthquake caused a tsunami. The Tsunami knocked out the cooling system of three of the six reactors at the Fukushima site.  What happened then is that the zirconium cladding of the fuel rods began to react with the water that it’s contained in. This releases hydrogen gas which creates a highly explosive mixture with air. The explosions you see in the footage of the Fukushima accident are hydrogen gas explosions.

Thousands of people died directly from the tsunami. The accident at the nuclear power plant killed one person at location. A few others, mostly elderly and ill, died later during the evacuation.

And the Fukushima reactors were built in the 1970s. Such an accident couldn’t happen with newer designs which have passive cooling features. Passive cooling means that, in the event of a loss of power, the default is that the fuel rods will be cooled. For example, in molten salt reactors the fuel is in liquid form. If the temperature in the system raises beyond a safe level, there’s a plug at the bottom of the reactor container that melts, and the fluid drains into a container to cool. This release happens without human intervention.

Basically the lesson from Fukushima is that when push came to shove and the thing that everyone was afraid of could happen actually did happen, very little happened. In my mind Fukushima is a good argument *for* nuclear power.

The bottom-line is this: Yes, nuclear power plants are expensive, and they take longer to build than other power plants. This is mainly because we don’t want to compromise on safety. But if there was a political will, it would be totally possible to speed up the process because evidence shows that other countries can do it.

None of what I said today is new to people in the industry. The solution that they’ve come up with are the small modular reactors I have talked about previously. The idea with those is that rather than building big power plants one a piece and having to go through the entire approval process from scratch each time, you build smaller reactors at a central factory and ship them to the places you need them. Then you combine as many as you want.

This makes no sense if you think of it from the viewpoint of compactness, and with that material efficiency because stacking several of them together will require more material in total, which also means more radioactive waste. However, it makes total sense if you take into account that the main issue with today’s power plants is the long construction time which itself comes from regulation issues.

Small modular reactors are not presently being produced in large numbers but some trial projects exist. Some of those have run into problems because they turned out to be more expensive than expected, so, well. I guess we’ll see how that works out.

In summary, it is true that nuclear power plants are more expensive and take longer to build than any other type of power plants. But the difference is not a factor of ten, it’s more like a factor two to three  in terms of costs, and a factor 2 in terms of time.

These factors, however, depend strongly on the country. In the North-Western parts of the world, the major issue with nuclear power is the difficulties imposed by safety requirements. It would be possible to alleviate this issue if political will would exist, and small modular reactors can also help to address the issue.

The quiz for this video is here.