Price(size) = fixed + K*size^b
Where b is almost always less than 1, can go as low as 0.6.
So building 6 small reactors will be more expensive than building a single reactor 6 times the size.
This is the difference between vertically scaling (bigger thing) vs horizontal scaling (multiple smaller things).
Horizontal scaling will almost always cost you more.
We build so few gigawatt-size reactors. In the US only 4 came online since 1990. For most of the construction crew of such a reactor, that job is the first of that type in their career, and the last. They get trained on the job, and that (ultra-expensive) training is then never used again.
Contrast that with the building of naval reactors, which are essentially SMRs. The US builds about 1 or 2 per year. In factories. The workers there participate in the building of numerous reactors during their careers. For each reactor, the majority of the crew has already done the job at least once before, and they know they will continue using their skills for many years to come.
The cost of naval reactors is classified, but one could infer from this congressional report ([1], page 6) that an A1B reactor cost about $400 MM in 2011 money. Adjusted for inflation, that would be $560 MM today.
A single A1B reactor is equivalent to about 3 NuScale modules, so 2 A1Bs would be equivalent to the 6-module package that was canceled here, because its cost ballooned to more than 9 billion. The US Navy is able to procure them for less than $1.2 BN.
So, no, the economics don't fundamentally doom SMRs. It's just that you can expect the numbers to start working out only after you enter serial production. The first-of-its-kind costs are always high.
[1] https://www.cbo.gov/sites/default/files/112th-congress-2011-...
I am not disagreeing with the idea of Wright's law. The US Navy very likely benefits from it, but NuScale does not, and possibly never will.
However, a) this does include other prices such site licenses and environmental assessment.
b) Wrights law models a decrease with every doubling in units produced. At 1-2 a year, we're looking at a very long time to get several doublings in. NuScale meanwhile has 0 doublings under their belt. And NuScale is not alone in this field, there are dozens of companies, each one promising the same thing (SMNRs). The Navy has a single source for their reactors, NuScale has lots of competition and has to share the market, leading each of company to get even fewer reactors built per year.
c) Someone has to pay those early costs. For the Navy, there are not many alternatives; there are no PV or Wind aircraft carriers. It's nuclear or oil. The same is not true about power generation.
d) As you say, each navy reactor is about 3 NuScale modules. Why does the Navy not build 3 smaller modules and get even better scaling? Why not 30? I argue they don't because of the costs associated with horizontal scaling. NuScale meanwhile doesn't have a power budget they need to hit, they are trying to produce baseload power, and the bigger they make them the more economies of scale the nuclear reactor benefits from.
That's a good point.
Actually the Navy does build such smaller modules. The Navy uses 2 types of naval reactors, one for Virginia-class submarines, the S9G, which is very close to one NuScale module, and one for Ford-class carriers, the A1B, about 3 times larger.
Each Ford-class carrier uses 2 A1B reactors. Why not 6 S9G? I don't know. But then why not a single reactor twice as large as an A1B? I don't know either. It looks like there are some tradeoffs.
In any case, the fact that the Navy can build economically SMR-sized reactors is encouraging. Of course, the Navy can use technologies that are not available on the civilian market, for example their reactors use weapon-grade uranium. So it's not completely an apples-and-apples comparison.
Still, NuScale has produced more than one million pages of documentation to get their SMR design approved by the NRC. One can imagine they also put a lot of thinking in the economic viability of the project. One way or another they convinced a large engineering, procurement and construction company (Fluor Corp) to buy them.
Just because the pile of sh*t is massive doesn’t mean there’s a pony somewhere underneath it.
[1] https://www.nrc.gov/reactors/new-reactors/large-lwr/col/auro...
The NRC denied them, without prejudice might I add, a license based on an evaluation of the technical merits and safety of the reactor, not on its economics.
But, imagine that whenever you want a new car, the car company sent a crew to your house and builds the car in your backyard. All the parts needed will be shipped to your address, and workbenches, tools, machinery, too. They put the car together, paint it, dry it, and voila, the car is yours. Would it surprise you if it was 10 times more expensive than if it came from an assembly line?
Not much to do about a), so if this technology were to succeed, the government would have to play kingmaker and pick only one company such that they actually go through enough units to drop down the price curve. This might work domestically, but I foresee political difficulties when trying to convince other countries to abandon their homegrown tech in favour of yours.
My other points still stand, especially the one about competition.
But, NuScale has no experience building anything. It's just a startup. It's a miracle they got their reactor design approved.
BWXT however is the company that builds the Navy's naval nuclear reactors. They are cooperating with GE and Hitachi [1] to build a 300 MWe SMR. All 3 of these companies have plenty of experience delivering on numerous types of projects, including nuclear ones. I do think they have a chance.
[1] https://www.bwxt.com/news/2023/03/21/BWXT-Awarded-Engineerin...
The real total cost of exploitation of a marine (think: cooling) military (think: many hands available, and a quite specific way to manage dangerous equipment) reactor is also very difficult to establish.
Even the cost of civilian research (from the 50's to 2012) is very difficult to assess («les données recueillies ne sont pas toutes homogènes et leur fiabilité n’est pas totale.»).
Ballpark figures (page 270): 288 billions euros (value: 2010) invested, 118 of them being directly production-related investments (maintenance is not accounted for), with severe methodological restrictions («Il est difficile aujourd’hui de 'reconstruire' l'histoire du financement»), tackled by using conventions inherited from the nuclear industry.
It makes sense when the monolith is very complex to certify, so having discrete parts that are individually certified is more expensive only in theory when you dont factor in the exponential certification complexity. which i guess is high for nuclear related parts.
Historically each nation nuclear certification authority developed its own set of requirements.
There are links, and a push to federate efforts, however it remains true.
Case in point: the Olkiluoto-3 EPR reactor isn't exactly the same as Flamanville-3, the one being built in France, because the Finns wanted/needed various different thingies. Any of such requirement may have a significant impact on the design, and therefore impede the SMR approach.