Climate reporter David Roberts interviews Jigar Shah, head of the Department of Energy’s Loan Programs Office, on nuclear power.
(1) The loan office isn’t pushing nuclear or any other sort of energy, but they are responding to demand for it. Tech companies want to build out their server farms for AI and other reasons, and they are committed to being carbon neutral. So we’re going to be doing a lot of solar, a lot of wind, maybe some geothermal. But the modeling also says we’re going to need a few nukes.
(2) Shah groks learning curves, in a way that we discussed above. Emphasis added:
Let me take us back through time just slightly, because, you know, frankly, I’ve learned a lot in this job. When you think about what nuclear did wrong in the United States, and we can compare it to Korea and other places, is that we have 92 operating reactors today in this country and no four of them are the same. So we decided that every single nuclear plant should be a snowflake. It’s like the original snowflake. So we never got good at a design and made it stable. Compare that to Canada, which has the CANDU reactor, and every one of their reactors is a CANDU reactor and they’re all the same.
Me: Nuclear construction has only been economical under programs that picked a single or a couple of designs and ran with them. This was a mistake made back during the 1970s by the US nuclear power industry: “the fatal flaw was that all of the reactor makers created this sort of, like, menu of options. And a lot of utilities thought that that meant that they should actually select different features off a menu.”
The Canadians and the French grasped the importance of standardization. Maybe the US could have figured this out as well, but we stopped building new designs after the 1970s… until Vogtle.
So basically, the way nuclear works is you always want to build two or four, not one — for a variety of reasons. Some are resiliency based, right, in terms of if something has to go down or whatever. But the other reason is because at every single site, you have certain security costs and site costs and other things, and so you want to share that over other things. And then the last issue is that you always get a cheaper price between one and four.
So the first one is always more expensive. At every site, even when we’re building the 30th nuclear reactor, it’ll be the same that the new site the first one will be more expensive, the second will be cheaper, the third one is cheaper, the fourth one’s cheaper. So it is objectively the case that if you went to a 1200 MW coal site, which by definition only has 1200 infrastructure around it, and you built one AP1000, that would be more expensive than building four 300 MW reactors, because the second, third and fourth one will be cheaper than the first one.
And so it’s better to do it that way, and then you get the resiliency benefits, et cetera. So if you’re going to build the AP1000, you need a minimum of 2200 MW of infrastructure around that site.
I’ll add that Noah Smith pointed out that after a few nuclear plants, the French encountered diseconomies of scale - the plants got more expensive. So while the first is always more expensive, we should hold off on comparisons between the fifth and the 10th.
(3) SMRs. They aren’t especially small in practice. Nor are they modular. But!..
(4) The Inflation Reduction Act is really pro-nuke:
The IRA says if you actually are building a nuclear reactor, you get a 30% tax credit. Then if it’s in an existing energy community, which all nuclear reactor sites are, you get a bonus 10%. And then if you have majority domestic content, you get another 10%. So, a lot of these reactors are going to get a 50% tax credit.