Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

That's self-fulfilling. Solar was expensive. Then they built a lot of them, they got cheaper (+ better tech obviously). We know basically nothing about the learning curve of nuclear because we haven't met the conditions for the learning curve to apply, except maybe in China, where it seems to hold.

I just don't understand why everyone is so confident that nuclear power is the one thing that doesn't get cheaper with better technology and scale. Everything that is built is old tech, and, except in China and south Korea, built so rarely that the workforce is basically brand new every time. Of course they're expensive(in the West), they're one-off large engineering projects. All evidence points toward China, south Korea being able to do it cheaper and faster, so we should look there when there's sufficient (trustworthiness and quantity) data to test scaling laws. All signs point to cheaper and faster, but even then, the numbers of builds are relatively small, so you'd expect scaling gains to be modest. One of the main points of SMRs is to harness the effects of large numbers.

Why not at least try? Why are we so confident it will fail before even trying with modern designs?



seems like a similar narrative to EV's back in the 90's - battery tech is not there, the economics don't work, no charging infra etc. Cut to now and that narrative has flipped. We might be in the same adoption curve cycle with nuclear.


Even if that's the case, we don't have 30 years to figure it out. I'd keep a little money for continued research, but otherwise go all-in on solar.


Look at the cost per GW for the Bruce and TVA ”SMR”.

They are in line or worse than Vogtle, and they haven’t even started building.

Nuclear power has been experiencing negative learning by doing for half a century.

It is a known entity. All in all an extremely complex coal power plant.

New built regular coal power plants are not competitive in 2025 either. Now make it even more complex.


You're judging a new technology on the cost of a damn pilot?

Again, no to the negative learning. The learning curve theory has requisite conditions. They aren't met, except maybe in China and south Korea, where it seems so far to apply, if modestly.


How can a 70 year old mature technology be a ”pilot”?

Lovely that it takes an enormous corruption scandal for nuclear power to ”work”.

South Korea’s latest plants have only taken 10 years with correspondingly increased costs.

https://www.technologyreview.com/2019/04/22/136020/how-greed...


Thermal power cycle infrastructure has a pretty high lower bound on price even if you’re ignoring the cost of containing a fission reaction.

It’s not that there wouldn’t be learning curves, it’s that we’ve already seen the learning curve for the easiest part of the stack, and it’s not competitive.

I’m all in favor of continuing to throw VC money at small nuclear, to be clear. The regulatory situation everywhere outside of Texas is so bad that there will be niches where “power plant arrives on a semi, no external interconnect required” is a great business proposition.

But if we had a sane regulatory environment for building transmission and interconnect, nuclear wouldn’t get enough custom to ever climb the learning curve.


Look at what China actually builds. Nuclear power is declining in their electricity mix.

> One of the main points of SMRs is to harness the effects of large numbers.

How many trillions in subsidies should we waste to try scale when the competition we also funded decades ago already delivers?

We already gave a try building nuclear power again 20 years and at the same time invested in the nascent renewable industry.

The nuclear investment evidently did not pan out.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: