Look at France in their first build out; subsequent reactors of the same design cost more than the first. And France was considered a success by most. The few buildouts that have seen cost decrease have never produced learning like you get in manufactured products.
I wish I had access to my bibliography right now for the papers on this, sorry I'm making statements without the research to back them up. But those that have looked into which technologies develop learning curves find that the learning curve is found early, and it tends to be tech wheee lots of the same small thing is made, with minimal customization at install.
Nuclear is inherently a big tech, with some customization for site install. Combined cycle gas turbines found some efficiency of scale by being small enough that they could have standardized plant designs that don't require design changes for the site (though it does limit which sites can install the designs). SMRs might change this for nuclear, but what people are calling "SMR" today is still is as big as 300MW, and it's hard to see how that could be standardized and manufactured. Nuclear does not scale down very well, which is why all these places are trying to pass off really big reactors as part of the SMR hype wave.
Here is the most well-known: https://www.sciencedirect.com/science/article/abs/pii/S03014...
There is nothing "inherently big" about nuclear
1. Scales with square or cube laws.
2. Are essentially fixed.
A 100 MWe reactor requires a very similar amount of security as a 1600 MWe reactor, maintenance intervals are similar and so on.
Thus, it has been evident from the very first days that bigger is better, from an economic perspective because you get both a more efficient reactor and more output to spread the huge fixed costs on.
The problem has been that even the bigger reactors are hideously expensive. Nuclear power never breached the cusp where it would be worth it and today the cost to build new renewables are in the same range as the operations and maintenance cost of existing, paid off, nuclear plants. Let alone trying to build one.