This famously isn’t true for nuclear power. It just keeps getting more expensive.

The French nuclear case illustrates the perils of the assumption of robust learning effects resulting in lowered costs over time in the scale-up of large-scale, complex new energy supply technologies. The uncertainties in anticipated learning effects of new technologies might be much larger that often assumed, including also cases of “negative learning” in which specific costs increase rather than decrease with accumulated experience.

https://www.sciencedirect.com/science/article/abs/pii/S0301421510003526

And this research was done before Fukushima, which increased costs even further.