we don't do it because we forgot to do it cheap. At 3bn/unit like GE's ABWR or chinese hualongs/cap's it's a steal. At 20+bn/unit it's not that fun.
We didn't forget how to do it cheap. The ALARA (As Expensive As Reasonably Achievable) policy simply made it illegal to do it cheaply.
There's the "we forgot" hypothesis, but I think a more realistic hypothesis is the "we got too rich" hypothesis.
Construction productivity has stayed stagnant for more than half a century, while manufacturing productivity has sky rocketed and made us all fabulously wealthy compared to when the first nuclear reactors were built half a century ago.
I don't trust China's public cost numbers as much as I trust their actual capital allocation on the grid. And I will trust GE's numbers once they have actually produced something at those numbers, as pre-build cost estimates for nuclear are not believable due to their extensive track record.