Right, but the fuel cost goes up over time as extraction gets harder (eventually requiring entire new types of facilities, which themselves take a decade to build), even as wind and solar continue their exponentially decreasing LCOE cost curve.

The exponentially decreasing cost curve is the part that I think people haven't actually internalized. The arguments on the pro-nuclear side are pretty much the same as they were ten years ago. (right down to 'we'll have thorium in ten years!') But things are very different now: It's like arguing for large scale investment in punch-card sorting machines circe 1960. Yes, the punch-card sorting machines can be used to manage all your bank data, but the digital processing are getting exponentially better with time and have been actually better for a couple years. But the proponents of punch card sorting machines haven't internalized the message yet.

> Right, but the fuel cost goes up over time as extraction gets harder

Uranium demand is declining, there has been almost no Uranium prosection in the last decades; Therefore it's no wonder reserves are diminishing. If demand picks up, prospecting will renew, and a lot more uranium will be discovered. This is not accounting for thorium reserves, which have never been seriously prospected.

> even as wind and solar continue their exponentially decreasing LCOE cost curve.

What you think is an exponential curve is actually the left side of a sigmoid.

Your meme pertains to exponential /growth/, not exponential /decline/. Exponential growth in a population hits a limit as a population saturates, causing the sigmoid shape. Exponential decline hits a constant asymptote and bottoms out. For populations, that asymptote tends to be zero. For cost curves, as in the solar panel case, it will be something more like capital cost of production divided by panel lifetime, which can be a very small number indeed.