The point for nuclear isn't just to cover demand curves; it's that to decarbonize the entire economy you need to both switch current electricity production to low-carbon sources and increase the capacity by 3-6x to cover all other energy requirements (including sucking carbon from the air to counter chemical processes we can't decarbonize).
Nuclear has to be part of the solution.
The other thing that will happen is that energy-intensive industry migrates to lower latitudes where there's less seasonality. Sorry, Europe.
I don't think nuclear advocates have gotten through their heads just how dire the situation is for nuclear energy. It is just not anywhere close to viable now, or even needed, and you do yourselves a disservice by concocting flawed arguments that let you think otherwise. Nuclear is going to survive only by large reductions in cost (and those costs are not due to some conspiracy of your political adversaries; they are due to the complexity and difficulty of the technology). Otherwise, it's going the way of vacuum tubes and magnetic bubble memory.
[1] http://energy.mit.edu/research/future-nuclear-energy-carbon-...
> as it builds toward tens of TW of installed global capacity it could get there
And that's kind of the other problem. How much land is that?
Per this perhaps dated 2013 report[0],
It takes "3.4 acres of solar panels to generate a gigawatt hour of electricity over a year." Unclear on if that's only in sunny desert locales like California/Nevada, or average over the US, or whatever.
Anyway, if my math isn't radically incorrect, that comes out to 114 kW of capacity on average, in whatever these test conditions were.
Just 10 TW of capacity would require, uh, 1.2e12 m², or 13k "survey townships" (US land units get weird), or 466k "sections" (weird US land unit that equals mi²) or 298 million acres.
For some comparison, California is 164k mi². Texas is 269k. You could (hypothetically) cover the entirety of California and Texas with solar panels (433 mi²) and you still would not reach 10 TW of average production.
Admittedly, your scope is global and these examples are US states. But it is still a ton of land. Perhaps efficiency has increased slightly since 2013, but by a factor of 2 or more? 10? And how much space will we need in landfills for EOL panels?
Global demand was about 22,100 TWh/yr in 2011 (random figure), so say it's doubled since then; we'd still need something like 233k mi² of these panels to supply the world, assuming perfect capture from season to season and overnight. That's most of Texas in (2013) solar panels...
And no, an experience curve never reaches zero. It's a power law in cumulative production.
Unlike sunny or windy days, nuclear power is consistent 24/7 and output can be modulated (albeit, with delay) by raising and lowering control rods. Batteries (electrical or mechanical) can store extra nuclear-produced energy and use it to power the grid while output increases to match demand, and vice versa.