I think it's more a matter of there not being any VCs who specialize in fusion. There just isn't the deal flow to sustain any VC specializing in it, and so far the success rate has been 0%. Of course, because there are so few fusion startups (lets see, I can think of less than 10 in the past 25 years), its hard to know what the real statistics are: is the chance of failure 100% or really just 99%?
From the point of view of a fusion startup, if you try 100 software VCs, you might find 2 or 3 who will be interested. However, you will wait forever if you want to find a VC specializing in fusion.
The NIF has never reached "break even" in the sense that most people might understand that phrase. Orders of magnitude more energy went into the system than could have been extracted from the resultant fusion reactions. However, post-analysis of a bunch of shots showed approximately two where, if you accept a model for the amount of laser energy that was incident on the target, somewhat more than that amount of energy was produced from nuclear fusion. Several other shots in the same series, under the same conditions, failed to produce this "break-even" energy. This is all in the actual papers, not, of course, in LLNL's press releases.
I would guess either personal interest or PR, rather than a solid return calculation. YC doesn't want to be seen as the "web startup incubator", even if that's where they can best maximize ROI. And besides PR, the principals likely don't want to be that, either, but have loftier goals. Which is perfectly fine; maximizing investor returns is rarely a good way to do anything important or interesting.
why do vcs with a specialty in software suddenly stray into completely unknown territory? nothing left to fund in software? rarely does this end well.