His own thoughts about are here:
http://www.paulgraham.com/whyyc.html
http://www.paulgraham.com/ycstart.html
Whether you believe that or not is up to you, but back then there was no guarantee it would work. To look with hindsight and say it was just greed is easy for you to hurl, but not at all what it looked like at the beginning.
More generally, I know many many people (maybe literally tons by weight?) who continue to work long after they don't need to, usually because they're just deeply interested in the domain. I've seen this up close with my father, who continues to farm at 70, and with some founders who keep founding despite having more money than they will ever need. So when someone in that situation says they work on a thing because it is interesting to them, I generally believe them.
I mean, you're maybe 10% right. You take that 10% and inflate it to the point that it's almost completely wrong, though.
I’ll take my -4 for hitting close to home. The grayer I get — silencing unpopular ideas by making them grayer was his idea too — the more confident I’m right. We stopped telling you folks you were following someone who only cared about your impact on his RoI because that sort of unwashed heathen isn’t welcome here. Paul Graham, YC, and this creation of his on which we speak now has done more to distort and destroy any foundation of computing remotely paying attention to making the world better than almost anything before it, and while I’m happy to see people finally figuring out that it just might be bullshit, it’s a bit too late.