Having spent a 20+ months working with Thesis and Revolution themes, I've been tearing them apart and not feeling like I could actually build my own.
I'm not really enthralled at the prospect - nor do I really have the time - but there are certain advantages to building my own - like optimizing the code to be light and effective SEOwise. And there are certain advantages to buying one or hacking one - less time to implementation.
Having done this kind of thing for 20+ years, I build my own (stealing new ideas from time to time, of course--though those are few and far between) and that's because I have already built a formidable kit of html code/themes. These days the eye candy on the web are only variations of existing themes. After all, there's a basic functionality that is necessary for it to be HTML (any version) or it becomes JS ASP NET or other...
We have evolved from themes, to customizing themes to just designing and coding themes ourselves. We recently realized that designing and coding custom themes actually saves time over customizing pre-fab themes. Personally, I won't go back to tweaking themes.