It's true that you don't have to prove yourself after the first semester of grad school, but there are many "release"-like hoops you have to jump through in those 5+ years. The main one is your thesis proposal. I also had to do a plan of study a couple years before that; nowadays, my school (Northeastern) instead makes you publish a paper (and you're encouraged to publish multiple papers). There's also all the things your advisor asks you to do as his research assistant, which isn't always directly connected to your thesis work.
Your adviser, committee members, and conference/journal reviewers are all "fickle customers". And these aren't customers who came to you because they want your product, these are more or less assigned to you and you're stuck with them (unless you change departments or schools). It's true that you get a stipend, but I'm not sure that counts as "full support". And while the drop-out rate is obviously lower than the startup failure rate, it's far from zero.
Still, I agree that grad school is less scary than a startup. But don't think it's not all about pleasing others, with many non-trivial deadlines and constraints.