Actually, this was a puzzle at the beginning, and I'm not sure when it stopped being so (though presumably it stopped being so long ago).
In the early days, there were just two of us programmers, and we began to sense that there was a backlog of programming tasks that we just didn't have the time to get to. We proposed the Jeff that we hire a third programmer. He wanted someone with a PhD in CS from one of the top 5 CS schools. We said "Jeff, we're a tiny startup in Seattle (remember, this is 1994, Seattle was different), we're never going to hire people like that". He insisted. We interviewed several of those people - they weren't interested (as we predicted). We bugged him and bugged him and eventually he relented and let us hire a less qualified, less stellar guy, just to get some tasks done.
In retrospect, Jeff had been exactly right. We should have held out for another A+ programmer, with the skills and vision to drive forward whatever tasks they were given (or self-identified). Instead we hired a capable, skilled person who needed a lot of hand-holding and encouragement (not to worry, I think he did pretty well for himself).
Jeff knew we had made a mistake (albeit not a very big one), and asserted very affirmatively that the company would not do that again. When we started hiring editors, he wanted people with "multiple years experience at a major national publication". The two of us founding programmers didn't try to talk him down from that, and eventually he was able to recruit such people. I don't know how he convinced them to join Amazon, even to this day.
Also, recall that in 1994, "the valley" wasn't what it later became. The people amazon was hiring in the 1995-1998 period weren't looking to the valley for alternative employment, for the most part.