Amazon was hardly a startup 5 years ago. And we're talking about a service and a space that's difficult to replicate and get into. S3 had resources most startups won't have to get that 5-year lead.
* who, even worse for their case, already had the infrastructure, resources, engineers, etc. to get it done.
The statement should say that startups with innovative, hard-to-reproduce products have nothing to fear; you'll either produce a product that they'll choose not to compete with/can't compete with (Twitter, Facebook) or get bought by them.
How many startups get killed because they can't compete with the big guys, though? Here's an example: http://family.go.com/assets/bubbleshare/
I find glowing reports about it from TechCrunch, then.... nothing, it just dies.
So what happened?
BubbleShare was a Toronto tech darling started by the always brilliant Albert Lai. They were WAY ahead of their time.
All I'm comfortable sharing here is that they received buyout Offer A, which Albert rejected in favour of doubling down for a better deal. Unfortunately fortune did not smile and BubbleShare ultimately accepted Offer B, which was significantly smaller than Offer A.
The site was sold to Kaboose, which is a Canadian family content company. As usually happens when startups are acquired, the key talent left (Albert started Kontagent) and innovation on the site halted.
I'm not an analyst, but these things usually distill down to "too early / too early / too late". I'm not sure that people were ever lining up to pay them money to use the service, and that might give you pause before going down the same road. Many products are cool but aren't solving a problem causing paying customers real pain.