Jesus Christ, this is not even close to being true and a paltry 5 minutes of research could have confirmed it. Starting from: http://techcrunch.com/page/311/, looking at stories from September 29th, we have:
SocialBakers: Still around
SweetLabs: Still around
Networked Insights: Still around
Hipmunk: Still around
Fameron: Still around
Vostu: Still around
Storydesk: Still around
Shazam: Still around
Kred: Still around
TechCompanyPay: Still around
GetComparisons: Still around
In short, this article isn't just wrong, it's dangerously wrong. 100% of the startups in this sample still exist.
I do think the 90% figure comes with a bit of hyperbole, but I also think there are more startups that “fail” than people realize. Most of these companies have such a low maintenance cost that the service continues even after the founders called it quits a long time ago. I personally know one of these companies.
By the way, the point of the article wasn't really about the 90% figure, but more about how people are seeing an article on TechCrunch as a means to an end, rather than a beginning.
If he were within an order of magnitude of being correct, then we could nitpick about methodology but he's not. Even with his revised interpretation he is wrong, just plain wrong.
