OpenID hasn't failed at all -- I've been noticing more and more sites allowing sign-ins using it, and even a few (primarily for tech audiences) which are OpenID-only.
Of course, there are a few sites which claim OpenID support but implement it poorly (such as news.yc), but that's not a reflection on the protocol.
I do not consider getting a tiny subset of the tech audience using OpenID a case for it being a success. In fact, it did that years ago and has gone nowhere since.
OpenID 2, which was the first version intended to be broadly useful outside of LiveJournal, was published less than two years ago. In that time, OpenID has gained support from every major identity provider and most major consumers. Anybody who uses LiveJournal, Wordpress, StackOverflow, or thousands of other sites can use OpenID instead of a traditional login.
You're making a case that OpenId hasn't succeeded yet. that's not the same as making a case that it has failed.
Of course, success is subjective and relative. I see a little success every time I log in to stackOverflow via OpenId, or get a OpenId guest comment on LiveJournal.
Of course, there are a few sites which claim OpenID support but implement it poorly (such as news.yc), but that's not a reflection on the protocol.