Wikipedia does not need to be globally consistent like Betfair does and the ratio of writes to reads is nothing like 10%, I'd guess at one write per million reads or less. There are several pretty obvious ways to architect a site like Wikipedia for effectively unlimited scalability. The main trick is that it doesn't matter if a page is slightly stale and you can queue edits in the backend for quite some time (many seconds) without severely harming end users. Given those constraints it really isn't rocket science given the plethora of amazing tools we have to hand.
What I'm NOT saying is that I could build it in a weekend. It would clearly require a few teams of skilled engineers to put it all together and, crucially, operate it. My initial comment was in the context of Wikipedia having 100 engineers, and I think it's reasonable to say that a team that size is easily capable of such a feat.