While I disagree with the notion that it's a requirement to have an IQ in the top 5% to get a PhD, jobo's claim is actually somewhat supported by what little information I've seen on the topic. One
source from the 1970's estimated the average IQ of a graduate degree holders to be 125, which is right around the 5% mark. I've seen a few more recent popular articles list 130 (in the 2-3% range) as the average for PhD holders, although people rarely cite their sources, so I'm not sure where they are finding their info.
However, as these are average values, that means there are people with both higher and lower IQ's who have PhD's, so an IQ of 125 or above is hardly a requirement. I would say this matches up with my limited exposure to graduate students at two universities, plus a variety of professors I've had along the way. There are plenty of people who seem particularly brilliant and everything comes easy to them, but there are also plenty of people who struggle a lot more with the coursework side of things but work their butts off to make sure they maintain a 3.0 GPA. They also work hard at their research and you don't need to have an IQ in the top 5% to think critically and come up with interesting new research questions that will expand human knowledge.
Of course, IQ tests are biased and favor the socioeconomic groups in power that design them. This
blog gives a brief overview of bias in intelligence testing. That's not to say that IQ testing isn't a useful tool for what it was designed for as well as other applications, but using it as a cutoff for being able to achieve this or that is a mistake.