I can't seem to find any real data to back up that 35% figure. It appears to all lead back to a blog post about a survey of a self-selected group from Infoworld's "
exo.performance.network community." Self-selected surveys result in completely useless statistics.
There is some legitimate data from NPD showing that retail box copy sales of Vista were down compared to XP in the first 6 months of sales, but that's not very relevant data either. Retail box copies have always been a very small fraction of total OS sales. What we need is hard data from the big PC vendors showing exactly what they've been selling.
Given the negative utility of the Infoworld data for inferring general market trends, I'd think it's fairly safe to continue assuming that the vast majority of non-corporate computer sales continue using the default operating system, without any effort taking on the part of consumers to choose something else. Corporate sales are a different beast, naturally, since corporate IT has very long upgrade cycles during which they keep everything using the same set of disk images. A lot of them are still just barely now upgrading from win2k to winXP.