This is interesting, and I'm a big fan of the Kindle, but I think the statistic somewhat overstates things, because there's a bias in the data: when someone has a Kindle, they're going to purchase ALL of their kindle material from the 200,000 items that are available for the kindle. When someone doesn't have a kindle, they draw their reading material from the long tail.
Imagine that there were just 10 titles available for the Kindle - I bet that 70 to 90% of all copies sold of those titles would be kindle titles.
Plus, those are 35% of Amazon's sales of those books. If you have a Kindle, you are defintely buying from Amazon; if you're buying paper books, Amazon is just one of your suppliers.
Even if those ten titles are at the top of the NYT best-seller list? There should be a limit to how many sales are attributable to the Kindle version, which would keep the percentages lower.
It would be nice to see the unit sales percentages for books with a Kindle version, instead of averaging the percentage across all books - that would weight it more toward the high-volume titles.
Imagine that there were just 10 titles available for the Kindle - I bet that 70 to 90% of all copies sold of those titles would be kindle titles.