I'm saying that the PIs of most university labs would gladly pay several times the retail price of XP just to be able to continue having support for multi-thousand-dollar instruments.

But I think this nitpicking over a relatively minor point--the important part is that upgrading from XP to 7 costs tens or hundreds of thousands of dollars per computer for some people.

Edit: In case you were referring to the sentence "$10k buys a lot of...," I can see how my intent was unclear (and I was sort of talking about both small and large labs at the same time). My understanding was that there is a minimum buy-in for extended support (something like $100k per year), but that such a contract would cover hundreds or thousands of computers. That's on the order of $100 to $1k per PC.

A medium-sized university lab might have dozens of instruments costing on the order of $10k to $100k. Now, paying $100k to support $100k worth of instruments for a year doesn't make too much sense, but paying a few hundred dollars per $10k instrument does. Even at $1k per PC per year, avoiding an instrument purchase buys a least a decade of extended support. It's just a question of getting a critical mass (or, by convincing Microsoft to open up the program to smaller organizations).