Mention this product and most folks will go "oh give me a break". But it is a real product and it has some unique qualities that put this kind of high-powered computing within reach of many businesses. Think data mining, advanced design on a scale less than Boeing and Airbus and yet greater than any level you've been able to afford before. I'm aware of one acedemic application being considerred in biochemistry. The list of potential customers is very large.
First off, while it's not as efficient as Linux, Solaris and Unix cluster solutions, it can be managed by a tech with decent skills and a couple manuals. No, really! I have that on good authority. And while it's not the whiz-bang solution for "big iron", it has a "sweet spot" of 4 to 64 way machines. For the price and ease of administration, many companies will gladly sacrifice some speed and power because they'll still have much more than ever before and in a form they can use.
Second, while developing for WCC03 is not all that close to doing it for the Window server and desktop systems, Microsoft is rapidly developing modifications of Visual Studio to accomodate those who need to write for this platform. That is just about everybody. There is very little of what we think of as "ready to roll" software for this sort of setup available. Nearly everyone will be using custom applications. Even with that, Microsoft is developing some applications in an attempt to make the core product more attractive.
I never did look at the price. I suspect it's one of those "if you have to ask, you can't afford it" things. And there's no way i could run it here. My little cluster of 486s wouldn't cut it, even if it was still in existence. (This is also a thinly veiled hint to Joe that I could use a few 8 or 16 way Opterons for my birthday, which is coming up.)
It's new territory for Microsoft and they appear to have a product which will appeal to a market sector which did not previously have access to such tools. As such things go, it's the only nearly ready solution to a lot of problems.