The better option, to me, sounds like to use Linux, which you can set up today on a honkin machine if you so choose. At every level tested, the Linux solution is cheaper and more stable than the MS solution. There's no reason to assume this will change, especially since Linux has proven to scale better than Windows at lower levels.
Would you rather be married to Microsoft or Oracle?
That's easy, the Oracle/Linux approach. With SQLServer/IIS/Windows you're locked into a single vendor for all products, and that vendor is known for stability problems. And expect to be forced into an upgrade path -- .NET -- that will lock you in even tighter to that single vendor. The one SQLServer db I have to deal with creates 80% of our DB problems.
With Oracle/Linux, you have a wide variety of choices of Linux vendors, a wide variety of web server choices, and Oracle is perhaps the most stable big DB around. The solution is more scalable without a doubt.
The single biggest thing to avoid as a businessperson is 'single vendor lockin'. Because then that vendor owns YOU. You're at the mercy of a single corporate entity. For example, if they decide the next version of their product -- like Longhorn -- isn't going to be backward compatible and you have to pay for all new copies of the same software, you don't have much choice.
If you just went with 4 of the Linux clusters tested here, total cost $9,522,184, would yield you 553,348 tpc.
You're well over 500 tpc, and saved over a million dollars.
And have a more stable OS, and a far more stable DB.
And that's not even going with a more honkin machine. Going up on the hardware is likely to provide even more cost savings and better performance gains.
I've never known anyone who's worked with both Oracle and SQLServer who'd reccomend using SQLServer for enterprise work. Of course, the people with a financial interest in MS will disagree, but what else would you expect a salesman to say?