Look at the comments of any recent LAMP (used broadly) vs .NET discussion and it won't be long before Microsoft's BizSpark program gets thrown around like some great equalizer. At first glance the idea seems good - especially when put into perspective Microsoft's history. When enrolled in BizSpark you get access to Microsoft software for free, for both your development and production environments, provided you meet certain eligibility criteria. For the most part, the criteria are reasonable and broad, so chances are good that you can make use of the program. But a single, unavoidable limit on the program makes the whole thing worthless: a three year limit.
After three years of being part of BizSpark the privilege to use Microsoft software in production, for free, is revoked. Its a massive bait-and-switch and truly a stroke of sales and marketing genius. In fact, since its introduction I've seen BizSpark for what it really is: a sales vehicle. I won't claim to know what motives Microsoft had - maybe they really thought they were helping developers - but the reality is they wait long enough so switching is virtually impossible and so that you've built up a nice inventory.
Point this out to people (a lot of whom don't realize they're on the hook for a potentially massive software bill), and they'll likely argument that in three years you should be able to afford it anyways. Its true that in 3 years, a lot of startups will have died, and a few will have succeeded - but a lot (most?) of them will still be trying to make it. Success doesn't usually happen over night, or even in a year. Maybe at that point they aren't called startups anymore, but that doesn't change their need to be smart with money. People who associate startups with millions of dollars of funding and/or profit over a short period of time haven't ever been part of the experience.
Also, and this might be my frugality talking, but even if you can afford it, why would you spend $100K in software licensing when you can get the same, or better, for free? You could put that money to better use within your company, or donate it (for tax reasons, of course)
This brings up my other problem with BizSpark: ambitious developers - those most likely to have a successful startup - aren't likely to let something as trivial as learning new technologies get in their way. I say this because the other argument you'll often hear in conjunction with BizSpark is that they are already familiar with the MS stack. But startups are highly competitive and in most cases as reliant on execution as they are on the idea (there are always other cases, such as those who are able to rely on their celebrity in lieu of good execution and an innovative idea).
Let me put it another way, if you can't get very familiar with PostgreSQL on Linux in 1 day, then odds are stacked against your startup. It isn't that you aren't a good enough developer - you are, trust me, this stuff is dead simple now - its that you lack the ambition and lust for technology.
If you plan on failing within three years, BizSpark is the right program for you. Otherwise, alternatives will always be more cost effective and, anecdotally, more productive.
My advice, whatever its worth, if you really want to start a new project using .NET is simple: don't use SQL Server. Use MySQL or PostgreSQL (or a NoSQL alternative) on Linux. There are solid .NET drivers for all of these technologies. This will shield you from a significant percentage of your licensing (and vendor tie-in) costs. That way, when the 3-year hammer drops (and it will), the damage won't be nearly as bad. Oh, and if the first thing that comes to mind is SQL Server Express - than you are beyond hope.