A friend asked me if his company ought to switch from .Net to LAMP. He's firmly convinced that in the long run LAMP will scale better. There's apparently some ongoing discussion within the company about what platform to use for an Internet-based business that expects to have millions of users.
My answer probably wasn't as definitive as he'd have liked. If I were starting a company and had no legacy, I'd choose LAMP. I believe it scales better. I agree with Peter Yared that there's no better tuned Web platform on the planet than Apache on LAMP.
That said, the choice for my friend wasn't clean. They already had a first generation architecture built on .Net and the engineers on staff were more comfortable with IIS and XP. Changing at this point will be painful and will likely be met with considerable resistance. In that environment, I think that the benefits of changing have to be significant since small companies can't afford distractions.
That's too bad, because in the long run, I think the company will be better off with Apache on LAMP. Not only will future hardware costs be deferred due to better scaling, but the operational aspects are simpler.
In the end, however, these things are not as likely to be as important for the company as getting a great product to market. For that to happen, before funding is exhausted, the better course is to stay the course and let the engineers, who were presumably hired because they are innovative and understand the problem domain work in an environment that makes them comfortable. That's the sense in which platform matters.
When you choose a platform, you choose much more than just an operating system. You also choose the pool of engineers who'll build to it and the pool of folks who can operate it effectively. Many companies are started with at least one technical founder who probably has strong feelings on this issue, but for those that are not, the choice is often made without much real thought or planning. Then, months down the road, they find themselves questioning their choice, like my friend, and find it's too late.