That said, the work various .net-related people have out or are working on have the very real availability or promise to lift that dependency, if not provide portability, with DB choice. See RavenDB (document DB) & biggy (Postgres 1st class). Supporting this for more traditional takes on relational store is that there is increasing experience in the .net world for applying coding patterns that can help projects pivot/abstract out the data tool dependencies.
If you have a sensitivity in this area, there are options which, when calculated with the ancillary benefits of applying existing .net experience and the great MS free + cloud tools, would pan out very well for many problem spaces.
Obviously, YMMV, but if you are doing a wide spectrum of cloud + mobile, Azure is very compelling and Xamarin can become nearly as large a burden cost-wise. That total burden still can be extremely competitive and the ramp for existing .net shops a pretty clear winner. But being a C# developer has never been so promising - and it wasn't ever that awful to begin with.
If anything, the problem of coming from the old MS-dev platform-planning and having to understand the range of options and optimizing for costs around those is the weak spot.