I hope you're being facetious. While (for instance) Facebook's ubiquity will insulate it for many years, come what may, raw connection numbers aren't nearly as important as local connection numbers, and "local" can mean many things, including distance on an interest graph. For instance, I joined Twitter (and Google+, for that matter) because there were interesting conversations happening there and I wanted to take part. Those networks continue to sustain interest for me, so I contribute, and that brings interest to other people, etc.
There might not be a steady state (either becoming abandoned over time, growing to monstrous proportions of popularity, or changing character enough to drive old timers away, or some combination of the three), but all the evidence for the entire history of social groupings suggest that other social groupings will pop up to take the place of any that fail and to fill in niches no longer served. It's certainly not zero sum.
Metcalfe's law was also only ever meant to be an observation (and really just boils down to a description of any network, with "value" being some nebulous term), and certainly was never meant to describe humans' social connections, if only because they cannot sustain an arbitrary number of them. Applying it to social networks for predictive purposes is just sloppy sociology, which is about as low as you can get.
