Efficiency of affordable solar cells can't increase by more than about a factor of 2 from where it is now. Even to do that, we'd have to figure out how to cheaply mass-produce multilayer cells, which are currently expensive one-offs made in labs; I'll be very surprised if that happens in 15 years.

What's more likely to happen, as I understand, is that cells of about the same efficiency level as what most people are using now, or maybe somewhat less, will get much cheaper (e.g., perovskites). But that won't be a reason to replace existing installations.