I don't know a great deal about Power7, but I've read interviews with IBM engineers who've said that they've taken lessons learned from Cell and incorporated them into the Power7. We know already that Wii U's processor is based on the Power7, and yet still retains backward compatibility with Wii games, so it's probable that they could do the same with PS4.
Given the critical importance of the SPU's in Cell, which developers have to code specifically for, I'd think that any new variant that incorporates Power7 core changes would still need to keep the SPU's. It's possible that they may be able to adapt the out of order execution units to the primary core and keep the SPU's (maybe even adding a few more), or they could introduce logic to map calls for SPU functions to equivalent (or alternatives) on a more native Power7 chip, but I'm not sure what would be the more cost effective (or compatibility-effective) method.
Looking at the huge quantity of content now available on both Xbox Live and PSN, though, I just can't see them moving to completely new architectures right now. There's just too much work that they'd undue if they lost compatibility with the past 6 years worth of software that's already deployed. Before PS and Xbox are anything else, after all, they're *products*, and businesses sell products to turn a profit. They know there's money in this content library, but only if it continues to be usable once today's consoles are yesterday's news.
I hope you won't see that as a bad thing, though, I certainly don't intend it as negative judgment :).
