IBM didn't need to hire Bain and McKinsey to tell them to focus on mainframes, because they've been doing it all along. Any old timers who were around in the PS/2 and OS/2 days can recall how IBM treated the personal computer as nothing more than d-mb terminals next to the "real computer" (the mainframe). That attitude existed in the 1980s and 1990s, and it still exists today.
The mainframe-only path will focus IBM to a pinpoint of irrelevancy in the world IT market. The "hybrid cloud" and "port stuff to Red Hat Linux" approach may extend IBM's life for a while, but customers will keep on working to find alternatives outside the IBM ecosystem. To many modern businesses, IBM is simply TOO MUCH. The machines are too big, the staff requirements are too many, it takes too long to implement, and IBM wants TOO MUCH money.
While IBM tries to sell big-a-s machines for huge amounts of dough, look for the rest of the market to do everything they can to replace the hulking IBM stuff. New processor, computer and accessory designs will come out from other vendors to implement the RAS, crypto and other features that were once available only on mainframes. These features will be dirt-cheap (relatively speaking), available from companies like Nvidia. They will be available on a piecemeal basis (buy just what you want), or if you want you can buy them as part of a cloud deal from one of the successful cloud vendors (Google, Microsoft, Amazon). New software will be written for the new environment(s), and eventually IBM's stuff will be ditched.
This may take 15-20 years, but it will happen eventually. IBM's cost footprint is simply too big for most players to support.