Watson is just a brand name attached to tools for applied statistics. Sure, hardware and datasets have got bigger and faster opening up a few new opportunities, but practical value continues to elude most of the more complex applications. How much money has your corporation blown chasing Big Data for much promised but little delivered returns? With R or umpteen other open source tools readily available why would you pay a ton for licensing plus staff, when one can pay just for the staff to get same results? Python/Numpy/SciPy are popular for a reason!
The last 10-20 years of R&D at IBM has taken them nowhere. PPC used to be a moderately successful line of business. Mac G4 & 5, PlayStation 3. Power9 is a fantastic architecture, but sales are weak and takers are few. A deal akin to how X86 exploded in the 1980's might be what's needed to make it competitive. A Power9 desktop at sensible price (say £1000) would have me jump like a shot off X86. Can't help but also think there's a market for the well priced bedroom computer too. We all love our Atari's and Amigas!
IBM Mainframe persists in a few environments. Their desktop tech has almost entirely been sidelined.
I would suggest failures to market Applied Statistics are now forcing IBM to use their buying power to find a few (guaranteed) revenue streams instead. RedHat, probably amongst others to come.