1. To make significant profits, we need to sell services on top of our software products (this is essentially GBS and their "strong" sales people)
2. To make very good profits, we need to make highly customizable software (for example AI and BI offerings).
3. To make even more profit we need to make sure the software is tuned to the hardware we make.
If one of those weakens the entire IBM portfolio and profits weaken dramatically.
Here's problems in last 7 years tho: 1. People moving to the cloud so the hardware business flatlines. 2. Because people moved to the cloud they found replacements to IBM software. 3. At the end of the day IBM is forced to deliver professional services on top of other companies' software and hardware (and services employees are not cheap).
At some point the IBM execs must have had an epiphany that their AI offerings don't sell because they don't have a platform that sells other commodity cloud services on top of which AI components can be sold as high-priced addons.
So thus IBM decided to do what it does best --- take control of the entire stack.
With this acquisition IBM has the potential to become a next gen. cloud vendor. For example IBM has been trying to sell Bluemix as a hybrid PaaS/IaaS but haven't been very successful. The engineering team in Bluemix is weak and one way to really up the ante is getting access to top talent in the industry to do this (CoreOS team, Openshift.io team, linux kernel devs, distributed storage devs).