VPS hosting is a commodity business, and Digital Ocean is driving every penny out of it (they are sort of the Dell of VPS it seems). But the "cloud" is a services and consulting play I believe. A couple of the things I learned at Google was that uniform server design didn't work well, but integrated compute and services did. A Gmail cluster was a thing to behold. The challenge is that if you're building petabyte and exabyte scale infrastructure you can't have your "storage cluster" over here, and your "compute cluster" over there, ever move a petabyte of data over a 10GB link? It takes forever. So your goal has to be to build an API/Services rich environment in a high bandwidth fabric but to ruthlessly use locality as a "feature" in your scheduling.
With Softlayer's gear, and IBMs resources that is really doable. Remember that IBM is still going to have 12 - 13 billion in free cash flow. So they have the horses to build infrastructure just as fast as Google does (which calls out that expense in their quarterly reports), but IBM also needs the services infrastructure play which is something they are really really good at. I'm not counting them out until the decide to divest themselves of SoftLayer :-)