Last spring, after Arvind Krishna took over as CEO, we had a Twitter Chat with Constellation Research analyst Holger Mueller where we seconded his idea that IBM should separate its software and services businesses. It gave us the excuse to convene the type of bull session that we have when we meet on the conference circuit, but in this age of COVID, Holger was generous enough to reconvene this as a very enjoyable 45-minute fireside chat on YouTube, where we both tossed around our ideas on the merits of such a split.
OK, we're not going to call it telepathy.
Last week, IBM announced plans to actually do at least some of that. It would separate the Managed Infrastructure Services unit of its Global Technology Services division (fancy speak for the core of GTS) into a new public company ("NewCo"). Wall Street liked the idea, as IBM shares jumped 6% on the news. Conventional wisdom is that this would allow IBM to more effectively leverage its Red Hat acquisition. It paid $34 billion, not to add $1 billion of revenue to a $77 billion business, but instead to capture Red Hat's mojo to transform IBM. Freed of at least some of its services business, the rationale is that at least one of the distractions toward making the Red Hat deal the reverse acquisition it wanted has gotten cleared away.
On the numbers side, the GTS business, with $19 revenues and $60 billion backlog, has clearly been the lag on IBM's growth. It's a mature business, one with low margins and low growth, but it takes a lot of resolve for a company to give up a unit that might otherwise look like a cash cow.
Beyond the slim margins of the services business, separating GTS could be one step toward helping IBM get rid of the internal cognitive dissonance between services and software. When it puts its ducks in a row, IBM can come out with phenomenal software. Cloud Pak for Data includes several SaaS services, including Watson Knowledge Catalog for data discovery; Watson Studio for developing and operationalizing machine learning models; and Watson OpenScale for tracking the outcomes of those models. In a hybrid cloud environment, these are exposed as services rather than complicated software that requires consultants to set up.
But when we asked IBM if there might be a vendor-managed option for Cloud Pak, much like AWS Outposts or Oracle Cloud@Customer, the answer was that this would be a GTS engagement. Hopefully, this is the type of thinking that will end once GTS is separated from IBM's new core, which would be the software business. For the hybrid cloud business, IBM needs to embrace self-service rather than service.
In our fireside chat, Mueller and I agreed that relying on the services business for such a large chunk of revenue gave IBM Software less incentive to make its software less complex to implement. But the pull of the cloud creates a centrifugal force the other direction, at least if the goal is building a SaaS portfolio.
There are other issues to being a combined software and services business. We've seen over the years that IBM, like many services-oriented companies, tended to operate as fiefdoms, and that creates its own internal channel conflict. That conflict comes from localized priorities. Consulting groups are incented to find the best fit for the specific customer; in a pure services business, the main downside is the hurdle to scaling best practices or jumpstarts that can be replicated to other consulting teams. When you combine a services company with a software company, you still have that problem, and then you add the question of priorities: should the consulting team recommend a custom or third party solution at the expense of a solution from the software side of the business? Should the software folks recommend throwing integration business to external channel partners to build up a larger third-party ecosystem?
Not all "channel conflict" need be bad. The major cloud providers have proven that in an era of as-a-service software, coopetition can indeed be a very profitable strategy. For instance, while Snowflake competes with Amazon Redshift, AWS still gets compute dollars for each EC2 instance consumed through the rival service. For IBM, there is promise that Red Hat OpenShift could provide a similar path to monetization with the best of rivals.
With the spin off, IBM is still hanging on to $8 billion of revenue from the cloud services side of GTS that is closer to IBM's mission. And there will still be the $16 billion from the more solutions-focused Global Business Services (GBS) unit. So, the GTS spinoff won't be a panacea for purists where IBM suddenly converts to a pure software company. It simply gets rid of the slower growth, less strategic part of the business. It still means that IBM must continue to work on sharpening its focus. As noted above, that means going all in on self-service for its hybrid cloud platforms. Shedding GTS may eliminate a distraction, but it's just one step toward the transformation it is seeking.