Power Systems: Driving More Revenue Than Initially Thought
February 25, 2019 Timothy Prickett Morgan
Any model takes refinement, whether it is something a human spreadsheet jockey puts together or it is a distributed neural network that is trained with machine learning techniques to do some kind of identification and manipulation of data. So it is with the Power Systems revenue model I put together a month ago in the wake of IBM reporting its financial results for the fourth quarter.
I did not really mean to get into it at the time. I was just going to assemble a short table of the constant currency growth rates of the Power Systems business and I just kept going back in time and wondering what this data really meant. Constant currency growth rates are interesting for month-to-month and year-to-year comparisons for a business that does business in many currencies around the globe, but it doesn’t really tell you the size of the Power Systems business. As a refresher, here is what that growth chart for Power Systems looks like:
So I went back in time and took my best stab, based on information from the analysts at Gartner and IDC, on reckoning what the quarterly revenues for Power Systems were in 2009, and I converted the constant currency growth rates that IBM supplies each quarter with the as-reported figures, which are reported in multiple currencies and converted to U.S. dollars at the end of each quarter according to the relative (and often fluctuating) values of those currencies against the U.S. dollar.
I made what was a pretty good model from this. But after getting some feedback and also giving it a bit more thought, I came to the conclusion that the initial revenue model was a little short on the external sales – meaning those that are reported as external sales by IBM when it is talking to the Securities and Exchange Commission – in a few different and important ways, some of which are easier to guesstimate than others.
The first way it was shy is just that it was simply too low for the external sales. Not much, but a significant amount that requires the model to be adjusted for 2018 and backcast all the way back to 2009. My initial model reckoned that external Power Systems sales (again, meaning those not sold to other IBM divisions but those sold to end users and channel partners) in 2018 came to a tad bit more than $1.6 billion, but I reckon now that it is more like $1.78 billion. That may not sound like much, but it is an 11 percent difference in the model, and I pride myself on being within 5 percent or less in most things. But this is very tough to do in the absence of data, and all I can say is that I think it is more accurate now based on feedback and new data.
But that is not all of the Power Systems sales that IBM does, and the picture is more complex, and this week I want to try to take on some of that complexity to present a more accurate picture. In addition to those external sales of Power Systems gear to channel partners and users, IBM also “sells” Power Systems machinery to the Storage Systems unit that is part of Systems group as the foundation of various storage arrays, like the DS8800 series disk/flash hybrid arrays, and software-defined storage like Spectrum Scale (GPFS) and Lustre parallel file systems as well as various object, key/value, and block storage engines. Back in the day, IBM used to give hints about how much of its as-reported revenues came from servers, storage, and chip manufacturing, but it no longer does this. It does talk about growth in storage hardware, so we can move forward from the old data to the new and try to figure out how much Power Systems iron, and its value, is underpinning various IBM storage. It is hard to say with any precision, but the Power Systems portion of storage looks to be somewhere north of $200 million in 2018 – my guess is $226 million, up 15 percent from 2017 levels and considerably higher still than levels in 2016. In any event, when you add that storage part of the Power Systems business in – which IBM does not break out itself – then the Power Systems division probably brought in something north of $2 billion in revenues in 2018.
Here is what the chart showing external Power System servers and internal storage-related Power Systems revenues look like together:
Those storage-related Power Systems sales are like icing on the cake, as you can see, ranging somewhere between 8 percent and 13 percent of total Power Systems sales (with just these two items, which is not the complete picture).
Here is what this data looks like if you annualize it and consolidate these Power Systems sales:
That gives you a better idea of the slope of the revenue bars. And if you like real data, here is the table of the data behind that:
If you want to really complete the picture on Power Systems hardware sales, there is one more thing that needs to be added in: Strategic outsourcing contracts involving Power Systems machinery. There are some very large organizations that have very large compute complexes based on Power iron, and in a lot of cases, they are much larger aggregations of systems than even System z shops have. And many of these customers have IBM manage these systems under an outsourcing contract through the Global Technology Services business. And when GTS buys iron to upgrade Power machinery for customers, this is not included in the externally reported figures. It is hard to figure how much Power machinery GTS consumes, and at what price, but here’s what we can say. IBM could make that price anything it wanted, any quarter that it wanted, so there are probably practices in place to assess that gear that GTS buys at a fair market value to avoid the appearance of impropriety. If you look at the annual revenues for Systems group, which includes Power Systems and System z servers, operating systems for these machines, and storage, IBM sold a total of $8.85 billion in hardware and operating systems, with $814 million of that being to internal IBM groups; I reckon that most of that went to GTS for outsourcing, and further that about half went for servers, a quarter went for storage, and a quarter for operating systems. It is not hard to imagine that several hundred million dollars in Power Systems iron was “bought” by GTS for outsourcing contracts last year. So maybe the “real” revenues for Power Systems hardware is more like $2.3 billion, and with maybe a quarter of the $1.62 billion in operating systems being on Power iron (the other three quarters comes from very expensive software on System z mainframes), the breakdown of the $2.66 billion or so in Power Systems revenue might look like this:
This is a larger business than many might have expected, and it is profitable and growing. It could be worse. And it has been. And it is getting better.