IBM’s Q3 in Servers, Redux: The i and p Platforms Do OK
October 27, 2008 Timothy Prickett Morgan
When first we practice to deceive. . . . . Since the beginning of 2008, IBM has tangled itself in a web of nonsensical numbers to report its financial results for its System i, System p, and Power Systems lines. I know that the top brass has given some bookkeeping reasons why it started calling things Legacy i and Converged System p instead of just Power Systems running i, AIX, and Linux. And as I have said since the beginning of this process, it is horse hockey.
Unlike the relatively few people who have picked up on this theme and just banged on a drum, I actually do some math and try to figure out IBM’s server sales and then its sales by their dominant operating system, which for the purposes of this story I still call by their old-fashioned IBM names–System z, System i, System p, and System x. IBM is still using the System z brand for mainframes, and System x is still used and includes rack and tower servers as well as BladeCenter blade servers. I have more than a hunch that my numbers are pretty good, and I am going to leave it at that.
Back to my grumpiness. Not only did I contend that IBM started using this bogus Converged System p line category in its books, which included any high-end System i sales, plus any Power6-based entry or midrange gear, plus any System p or Power Systems box running AIX or Linux (regardless of processor technology), to try to pump up “Unix” sales (and I put that in quotes for your amusement, since neither i5/OS nor Linux are Unix). Why? Because if IBM didn’t do this, it would have had to admit that its AIX and Linux Power-based server line actually had a sales drop in the first quarter of the year, which I estimate to be down 11 percent to $871 million. IBM reported that Legacy i sales dropped by 21 percent in the first quarter, but “System i” sales, by the old definition meaning a Power-based box with i5/OS or i 6.1 as its primary operating system, only fell by 12 percent, to $161 million.
IBM never gives out specific revenue figures for servers, storage, and other iron it sells in the Systems and Technology Group. But the company has, to its credit, given enough information that you can now at least reckon how much money it actually makes selling servers, storage, retail systems, and other OEM technology like chips. You have to read the presentation carefully, but if you do, you can get out a spreadsheet and start making things fit.
IBM said that STG had sales of just over $4.4 billion, down 10 percent, in the third quarter. As far as I can tell, servers across all operating systems and brands comprised just under $2.9 billion of that, at 65 percent of the total and actually up 1 percent compared to a year ago. Storage, according to IBM, was 19 percent of STG’s sales, which works out to $842 million (down 3 percent), followed up by about $177 million for retail systems (down 24 percent), and $532 million for technology and other OEM sales (down 27 percent).
So, first of all, while some brands, like System x and Legacy i, took a dive in the third quarter as I said last week, as far as I can tell from my math, servers were up a smidgen, which is good considering the economic environment. And as far as servers are concerned, IBM has one culprit in Q3: System x (including BladeCenter).
I have spreadsheets that go back to the late 1980s tracking IBM’s server sales by product line from a number of different sources, and over the years I have looked at this data and tweaked it here and there with some help from various sources (including IBMers) to build my own models. Once you have four quarters of estimates, you can pretty much keep going indefinitely with the model as IBM announces results.
That’s why I think it is very likely that IBM made something like $775 million in System z mainframe sales in the third quarter, up 25 percent like IBM said, and made something close to $939 million in System x sales, down 18 percent just like IBM said. Once you know how much server money IBM raked in ($2.88 billion), you can subtract those two numbers and now you know how much Power Systems sales IBM probably had–and that works out to $1.17 billion in Q3, up 6 percent by my math.
Coming up with a System i-System p split for this is a bit tougher, and it involves some hunches based on past trends and current market conditions. My best guess is that what we would have called System i sales came to $247 million in the quarter, up 4 percent, and what we would have called System p sales (by which I mean any AIX or Linux box, regardless of chip) came to $919 million, up 7 percent. The crash in the Legacy i category (down 82 percent, but against a small number) was offset by the increase in the Converged System p category thanks to the addition of the new Power Systems i products and high-end System i iron from the Power Systems division. (My assumption, and I think it is a fair one, is that the silly categories basically wash out in this quarter.)
Now, here is where the tangled web parable comes in. Let’s say that I am wrong, and “System i” sales were lower in the third quarter. Maybe high-end System i and Power Systems i sales fell because of the skittishness in the banking industry, where the AS/400 has had a lot of play at regional banks for many, many years. Or maybe manufacturers froze their budgets. Maybe “System i” sales were as low as $150 million. Well, that would mean System i sales would have fallen by 37 percent and “System p” sales would have risen by 18 percent–and the aggregate Power Systems and the Converged System p numbers would not change one bit. And if something like this did happen, then IBM artificially imposed categories that made its Unix business look weaker.
Here’s a pretty picture that shows the trend data since the first quarter of 2006:
So, the i platform did not do so badly as far as I can tell, and the AIX and Linux platform did OK, too. The real downer in the quarter was IBM’s X64 server business. With Lenovo starting to sell clones of IBM’s System x tower and rack machines (see Lenovo ThinkServer: The Sales Pitch Sounds Familiar for more on that), and IBM’s STG under margin pressure–pretax income was down 21.6 percent to $223 million in Q3–you have to believe that IBM is wondering why it is building any low-cost, low-margin servers at all.
The answer, as Hewlett-Packard shows, is to have a vast base of customers that buy your gear and to cross-sell and up-sell the heck out of them. But IBM’s abandonment of low-end and then high-end printers, networking gear and network services, disk drives, and PCs over the years shows IBM doesn’t have a lot of patience. And I can totally see IBM deciding that the only boxes it needs to actually make are high-end Power and mainframe boxes. The company hasn’t made its own X64 servers except the BladeCenters and its very largest System x Xeon MP boxes for many years.
Here’s the important thing to see: As best as I can tell, “System i” sales have increased sequentially in all three quarters of 2008, just as they did in 2007. The bad part is that the System i had a decline in the first quarter, but you have to expect that as everyone was waiting for a Power6 lineup to come out in March or April. No major changes will come to the line in 2009, in terms of hardware, unless IBM has some other reasons for doing it–maybe a Power6+ speed bump, but that is a good thing. So there will not be a big transition to cope with. While the economy is not doing so hot, and business partners are seeing deferrals on deals in Q3 and Q4, if the economy stabilizes, people will find budget money in 2009 and IBM will probably have a better year in the i platform. And, hopefully, that stupid Legacy i category gets buried in Q1 2009, when Power5 and Power5+ machines are out of the catalog.