How Big Is the OS/400 Ecosystem?

by Timothy Prickett Morgan

As the editor of The Four Hundred for eight years, then AS/400 Monday Morning Update for three years, and now the second incarnation of The Four Hundred for nearly four years, I get asked a lot of really hard questions about the OS/400 installed base. And I really like them because it is fun to try to figure things out. But, sadly, there is just a lot less information available about the OS/400 installed base and its larger economic ecosystem than there has been in the past.

I was asked recently by a prominent OS/400 software supplier how large the market the company plays in was and how it was doing. And I realized that no one has an answer to the salient base and economic statistics that make up what I like to call the OS/400 ecosystem. Part of the reason is that the IT ecosystem has rapidly expanded since I have been watching this market, while at the same time the OS/400 ecosystem has become considerably smaller. There are a lot of reasons for this, but the shift to open systems in the 1990s and the porting of OS/400 application software to Unix and then Windows platforms and the advent of whole new applications for Unix and Windows platforms that compete with OS/400-based applications is greatly the cause of diminished sales of OS/400 servers. So is the continual improvement of price/performance in the AS/400 and iSeries line. Even holding server shipments steady--if IBM could have done that with the AS/400 and iSeries line in the past several years--means losing ground in this cut-throat server market and the related hardware, software, and services ecosystems that are part of a complete platform's overall ecosystem.

Back at Penn State, when I was studying aerospace engineering, I had the great joy of having my first two semesters of physics at the ungodly hour of 8 a.m. three times a week. I am not a morning person now, and I certainly wasn't as a freshman in college. I think that getting up at 10 a.m. is a fine idea, because the sun is warm and no one is around. To have to be awake at 8 a.m. was bad enough, but to have to be alert enough to do physics was even worse, even though I loved the subject. (And still do. I am puzzling out superstring theory in my copious spare time right now thanks to Brian Greene's The Fabric of the Cosmos.) To top it all off, my physics professor was from England, complete with the thick glasses, the wool tweed jacket, and a midlands accent softened by his own years of education in his home country and his years abroad. And he was a very happy physics professor, and he smiled a lot and laughed a lot and he was clearly thrilled to be there. We were mostly tired and sometimes hung over, if I remember correctly. But this professor--heaven help me, I cannot remember his name--taught me some important lessons that I still use decades later.

First, he never worked out the problems after he explained a theory. He made us do it, even though we were always new to the subject. Second, he made one of us, chosen randomly, come up in front of the class to go over homework sets. You never knew when you were going to be chosen, so you always came to class prepared. And if you didn't have your sets done, you had to still get up in front of the class and do it in real-time, wasting the professor's and your classmates' time and perhaps making a fool out of yourself if you screwed up. And third, before you started a new problem with a new theory, he made you take a guess at what an answer might look like, explain the assumptions you were making, plug in some estimated numbers, do the math out loud to the class in real-time, and come up with one or more estimated answers before you actually solved the problem by going through the motions and cranking the equations. This is a kind of error correction, I learned. And it helps you arrive at reasonable answers reasonably fast.

This last bit is the part I still use today. I internalize the lessons people teach me, sort of like you programmer types make use of subroutines in a program. (I have had so many teachers, and inside this crazy mind of mine there are a lot of subroutines running.) I can't remember that professor's name, but I can still hear his voice:

So, how big is the OS/400 ecosystem, Mister Prickett Morgan? Are you trying to sound British with a name like that? We kicked your kind out of the country in the 1600s, you know . . .

Well, we actually left willingly, Cromwell being something of a bummer.

But any way.

Estimate one is based on history. Because I have been doing this so long, I have mountains of data squirreled away in old paper copies of The Four Hundred. (Yes, we used to print the newsletter and actually put it in snail mail.) Back in November 1993, I was hooked into the competitive analysis people in Big Blue's White Plains, New York offices, and if you built some market data that made sense, they would take a look at it and tell you where the numbers were wrong. At that time, the question I was trying to get an answer to was a little different, which was how much revenue does the AS/400 customer base generate for IBM? The answer was $14 billion, and the pie chart I made looked like this:

Figure 1: Right now [which was back in 1993, not 2005], AS/400 software represents a very small part of the AS/400 division's revenues, but it is a very profitable business for IBM. These numbers are estimates based on market information supplied by IBM. Maintenance figures include fees for System/3Xs and AS/400s.

(If you want to read the story that the pie chart above came from, I have reprinted it: TFH Flashback: Critical Mass, November 1993 issue.)

OK, at that time, the combined sales of AS/400 servers, disks (mostly external then, remember), other hardware, and systems software was a whopping $7.7 billion, and the IBM AS/400 ecosystem (including PCs and CRTs) was $14 billion. At the time, IBM had a big AS/400 channel, and there were lots of AS/400 software houses providing services and applications--something like 8,000 business partners and 20,000 applications, if you remember--but IBM still got a lot of the sales and had a pretty impressive sales force. Given all of this, I think it is safe to say the OS/400 base generated about $20 billion in total ecosystem revenues in 1993. (The AS/400 platform was never very big on services, and still isn't, and that number includes PCs.) If you assume there is a reasonable, if somewhat loose, coupling of total platform ecosystem sales to base hardware/software platform, then the ratio in the OS/400 ecosystem was roughly 2.6 to 1.

Now, I have a pretty good idea of what the worldwide market for iSeries hardware and upgrades is. I figure basic machines brought in about $1.5 billion in 2004. This is consistent with estimates made by Merrill Lynch's Steven Milunovich, who says IBM booked $1.478 billion in iSeries server sales in 2004, down 16.6 percent from the $1.772 billion in sales the company had in 2003. By comparison, 2003 was a good year, with sales up 3.9 percent from the $1.705 billion level in 2002. (Just for the sake of reference, in a normal year in the late 1990s, IBM pushed around $3.5 billion in AS/400 server sales, by my estimates.) Milunovich might not be aware of IBM's commitment to double-digit iSeries revenue growth this year, and that is probably why he is only predicting growth of 8.9 percent to $1.61 billion for the iSeries in 2005. IBM posted an uptick of only 1 percent revenue growth in the first quarter of 2005, and that was also with some pretty strong currency effects, which means sales in Europe are still declining in local currencies. In China, India, certain Asia/Pacific countries, and Eastern Europe, IBM has growth rates in the 20 and 30 percent range, but these are very small bases. In other words, iSeries server sales seem to be at a new steady state that is about half of where they were a decade ago.

If you assume this ecosystem ratio between OS/400 platform sales (servers plus systems software plus other hardware) and the overall OS/400 ecosystem is about the same today--and I realize this is a big assumption--then we can come up with a number. IBM booked around $1.5 billion in server sales and another $300 million to $350 million in OS/400 and DB2/400 sales, and probably a few hundred million dollars each in external storage, tape drives, upgrades, and other hardware in 2004. Call it $2.5 billion for all of these added together. Do the math--$2.5 billion times 2.6--and that is around $6.5 billion. I think the scaling factor (which has PCs and other elements in it) might change over time, especially because price/performance has improved so much over the past 12 years. Software prices have gone up, not down, and server prices have gone way down, not up or even held steady. Moreover, IBM has moved storage largely under the skins of the iSeries, so external disk sales today would probably be less than scaling factors based on past data might suggest. But, then again, the changes in portions of these scaling factors may cancel each other. It is hard to say, and real data is always better than making such estimates.

Estimate number two is also based on history. IBM used to have about 8,000 ISVs and 20,000 OS/400 applications and drove about $14 billion in ecosystem sales within the company. IBM now has 3,000 OS/400 ISVs and 8,000 OS/400 applications. If the ecosystem numbers scale with ISV count or application count, the IBM ecosystem for the OS/400 platform in 1993's categories would be around $5 billion inside IBM. That scaling factor tracks with what I independently guessed for OS/400 server, system software, maintenance, services, and other key OS/400 ecosystem sub-categories, but does not scale so well with PC sales (which have ballooned as the PC has grown in importance). Here's the 1993 data, the scaling down of those numbers as OS/400 ISV and OS/400 application counts declined between 1993 and 2004, and my guesses for 2004 before I even thought of this relationship:

Figure 2: AS/400 sales in 1993 and how sales might scale if there is a correlation between OS/400 ISV count (that's ISV scaling, which reduces it by a factor of 3,000/8000) or OS/400 application count (which reduces it by a factor of 8,000/20,000).

The reason I bring this up at all is that I think ISV count, application count, and ISV focus on the platform were key factors to the success of the AS/400 and are vital factors for the rejuvenation of the iSeries platform.

Now, the question the OS/400 software vendor asked me--remember, this is how this story started way up there in the first two paragraphs--was to estimate the size of the OS/400 software ecosystem. This is a bit trickier. I think IBM probably pushes another $300 million to $350 million in OS/400-related systems software, as I said above. As for how much money the OS/400 ecosystem spends on software in a year, I do not believe anyone has the foggiest idea. The vast majority of OS/400 software vendors are privately held and most are small companies with a handful to dozens of employees; there are a few hundred with hundreds of employees.

Here's a wild guess. There are about 3,000 core ISVs in the iSeries market. The top six iSeries ISVs together account for about $2 billion in sales (the former JDE, plus Intentia, the iSeries chunk of SAP, International Business Systems, MAPICS, and SSA) and I figure about half of their aggregate revenue is specific to the iSeries. So that is about $1 billion. Very quickly, we run out of big application software companies. So adding another 600 ISVs probably does not boost revenue by more than a few hundred million dollars. Call it $500 million. So that is 20 percent of the ISV installed base, and about $1.5 billion. If you use the 80-20 rule (I love that rule), then the remaining 2,400 or so ISVs only bring in another $400 million. That's about $1.9 billion in OS/400-related software sales outside of IBM, including applications, tools, and so on.

I built two different models in spreadsheets to characterize sales among those key 3,000 OS/400 ISVs. One model comes up with just over $1.8 billion in OS/400-related software sales among the 3,000 ISVs, while the other comes up with $6.5 billion. I created this table as a check of sorts, and I think that Model A is reasonable and Model B is probably wildly optimistic. I present them so you can see how I am trying to guesstimate this and provide some suggestions on how I might build a better model.

Click here to see the models for the estimated ISV OS/400 software ecosystem.

So, if you add up the IBM OS/400 ecosystem of about $4.5 billion using my guesses (as cross-checked with the ISV and app count scaling factors against the 1993 data) with about $2 billion in OS/400-related software sales (from the other model), you get an OS/400 ecosystem of about $6.5 billion. About the same as the $6.5 billion I arrived at from the other set of estimates. I am not saying this is correct, mind you. But that it is consistent.

It's a pity, of course, that having made estimates, there is no way to perform a calculation to verify this. Server platform numerology is not physics, after all. All I can do is give you the benefit of my guesses.

And, because I am obsessive about this whole notion of platform ecosystems, I went one step further and created a more detailed OS/400 ecosystem model--one that compares and contrasts with overall IT spending and that takes into account the costs of people as well as hardware, software, and services. I will show you this model next week, and after that, I will compare and contrast the OS/400 ecosystem with those of other platforms to try to make some sense of how they all fit together, and butt heads at the same time, in the overall IT ecosystem.

In the meantime, if you have any suggestions about how to build a better model, or some data that we can plug into the model, please go to the IT Jungle contact page and send me a message.

RELATED STORY

TFH Flashback: Critical Mass, November 1993 issue