Some Thoughts on i5 Spending Patterns
May 14, 2007 Timothy Prickett Morgan
Two weeks ago at the COMMON midrange user group meeting in Anaheim, California, Mark Shearer, the general manager of IBM‘s System i product line, divulged some internal data that the company has developed concerning spending patterns at i5 shops. I have been mulling this data, and I wanted to share some thoughts about it.
As I explained in last week’s issue, Shearer and his marketing team are taking what they believe is a practical approach to i5 sales: They are going to start focusing more on IBM’s share of total revenues at i5/OS and OS/400 accounts and less on quarter i5 sales.
The reason why IBM is doing this is twofold. First, no one is convinced that IBM can bolster i5 sales by anything near the growth that would be necessary to return the i5/OS and OS/400 platform to the heydays of the early 1990s, when IBM had close to $5 billion in sales of AS/400 systems, related storage, and systems software, or in 1996, which is when IBM made the transition to Power-based AS/400 servers and it came close to hitting $5 billion in system sales (again, that includes hardware and systems software, but it does not include maintenance). Back in 1996, the Power AS/400s offered customers a lot more performance at about the same price, yielding a factor of 2X or more better price/performance. This is the last time IBM was so generous, and that is also pretty much the last time the AS/400-iSeries-i5 business grew appreciably. I have always said that these factors are related, and I have said it to every general manager I have ever met who ran the IBM midrange biz. I believe in price elasticity for computing platforms, and I need only look at X86 and X64 server sprawl to know I am right. Back in 1996, IBM was number two in the X86 server space, behind Compaq, and that business was one-fifth to one-fourth the size of the AS/400 business. The RS/6000 server biz was half the size of the AS/400 business, which rivaled the mainframe business of 1996 in terms of revenues.
Ponder that for a second. Let it sink in.
Shearer said in an interview with me two weeks ago that based on surveys of the i5/OS and OS/400 customer base in the United States, IBM has calculated that for every $1 that customers spend on i5 systems, they spend an additional $2.60 on Windows systems and another 40 cents on Unix systems and external storage arrays. If you extrapolate that to a worldwide scope, and measure it against revenues, then you would calculate that against an annual spend of about $1.8 billion on i5 servers (I’ll be more generous than other numbers I have seen), the i5/OS and OS/400 customer base then, in aggregate, spends around $4.7 billion on X86 and X64 servers running Windows, and then another meager $720 million on Unix servers and external storage.
The funny thing I saw in that is just how allergic the i5/OS and OS/400 base has been to Unix–except, of course, at accounts where the i5/OS and OS/400 plus DB2/400 combo has come up against a Unix plus Oracle or DB2 combination and lost mightily. These guys are removed from the pool once the AS/400, iSeries, or i5 is unplugged. And best I can figure, anywhere from 50,000 to 75,000 shops in the past decade moved off the i5/OS and OS/400 platform, and probably half went to Unix with the remainder moving to Windows.
Now, work back to 1996. I saw surveys that showed IBM had dominant market share as an X86 server supplier into the AS/400 base. A survey I did in January 1996 against a massive customer database owned by Computer Intelligence (now part of Harte Hanks), showed that the IBM label was on nearly half of the 80386-based servers installed at OS/400 shops, but IBM’s share had dropped to 30 percent for 80486 boxes installed at the same time (compared to Compaq’s 42 percent). The Pentium was the hot processor back then–running at a blazing 100 MHz in top-end machines–and IBM had under 15 percent share of boxes at the same AS/400 sites. IBM was sliding pretty hard back then, and it had to have been terrified at the account control Compaq was getting at its AS/400 sites, which were installing NetWare file servers and just starting to play around with the Internet. The File Serving IOP, which has evolved into the Integrated xSeries Server, helped a bit.
According to Shearer last week, about 70 percent of the X86 and X64 servers at i5/OS and OS/400 sites do not have an IBM label on them–but the good news, at least from IBM’s point of view–is that 30 percent do. IBM has not seen X86/X64 server market share like this in a decade, in fact, and it is a testament to the designs and marketing prowess of IBM in the System x market that it has a share even that high.
The funny thing to me–for a loose use of the word funny–is that back in January 1996, I was railing against IBM for charging too much for AS/400s and for splitting the line in two–expensive machines that run RPG workloads, which were called Advanced Systems, and cheaper machines for running other kinds of code with no 5250 support, called Advanced Servers. I don’t often quote myself, and I don’t get a lot of satisfaction from saying “I told you so,” but I warned IBM what would happen 11 years ago:
The PC servers available from IBM, Compaq, and HP are particularly attractive to AS/400 customers who want or need to run a set of applications under NetWare, NT, OS/2 (LAN Server) or SCO Unix. For instance, while a manufacturing company may have an AS/400 keeping its books and running its shop floors, there is probably no good reason why that company’s personnel applications can’t be put on a PC server. While the AS/400 can do all of the same things that a PC server can do–it’s a database server for PC clients, it’s a file server, it’s a print server, it can get you out on to the Internet–AS/400 applications may not be as slick as those available for PC servers. This is especially true for Windows NT apps. Moreover, AS/400 servers may not be as cost effective as PC servers. I am aware that IBM has shown very good price/performance with the first generation of Advanced Servers on the Bapco client/server benchmark tests at Comdex last winter, but times have changed. IBM’s own PC servers are just as impressive as its AS/400 Advanced Servers, and they blow away the Advanced Systems. Both Compaq and HP offer PC servers that are even less costly than the IBM alternatives.
The AS/400 division’s response to the direct and indirect competition coming from PC servers involves a lot of public relations and forthcoming announcements.
IBM constantly reminds us that while AS/400 hardware and software is often more expensive than equivalent PC server technology, the AS/400 usually has an overall lower cost of ownership. Rarely have I seen this claim backed up with hard data. And the data I have seen did not and could not apply universally across all sizes and makes of midrange machines at all times. Yes, an E02 was a great machine compared to a 486 server in 1992. But can IBM hope to get into new accounts with a 510-2144 with a $195,000 price tag when an HP or Compaq P5-133 server with about as much power and memory costs $25,000? Even a base 50S-2121 server (roughly the same as the 510-2144) costs about twice as much as the PC server.
I believe that IBM has lost its edge in the less tangible aspects of cost of ownership, too. The services and hand-holding that used to come free from IBM in the past now come with a price tag. IBM’s AS/400 sales and service reps are spread just as thin on the ground as those from the IBM PC Co, Compaq and HP. Moreover, there is no indication that RPG programmers are any less costly than Cobol and C++ programmers; I doubt that systems analysts who are intimate with OS/400 are any more numerous or less costly than their counterparts in the Unix or NT markets. Programming tools for other midrange markets are just as productive and slick as those for OS/400. And finally, I doubt that end user training costs are much different on an AS/400 than on a PC server running OS/2 LAN Server, NetWare, Windows NT or SCO Unix. In the client/server world, everybody has the same client–Windows, and if Bill Gates gets his way, Windows 95–and all applications look about the same.
While I applaud IBM for its intent to launch a low-cost, low-end RISC AS/400 aimed directly at the Compaq-NT market, I wonder if IBM has seen that Compaq and other PC server vendors aren’t just aiming at little AS/400s, but across the whole AS/400 range. And perhaps more significantly, the PC server vendors are aiming at each other, which drives the price/performance curve down even faster. This leaves IBM with a simple choice. If IBM just wants to maintain its AS/400 franchise, keeping relatively high margins, it can continue to charge what it does for AS/400 products. Or the company can come to the conclusion (as I have) that the AS/400 is its best commercial server. IBM can realize that its consistent and persistent strategy to charge more than any other vendor in the midrange is what really keeps the AS/400 from selling 500,000 units a year instead of 60,000. If IBM was the price leader as well as a technology leader and the volume leader in the midrange, who could touch it?”
I was a lot younger back then, but I still didn’t have any hair on the top of my head. And the beard didn’t have little flecks of white like it does now. And I am smart enough to know that the System i line is still profitable for IBM, and probably very much so because of one word: maintenance.
With over 200,000 accounts and an installed base of hundreds of thousands of machines, IBM is still probably raking in as much per year on maintenance as it does on hardware and system software sales in the i5 line–just as it did back in the 1990s. The number was just twice as large a decade ago, and then external storage was way more popular, too. The point is: The cost of that iSeries and i5 maintenance and tech support is relatively low because the machinery and software is very reliable. This–and not hardware sales–is the gravy train. I would guess that for every single $1 that i5 shops spend on i5 systems, they spend around $1 for maintenance and other support. And if they spend $1 on systems every couple of years, they spend that $1 on maintenance every year. Which is why IBM stays in the game, and which is why IBM does not seem too concerned about boosting i5 sales.
Here’s the other thing I learned. Shearer told me at COMMON that about half of the System i5 revenues in a typical quarter come from large systems–by which he meant big i5 570s and 595s–while the remaining half of sales came from smaller systems. The bulk of shipments, of course, come from small machines. The AS/400 market was a little top-heavy, but big systems did not dominate the revenue stream and sales were more of a pyramid, with lots of small systems bringing in a relatively large sum of money. This more recent distribution of sales only shows how much Windows has displaced OS/400 as the operating system of choice for back-end systems at midrange businesses.
This is the real problem the i5 faced a decade ago, and it is the same one it faces today.