The Four Hundred

I've spent the last four weeks designing my iDeal iSeries servers, a sort of wish list of hardware, software, and pricing practices that I think points IBM in the right direction for making the iSeries a competitive platform, one that can actually increase market share each year in an increasingly Windows and Linux world. The only thing that remains to be done is to see how such a strategy might affect IBM's financial results. So let's get to it.

I'm not going to bore you with a full presentation of the financial models I have built for the iDeal iSeries machines, but I do want to share the results of my what-if analysis. To simplify matters, let's assume that IBM replaced the iSeries line with the iDeal iSeries line, and started selling these machines on January 1. IBM sold only about 30,000 iSeries machines in 2001, and it is hard to imagine that unit shipments will be higher in 2002. With 250,000 customers and an installed base of 490,000 machines, IBM should be selling more machines, and it is my contention that pricing is an issue and that there is a tremendous amount of pent-up demand for machines that are attractively priced. I think even a pessimist would agree that IBM could churn half of those machines in the installed AS/400 base in three or four years with machines like my iDeal iSeries line. Call it four years, to be skeptical. That's a run rate of 61,250 machines per year. IBM hasn't sold that many machines since the booming late 1990s. But when you consider that it has hit that level in the past, it suggests that when market forces meet go-to-market pricing, the iSeries can in fact sell. This is not a staggering number of machines. In the past four consecutive quarters, IBM has sold 75,000 pSeries servers, according to Gartner's Dataquest unit, and to a much smaller installed base of customers. Hewlett-Packard does about that volume in the Unix market, and Sun Microsystems does anywhere from three to four times that volume in a year.

After reckoning what unit shipments might be, the next problem to tackle is model mix. I think that the vast majority of customers will want the iDeal iSeries machines that support native OS/400 and Linux, and that are based on the PowerPC 970 processor. These offer the best bang for the buck and have that shadow excess capacity that I want to dedicate to transparent, embedded high availability for entry iSeries machines. (You'll have to read the past iDeal iSeries stories--the links for which are at the bottom of this story--to find out what I am talking about here.) All server product mixes are, generally speaking, pyramids. There are many more small customers than large ones, and they buy more, yet less powerful, machines. I have created two pyramids to characterize iDeal iSeries shipments. I also know that in each server subcategory, customers tend to buy more of the entry machines than the big machines--they want expansion room, and if they are worried about that, they will move up the product line to the next model and buy a lean configuration rather than a heavy configuration of a machine toward the bottom of the line.

For the flatter pyramid, more customers buy a lot more entry iDeal iSeries machines than bigger machines. I figured that 60 percent of iDeal iSeries machines--36,750 units--would be for Model 381 uniprocessors. (I actually have breakdowns in the model for each specific processor option.) Another 22.5 percent of shipments in a year (13,781 machines) would go for the Model 382 dual-processor machine, and the four-way Model 384 would comprise about 4.5 percent of shipments (2,756 units). The Model 831 (which is a rejiggered pSeries 630) and the Model 851 (a rebranded pSeries 650) would each account for 3,675 units (6 percent of shipments). The Model 871 (pSeries 670) and Model 891 (pSeries 690) would account for 613 machines, or just a hair more than 1 percent of shipments.

Given the prices I set for hardware and software, I reckoned further that 65 percent of customers buying their initial machines would be able to get by with my base configurations. They are configured appropriately--excepting storage expansion and other peripherals, like tape and CD storage--so there isn't much more to buy. I figured the other 35 percent would have to buy heavier configurations, like the ones I outlined in the previous articles in this series. I also figured that about half of the users in the iDeal iSeries installed base would be using the 5250 protocol to access databases and the other half would use ODBC, JDBC, or other data access methods (like SQL). In my model, the 5250 access is priced at $225 per seat, while ODBC/JDBC/SQL access is priced at $50 per seat. When you let the spreadsheet do the math on all of that, it works out to $1.82 billion in PowerPC 970-based iDeal iSeries machines (again, not counting storage, peripherals, other software like compilers, WebSphere, and so forth) and $1.42 billion in Power4-based servers (the rejiggered pSeries line). That's $3.24 billion in annual sales. I will remind everyone that IBM is counting interactive software as hardware right now and shipping 30,000 machines, and is lucky to break $2 billion in sales on hardware and base operating systems in the iSeries line. Some people think 2002 will look more like $1.6 billion to $1.7 billion. On the entry and midrange iDeal iSeries machines (Models 381, 382, 384, 831, and 851), hardware is about 40 to 50 percent of total server revenue, and on the bigger Model 871 and 891 machines, hardware represents 70 to 80 percent of the total cost of the initial configurations sold. To calculate the potential profits for sales of the iDeal iSeries, I assumed that the entry machines (Models 381, 382, and 384) couldn't have more than a 10 percent gross margin; gross margins on the Model 831 were set at 15 percent, on the Model 851 at 20 percent, and on Models 871 and 891 at 25 percent. The math works out to $1.55 billion in hardware sales at a gross margin of 14.9 percent, or $231 million. This isn't terrible in the server business. It is quite good.

On the software side, those initial iDeal iSeries configurations sales for 61,250 shipments a year, using my flatter model mix pyramid, results in $1.69 billion in OS/400, DB2/400, and data access client middleware sales. This software would easily have an 85 percent gross margin--probably higher, since it is already written--and that brings $1.44 billion to the bottom line. Add it up and the iDeal iSeries platform has a base gross margin of 51 percent.

I think this is the most likely scenario. But just for giggles, I created a different model mix pyramid that is taller, by which I mean more of the rejiggered pSeries machines in the iDeal iSeries line sell than in the other model. In this model, only 30,319 of the Model 381, Model 382, and Model 384 PowerPC 970-based machines sell, and IBM is able to sell thousands of gargantuan Model 871 and Model 891 machines because server consolidation really takes off and the number of partitions on the machines are radically increased. This is the kind of marketing plan that IBM's engineering focus seems to be aimed at, and the reason why this is the case will be obvious after I tell you this: When I plug this taller pyramid into the iDeal iSeries sales model, IBM rakes in $11 billion in server revenues on 61,250 shipments in a year, with hardware gross profits of 22 percent and overall gross margins of 45 percent. (The hardware sells for a lot more money, you remember, but it has lower gross margins.)

The interesting bit in these scenarios is what happens if you drop annual shipments to 30,000. With the flat pyramid iDeal iSeries model, you get $1.59 billion in overall iDeal iSeries sales, with just under $760 million in hardware sales and about $703 million in core software sales. (Sound familiar?) With the tall pyramid iDeal iSeries shipment model, you get $5.4 billion in iDeal iSeries sales--$3.4 billion in hardware sales and just under $2 billion in core operating system and database software sales--and the same 22 percent hardware and 45 percent overall gross margins on hardware and software.

It occurs to me that IBM is spending its engineering and marketing effort as if it were shooting for the tall pyramid model, and what it should be doing is using the flatter pyramid, where inexpensive yet powerful hardware drives more software seats per processor and therefore improves gross margins and boosts overall shipments. Call me crazy, but this can work.

I'll make you a deal, IBM--even if it isn't ideal. If you revamp the iSeries product line--in terms of hardware, software, and pricing--in any way that even moderately resembles the iDeal iSeries product line I have outlined over the past five weeks, I will give you all the rights to use that name. I'll give you all of my spreadsheets. I'll even give you free advertising in Guild Companies' newsletters to help sell them. Not that I think you'd need much advertising. You'd be struggling to meet demand, I think, which would be a nice change. Take a chance. Do things differently. Don't put governors on the database and go down the same wrong road. Use the innovation that we all know the Rochester Labs can cram into a great iSeries machine, and give the channel something that they can sell, and sell easily. Do this, and you won't have to listen to cranky crowds at COMMON again. Build the iDeal iSeries and they will come.