IBM pushes hard for mainframe sales - Mainframe Markets - International Business Machines Corp

Rethink IT, Jan, 2004

MARKET OVERVIEW

IBM ended the year in a frenzy of activity to boost its sales revenue within the mainframe division, The company was chasing new types of deals such as the ESSO, ELA or OIO, which force users into a lung term commitment to IBM and also more traditional deals based around 'an offer you can't refuse' or budget based offers that do not require a budget increase.

Some of the latter deals can be simply based upon a good hardware price together with lower software and maintenance costs into the future. These typically increase IBM's profits, particularly where any form of leasing is involved, and increase the current year's profits. But they will also reduce the mid term IBM revenues a little. Other budget

based deals, however, do nothing for either IBM's revenues or profits and can indeed dramatically reduce the profits--as will be seen later.

However, by far the biggest problem is that such deals contribute a wide range of pricing depending upon the exact circumstances rather than the fair market price. This range is currently a full 300%--that means one user pays $1m and another $3m for the same product. Other users pay prices that range between these extremes but, to be honest, at this time they are mainly closer to the top end of the range.

T-REX: A DESIGN PROBLEM?

A major problem that IBM faces with the z990 or T-Rex is that the cost structure is quite different from any previous systems and makes the pricing model very complex to implement without looking at the actual cost of a system--something they never had to worry about before.

For example, a user buying say a 2084-306 may well consider that this system will consist of a one book model A08. However, if he wants to buy two CBUs (capacity back up engines) and an IFL, there is a need for a two book 2084-B16 at almost twice the cost of an A08. This user has a price in mind for the 306 of perhaps 2.4m [pounds sterling] (400k [pounds sterling] per processor) plus say 210,000 [pounds sterling] for the two CBUs and an IFL (average 70K [pounds sterling] per processor). Indeed, IBM may have given him the impression this was the pricing. Bur IBM finds that to add the two CBUs and an IFL DOUBLES (well, to be precise nearly doubles) the cost of the system. Indeed, the cost increase far exceeds the price that it can sell the non $390 processors for. IBM has now had to ship twice the processors, twice the memory and probably more channels compared to the A08.

Consequently the price to the user will increase dramatically.

This design problem should not be underestimated. I discovered recently in one case where a relatively modest size user with two systems for availability reasons found that to provide sufficient CBUs to back up each system together with the required ICFs and IFLs required a D32 four book system with 32 processors! Quite a surprise as initially he had viewed the top end T-Rex as well beyond his likely requirements.

Readers should be aware of this issue and be wary of the promise of cheap ICF/IFL or CBU processors--they are only cheap if there are 'spare' processors in a book. Even then, when the next $390 processor upgrade (or indeed ICF/IFL/CBU) takes you to the next book the price will be high.

These cost issues should not affect the user, nor should users need to consider whether they are getting a one, two, three or four book system. The price should be decided by the $390 processor capacity and the additional 'special' engines selected. An example could be set up as below in Table 1:

However, this type of model causes major problems for IBM as can be seen in the example below where I calculate the cost and revenue for each of the four 'books' within the z990 range using the above table and on the assumption that all of the engines are sold as one type. See Table 2.

This translates into an acceptable profit margin for IBM (they would like 90%) where all engines are sold for $390 work but unacceptable profit levels for the other types of engine with the CBU actually producing a loss--and remember that only the $390 engines generate software revenue which will probably more than treble the real revenues created for IBM by each $390 engine sale. See Table 3.

With even relatively modest users with a rich mix of S390/IFL/ICF/CBU engines IBM will need to ship a D32 system and the overall profit level will reduce considerably unless they increase the price which they will--making all the low cost IFL/ICF/CBU claims a lot of hot air.

If they don't raise the price, based upon the above figures, even a 2084-310 user who wants to back up another similar size system and use a couple of IFLs and three ICFs will need a D32 to be shipped as he uses not only more than the 16 processors of a D16 but actually 25 processors which exceeds the C24 model from IBM. At 'normal prices this would produce no more than 70% profit margin on what IBM expected to be its most profitable model. See Table 4.

The problem of design versus cost is not a new one, for some time within the 9672 range IBM has built only 'fully populated' systems with all processors included. This did cause some problems at the low end and the 'book' approach was supposed to fix the problem but has failed dismally. In fact it has accentuated the problem as the z990 processors are so much bigger (in performance terms) and therefore when IBM sells an IFL/ICF/CBU at a similar price on the z990 to the z900, it effectively gives away an enormous amount of $390 capacity it could have sold for a far higher price.