IBM may seem to have got more things wrong than it has got right over the past five years, but no-one should be under any illusion that the task of running the world’s biggest computer company has got anything other than progressively more difficult, and the problems that the company is likely to face over the next five years would tax Superman’s abilities. Depending on how the rest of the business is going, IBM derives anything from 60% to 80% of its profits from its mainframe business, which accounts for about half the total by volume. But the mainframe business has already turned into a mature, low-growth one, and within five years, it threatens to become a no-growth one for IBM. Indeed Amdahl Corp, diverting increasing proportions of its research and development into its native implementation of Unix for its 370 architecture machines, may well be better placed for continued prosperity in mainframes than IBM itself. The immediate cause of IBM’s problem is the devastating failure of the 9370, which looks more and more irretrievable. By contrast, the AS/400 is a clear, if over hyped, success. Grey market We are beginning to hear word of bitter disagreements between IBM and some of its AS/400 resellers in the US, with all the concomitants of the Personal Computer reseller problems – dealers overstuffed with inventory, grey markets and IBM getting heavier and heavier with the ones that have failed to meet their targets or have been seen dumping surplus machines onto the market at any price they can get. But, as we have been saying since last October, all the signs were that most of the easy AS/400 sales, those to System 38 users, would have been made by last month, so there is no reason for anyone apart from gung-ho Wall Street analysts who believe their own forecasts to be surprised that the breakneck pace of market acceptance of the AS/400 should be beginning to slow. And none of that alters the fact that the AS/400 is a stunning success when compared with the 9370 – and indeed with the 4381. And that success – plus the fact that the machine is relatively easy to use, and – for IBM as well as users – to program for, means that increasing amounts of new software and functionality is appearing first on the AS/400 it was the first to feature a preview version of IBM’s forthcoming Office product, it features more prominently than 370 machines in IBM’s computer-integrated telephony announcement last week, and there is no reason to believe that that is not a pattern that will continue and accelerate in the future. After all, if IBM is going to sell a lot of the boxes, it makes sense to try and bleed the people who’ve got them of as much cash for add-on products as possible. Moreover the machine has gone down extremely indeed with third party software houses, who always liked the System 38 and were only frustrated that IBM put so little marketing muscle behind it. So why does the the AS/400 pose a problem for IBM? ICT and English Electric After ICT was merged with English Electric Computers to form ICL, the next seven years were spent in unproductive and debilitating rivalry between the ICT 1900 and the English Electric System 4 sides of the house, and even the launch of the 2900 in 1974 didn’t eliminate the rivalry. IBM has the good fortune to have reached its formidable size without the trauma of having to go through any major mergers since the launch of System 360 in 1984. But by the mid-1970s, it became so clear that upstart companies like DEC and Data General were beginning to eat IBM’s lunch in the low-end mass market that IBM made what then looked like a sensible and imaginative move – it created a General Systems Division counterpart to the bureaucratic and monolithic mainframe business, which became the Data Processing Division. The General Systems Division laid a couple of ripe eggs with its first two products, the System 32 single-user RPG computer and the 5100 desk-top machine which was actually IBM’s entry into the personal computer market, although no-one recognised it as such at the time. The System 32 sold so w
ell that IBM had made its money back several times over on it before people realised that they had been had, but the new GSD really got it right with the System 34, which was the first low-end machine from the company that was genuinely competitive with the host of multi-terminal minicomputer-based business systems and didn’t carry an intolerable premium for the little IBM badge on the front. But at the same time, GSD, which had been given a much freer and easier brief than DPD to be innovative and to try things out, had laid another egg with the Series 1 minicomputer, but it is the IBM way to plod on with its failures until it has recouped its investment in them, and having failed to dent DEC or Data General with the Series 1, the GSD boys looked around for another market and spotted that while the Series 1, if it had a strength, found it in its communications, an area where the batch-bound 370s were weak. So they went out and started selling the Series 1 as a free-wheeling alternative to DPD’s own late 1970s dog, the clumsy and inflexible 8100. Soon users were to be treated to the hilarious sight of GSD and DPD salesmen arguing like Kilkenny cats over their business, and found to their amazement that if they said a plague a both your houses and ordered a string of Syfas from Computer Automation, the DPD salesmen seemed to go away remarkably content. Then the General Systems Division really over-reached itself and introduced arguably the most innovative and best mid-range machine from IBM or any other manufacturer. Effectively a powerful microcoded relational database with a lot of input-output (and too little memory), the System 38 was pretty clearly the machine into which System 370 should have evolved by then – ideally it should have been built as a mighty co-processor to the 370 CPU so that users over time could migrate their on-line database applications across. Ossified Unfortunately, the Data Processing Division had at the same time been working hard with the latest chip technology from the IBM labs to come up with a dramatically cheaper successor to the 370/115, 125, 138 and 148, and in 1979 it came out as the 4331 and 4341 and early 1960s architecture swept all before it, including the late 1970s architecture System 38. And two or three years later, IBM decided that DPD and GSD were both becoming ossified and decided to reorganise and merge the two divisions. But, just as with the ICT-EE merger, old affiliations died hard, and only last month, we heard of a US user really keen to buy an AS/400, whose salesman argued adamantly that what he needed was a 9370. But the more successful the AS/400 becomes, the bigger the machine will have to grow until it starts stealing development dollars from MVS and VM. Because the failure of the 9370 and the rapid fading of the 4381 means that there are fewer and fewer new users being brought up in the 370 world. In five years, almost the only market for Summit will be 3090 users, slow growth will have become no growth, and with it, IBM’s main cash cow. But Unix users will begin to need mainframe power, and Amdahl Corp, but not IBM, will be waiting for them…
