Although the company still does not seem to realise it, IBM Corp has lost an empire and has yet to find a role but the rest of the world can see that in the new computer industry order, Intel Corp and Microsoft Corp share the role that IBM formerly held in the mainframe – and by extension, the whole computer universe. Intel’s current position mirrors remarkably closely the state of affairs when IBM and the 370 mainframe held sway. Intel sets the architectural standard to which the vast majority of users adhere, and just as IBM had to contend with Amdahl Corp, Fujitsu Ltd, Hitachi Ltd, Magnuson Computer Corp, Two Pi Systems Inc, Cambex Corp, IPL Systems Inc and sundry other now-forgotten would-be IBMulators, so Intel has to contend with Advanced Micro Devices Inc and Texas Instruments Inc, plus Cyrix Corp and sundry others that want to make a living by emulating the iAPX-86 architecture and picking up some of the crumbs left by Intel.

Would-be usurpers

Intel responds just as IBM did before it by brandishing its lawyers at the would-be usurpers while investing and working furiously to push the architecture forward so fast that the Intelemulators haven’t a hope of keeping up. And just as in the 1970s and the 1980s, there was a battered band of mainframers Sperry Corp, Burroughs Corp, Honeywell Information Systems, Control Data Corp, NCR Corp, gamely trying to persuade their dwindling customer bases to keep the faith and not convert to the overwhelmingly dominant IBM standard. Now and again, one of the mainframers came up with an innovation that cut its costs so dramatically that it was able for a time to reverse the tide: Sperry did it most notably with the 1100/60, a mainframe built of Motorola Inc 10800 ECL bit-slice microprocessors that by using heavy pipelining and the well-established trick of dividing the instruction stream in two and sending half down each of two arithmetic-logic complexes, created a machine so dramatically cheaper to build that it was able for a time to win new adherents to a non-IBM mainframe standard. But by the time that Burroughs swallowed Sperry to form Unisys Corp in the mid-1980s, it was clear that the game was up, the non-IBM mainframe was doomed, and it was game, set and match to IBM. What a pyrrhic victory! IBM had devoted so much effort to killing off the non-IBM mainframe and scoring silly debating points off the IBMulators that it had quite failed to notice that two of its own creations, Microsoft Corp and Intel Corp, aided and abetted by the likes of Novell Inc and Oracle Corp, had quietly been rewriting all the rules of the game, and the one that IBM had won was one that interested nobody any more. Now Intel finds itself in the same position as IBM in about 1980, with the one striking difference that it does not own the software that runs on its totally dominant chips. But Microsoft’s interests are so bound up with those of Intel that although it pays lip-service to the idea of putting, or allowing its software to be put, onto other chips, few in the Intel world are shivering at the idea of Windows NT running on Digital Equipment Corp’s Alpha chip. All the RISC chips are cast in the role of the non-IBM-compatible mainframes: will any be any more successful in turning back the tide?

Fatally compromised

This week, the champions of an alternative chip standard are furiously (to use an Australianism) barracking for the PowerPC, and there are a few analysts and investors seriously prepared to countenance the idea that this poses a threat to Intel. Sad to say, for all those that favour diversity, the direction in which Louis Gerstner has set IBM means that even IBM will be ambivalent about giving the PowerPC its head, and without IBM’s whole-hearted backing, the PowerPC is doomed to be just another non-Intel chip. The IBM problem is that if PowerPC-based personal computers were to take the market by storm, they would pose a threat to what over the past couple of years has become an enormous iAPX-86-based personal computer business – with many of those Intel-compatible chips actually m

ade by IBM. IBM, having given up rights to fabricate the Pentium, seems to be committed to having the PowerPC take over at the high end of the personal computer market, but as soon as it is clear that PowerPC personal computers threaten IBM’s own PS/2 business and yet command much lower margins, the old guard at IBM will ensure that its budding success is fatally compromised – just as the mainframe and AS/400 camps have made sure that the RS/6000 has failed to meet the predictions of its backers within IBM that the company would be number one in Unix this year – in fact it is number three and stalled. The belief that one chip could derail the Intel bandwaggon is anyway fatally flawed: as that wise old bird Dr Gene Amdahl advised us when we suggested that one mainframe standard might survive alongside IBM, there was even less room for just one alternative standard than there was for six. [There is a widespread assumption that the Precision Architecture, Sparc, R-series, Alpha, PowerPC each has to become a major player on the merchant market to survive, whereas so long as the bases of the company or companies that own and depend on any given RISC architecture continue to grow, enough cash will be generated to keep the chip current even without a large number of external design wins: the 88000 failed specifically because Motorola was not a significant player in the systems business.] The conclusion is that just as the first computer empire was ruled by IBM and MVS until the barbarians in the shape of the Z80, CP/M, the Apple II and VisiCalc arrived at the gates to clear the way for Intel and Microsoft to sack the citadel and create the second empire, the Intel-Microsoft hegemony will hold sway whatever IBM may try to do with PowerPC until something else equally revolutionary arrives to sweep the new established order away – and by the nature of things, there is no way of forecasting just what that will be.

Optical microprocessors

To try to do so is a mug’s game, but it could be that Intel will finally be tripped up by the ever-increasing complexity of its architecture, and that the next revolution will be driven by masses of dramatically simpler devices for which Microsoft’s software will no longer be appropriate. What breakthrough would be most likely to bring this about? It is axiomatic that the first optical microprocessors will have be relatively simple devices that can be used to build powerful machines only by running thousands of them in parallel. Why should Intel not be the first to anticipate and master this revolution? Things just don’t work that way: Intel will have so much physical and emotional capital tied up in iAPX-86 that instead of taking the lead in smashing the past to create the future, it will confidently rubbish the leaders of the coming revolution. But by then it will be a very big company indeed – as IBM was only the other day…