Science and Technology
Delays in the microprocessor revolution
January 12, 1980
Some horrendous problems in developing applications for the latest generation of microprocessors have been identified by Intel, the outstanding microchip innovator of the 1970s. Intel thinks they can be solved. But they are bound to slow down the so-called microprocessor revolution.
The problems arise with the very powerful new 16-bit microprocessors, designed to do anything from operating a robot to being the brains of a computer terminal. Intel's version is called the 8086. The basic processor costs around $35 but a typical customer would pay $200 or more for extra chips to go with it. The customer--designing all this into, say, his new robot--is faced with the task of choosing which chips to have, how to wire them together, how to write the software for them and so on.
Mr William Davidow, head of microprocessor systems at Intel, reckons that in the early 1970s a customer needed to spend only $0.3m or less to develop an application off his own bat. Today, however, with the new 16-bit microprocessors, it would cost him around $5m. The solution is for the customer not to do all the applications development on his own but to get the microprocessor manufacturer to help him.
This should bring the cost down to about $1m. But that is still far more than in the early 1970s. Many users of chips, especially small firms, will be restricted to applications involving simpler microprocessors and will have to eschew some of the latest technology.
Then there is another major headache. Partly in order to help the customer reduce his applications development cost from $5m to $1m, Intel reckons it will have to spend around $200m ''closing the application technology gap'' for its 8086: in other words, in order to develop a whole family of electronic devices to work in tandem with the basic processor, to design software, and so on. That $200m is just for R and D. It excludes the investment in manufacturing facilities that will be needed.
Concealed behind this cost estimate is a highly significant assumption. To justify spending $200m on closing the applications gap, Intel calculates that the 8086 and its related products will have to achieve sales of at least $2 billion during the microprocessor's lifetime in the marketplace. That may sound over-optimistic for a market in which Motorola, National Semiconductor, Texas Instruments and Zilog are among the competitors. How many companies selling $2 billion-worth of 16-bit microprocessors can the market stand?
Morgan Stanley's electronics experts reckon that Intel's 8086 family will be selling at an annual rate of $250m by the end of 1980. Intel is reckoned by some observers to dominate (by value) the market for the new, advanced 16-bit processors at present. Is it planning to be the monopoly (or near-monopoly) supplier of 16-bit processors? The chances are that, as competitors get their advanced 16-bit families ready, they will steal some of Intel's thunder. And the demand for 16-bit processors looks like increasing as customers learn to use them.
On the other hand, microprocessors do seem to be heading for a situation like that in mainframe computers, with few competitors able to afford to stay in the game--and certainly not much room for the Europeans.
One of the key restraints in developing all the applications to soak up the huge potential supply of 16-bit micros is people. Application systems are not the sort of thing that can be developed by somebody with a few months' retraining. They need people with years of experience on micros. There are not many of those even now. And some companies fear that development contracts announced recently by the United States department of defence (see box) will make the manpower shortage worse.
Copyright 1980 The Economist Newspaper Ltd.