Inside Intel. Tim Jackson
Читать онлайн книгу.
of the 1103. Jordan and Regitz had hit a rough patch at work: their boss had suddenly told them that his travel budget would not allow them to go to the International Solid State Circuits Conference in Philadelphia which every semiconductor engineer worth his salt had to attend. So the two men had simply taken a week’s vacation from work, and bought themselves tickets to the conference from their own pockets. After the sessions they had drowned their sorrows in a few beers, and had come to the conclusion that they would both be better off working for Intel than Honeywell.
Their telegram to Noyce said nothing about Honeywell’s internal politics; it asked simply whether he would be interested in seeing a business plan to set up a division of Intel to build ‘memory systems’ – integrated units that companies could plug into their mainframe computers to increase their memory capacity. The answer, of course, was yes. Although they did not know it, the two men’s telegram landed on Bob Noyce’s desk just when he had come to the conclusion that Intel needed to do more than offer only loose components. At his suggestion, Jordan and Regitz mailed over a business plan outlining in five handwritten pages how the new division might make money. Noyce and Moore then flew over to the Boston to see the two men. Within a month they had moved their families across to the West Coast, ready to set Intel up in the memory systems business.
On their first day at work Regitz and Jordan were told that Intel’s plans had changed. The yield problems with the 1103 were so severe that the company could not afford to spare both men for the memory systems effort. One of them, Regitz, would be assigned to ‘turd-polishing’ – the project to redesign the 1103. Jordan would have to work on his own to set up the new memory systems operation. (Almost from day one the new division was referred to as ‘MSO’. Intel was big on abbreviations.)
Poaching one engineer from a disgruntled Ted Hoff and more from Honeywell, Jordan soon put together the beginning of a memory systems operation and the beginning of a product range – essentially large boxes full of circuit boards to which thousands or even tens of thousands of 1103s were mounted, together with the appropriate driver circuits. Selling 1103s in this way brought Intel two great advantages. One was simply the strength of numbers: with a single sales call, a single invoice and a single goods shipment, Intel could bring in tens of thousands of dollars. This was a source of great excitement for a company which until now had never had a product that had been sold for more than a couple of hundred dollars. It meant that the work of Intel’s sales operation could be better leveraged.
The other advantage of building memory systems was that Intel could now make use of parts that did not meet the specifications laid down in the company’s data sheets. When it was selling only individual components, there was nothing to do with parts that ran a little too slowly or consumed too much power other than to throw them in the junk bin. Yet these parts could account for a quarter, a half, or even more of the output of the fab line. Now that Intel could build these substandard parts into memory systems, the only thing that mattered was to make sure that the system as a whole performed adequately. Half the parts inside might not meet the specification of the components sold individually – but customers would not even know, let alone care. On the contrary; for the extra convenience of having all those substandard 1103s built into a complete memory system, they were ready to pay a premium.
An added attraction to the memory systems business was the marketing strategy of IBM, the grand-daddy of the mainframe business. To make its computers look more price-competitive against those of the ‘seven dwarfs’, IBM accepted modest margins on the basic computer, but then made its real money on the add-ons that customers came back for in a year or two when they needed to upgrade their systems. This was a high-tech example of making money from the blades instead of the razors. But it also opened up an opportunity for Intel. Since IBM built all its own memory devices, and wouldn’t even consider buying 1103s from Intel, why not design some memory systems especially for IBM machines, and sell them straight to IBM’s customers? The margins would be astounding, and the division would gain an entry into a previously closed market that accounted for nearly 70% of the world’s memory sales.
Venturing on to IBM’s turf required a little more expertise and care than the average new product introduction. Not only did Jordan have to make sure that the quality was acceptable – he didn’t want any of his division’s customers to get fired, if something went wrong, for failing to buy IBM. Jordan also had to try to copy some of the excellent customer service that was part of the IBM package. For example, the division had to devise a leasing programme that would allow companies to rent the memory systems for a fixed period of months and then send them back to Intel. These things were not difficult in themselves, but they, like Jordan himself, were all new to a relatively small company like Intel.
It was only when the memory systems operation was up and running smoothly that stresses began to emerge. The issue was how the new division’s accounting should be done. Should the chips that it obtained from the components business and then built into systems be charged at their full arm’s-length price, just as if the memory operation had bought parts in the market? Or should they be treated as costing nothing, since the parts that were now going into Jordan’s memory systems had simply been thrown away before his arrival?
Both approaches had problems. Booking the incoming units at the full price hardly reflected their opportunity cost to Intel, since the company had never tried to find any takers in the market for substandard components that failed to meet its specifications. Equally, it seemed too generous simply to give the chips to the new division; this might give rise to such accounting distortions that the company would be tempted to pour money into the memory systems operation because the figures made it look considerably more profitable than it really was.
Clearly the sensible answer was to choose some number between the two. But since each division manager was supposed to be responsible for the profit and loss of his own operation, the question was more than merely academic. Every extra penny charged that Bill Jordan was forced to pay for his chips would increase the profits of the components business and cut his own division’s profits; every penny that he managed to negotiate away in price breaks would make the components business look worse but him look better. The outcome could make or break careers.
The question would remain unresolved for nearly seven years, and both Gordon Moore and Bob Noyce would resist pressure to arbitrate between the two sides. But as the years ran on, a couple of trends would become evident. First, memory systems might be a fine business, but even on a favourable set of numbers it could not claim the 55% gross margins that were imposed elsewhere in the company as the test of viability of a new project. Second, the growth rate of the division was not as great as the growth rate of Intel’s other semiconductor-based businesses. Mainframes were a giant market in the early 1970s, but the arrival of low-cost memory and low-cost processing would change the face of the entire computer industry – and offer better opportunities for profit elsewhere.
Sometimes, understanding why a device failed was as important as finding a way to make it work. One of Intel’s most clever engineers was a tall, thin physicist called Dov Frohman who had devoted most of his working life, including a PhD and a spell inside Fairchild’s research lab, to the study of an obscure metal-nitride oxide semiconductor technology. When Intel decided to bet on silicon gate, Frohman was reassigned to trouble-shoot the process. Looking at the broken devices, he began to wonder whether part of the reason the silicon-gate devices had failed might be that some of the gates in the circuit had become disconnected, or ‘floating’.
Some weeks later Frohman wheeled a demonstration trolley into Gordon Moore’s office. In the course of ten short minutes he demonstrated a way of turning the phenomenon that caused Intel chips to fail into a something that could form the basis for a new product. ‘We put together a 16 bit array with primitive transistor packages sticking out of the sixteen sockets,’ Frohman recalled. ‘There were red bulbs to indicate the bits. This was all new to us, and we were thrashing around. We showed Gordon that by pushing the button you could program the device, and we demonstrated that it would hold a charge.’ Frohman had discovered something truly astounding: a semiconductor that behaved like read-only memory (ROM), but could be programmed with incredible ease.
This discovery had the potential to change the lives of engineers who needed to store information permanently on a chip. The standard way to do this