Читать книгу Inside Intel - Tim Jackson - Страница 16

9

Public Company

FIFTEEN YEARS AFTER INTEL’S FOUNDATION Andy Grove published a book called High Output Management, in which he set out many of the lessons he had learned from his experiences with the company.

The book began with a chapter called ‘The Basics of Production: Delivering a Breakfast’. Grove set his readers a simple problem that he’d come across while working as a busboy shortly after his arrival in America. ‘Your task’, he wrote, ‘is to serve a breakfast consisting of a three-minute soft-boiled egg, buttered toast, and coffee. Your job is to prepare and deliver the three items simultaneously, each of them fresh and hot.’ Within a few pages Grove plunged into the complexities of production management, using the breakfast ‘factory’ as an example. Continuous egg-boiling machines, toast-making delays, problems with rotten ingredients and rude staff – these and many other issues were dealt with as the proprietor of ‘Andy’s Better Breakfasts’ learned how to deliver an acceptable hot meal at an attractive price.

‘Bear in mind’, the Intel veteran wrote, ‘that in this and in other such situations there is a right answer, the one that can give you the best delivery time and product quality at the lowest possible cost. To find that right answer, you must develop a clear understanding of the trade-offs between the various factors – manpower, capacity and inventory – and you must reduce the understanding to a quantifiable set of relationships.’

Not too many people in the restaurant trade could tell you with confidence the link between the number of waiters they employ, the number of pieces of bread their toasters can handle every minute, and the number of eggs in the cold store at the end of the day. But in the infinitely more complex business of chip manufacturing, where there are dozens of different steps to making a product, scores of people with different skill levels involved, and new production processes and machinery being introduced all the time, the job of ‘reducing the understanding to a quantifiable set of relationships’ is almost nightmarishly difficult.

It was Grove’s determination to succeed at this piece of analysis that made Intel’s first years of production so stressful. Theoretician that he was, Grove had no truck with the touchy-feely approaches to manufacturing that he believed were common elsewhere in the industry. Instead, he wanted to be able to express Intel’s production lines as a set of equations like those he’d published in his book Physics and Technology of Semiconductor Devices. To do this, he needed facts: statistics in huge quantities, regularly delivered.

But this was 1971, not 1991. There was no company Intranet, no spreadsheet software – not even any desktop computers. The statistics that Grove demanded had to be collected and tabulated largely by hand – by people who were having problems enough just getting through the day, producing any output at all from the rudimentary designs and processes in Intel’s fabs. Most Intel engineers were lucky if they got home before midnight in time to see their families. No wonder there were stresses.

The 1101, Intel’s first silicon-gate MOS product, had been hard enough. ‘We embarked on this new technology … because of its perceived superior characteristics, although nobody was using this type of technology at the time,’ said Les Vadasz afterwards. ‘And the damn thing didn’t work! Week after week, we just pushed wafers through the line with zero yield! We were seriously beginning to doubt the correctness of our technology direction. But perseverance did pay off …’

The problems Intel had with the 1103 were of a different order of magnitude. Grove, who admitted to suffering from weeks of sleepless nights while trying to get the part into mass production, looked back on it as ‘almost as much fun as your final exams at college’.

‘I can remember twice a day going out on the line and physically counting 1103s as the introduction date drew closer,’ recalled another engineer. ‘We almost knew each good unit by name.’

Others came almost to hate the endless measurements and endless tiny changes they had to make both to the design of the chip itself and to the process used to build it, in an attempt to find a ‘sweet spot’ that would deliver reasonable yields. So heartily did they come to hate the troublesome 1103 that they referred to the job of getting it into mass production as ‘turd polishing’.

The hourly paid people who worked on the production lines, mostly young women, could be offered the incentive of cash – delivered in crisp dollar bills at the end of every week, because Grove believed in the animal trainer’s principle of making the reward immediately and visibly linked to the good performance that had earned it. They could be told to watch Betty or Jane, who seemed to be producing a higher percentage of usable parts, and imitate exactly what she did. They could be forbidden to wear make-up in the fab area, and they could be forced to tie up their hair behind a cap, and wear gloves, booties, a ‘bunny suit’ and safety glasses. They could be ordered to pick up wafers with a vacuum wand instead of with tweezers, to make sure that no physical contact took place which could break off microscopic chunks of material that would contaminate the processes further down the line.

But professional engineers were less unthinking than hourly paid line workers in their acceptance of orders from above. They would do what they were told if they could see a good reason for it, but not otherwise. As John Reed’s reaction to a hostile performance review had shown, criticism that was too harsh could easily be counter-productive. So in 1971 Grove was caught between a rock and a hard place. On the one hand, he was answerable to Noyce and Moore as operations manager for getting prototypes into production at reasonable quality, cost and speed; on the other, there was a limit to how hard he could drive the engineers beneath him to meet his goals.

When the pressure built up, it was in meetings between manufacturing and marketing that the safety-valve would blow. Bob Graham, who had had time to go fishing so often in his first year at Intel, was now banging the table regularly, complaining that his customers in the computer business weren’t getting the parts they had been promised, and demanding to know why Grove had not met this production forecast or that yield projection. More frustratingly still, he would also complain when Grove delivered parts that he had not asked for. Graham had adopted the sales slogan of ‘Intel Delivers’ for use outside the company. He was damned if he was going to let Grove get away with failing to deliver to him.

On one occasion Graham was put in a particularly embarrassing position. Intel’s Japanese distributor had a number of customers waiting for matching sets of memory chips and the drivers that were needed to make them work. What Grove actually shipped to Japan was a number of incomplete sets. When an incomplete consignment arrived at Yokohama, the distributor sent Graham an angry telex: ‘HAVE RECEIVED MUSKETS STOP AWAIT BULLETS AND POWDER STOP’.

After a series of furious confrontations, in which Graham taxed Grove with sending parts that he couldn’t sell and failing to send parts that he had promised to sell, a compromise was brokered by Moore. In the company’s internal financials, Grove’s organization would receive credit only for boxed stock that was delivered in accordance with the production forecast. Any parts Grove produced that Graham had not been forewarned about would come ‘free’.

What made the disputes particularly bitter was that Graham and Grove were of exactly equal status. They were founder-employees, though not founder stockholders. Both of them had been given stock options along with every other Intel hire, but Noyce and Moore never even offered to sell them a chunk of shares at the outset. The pair also received promotion at the same time: when they were at a conference together in Tokyo, a congratulatory telex arrived from Noyce informing them that their job titles had been raised to vice-president.

From where Graham sat, it was clear that Grove knew how to take orders from Bob Noyce and how to ask advice from Gordon Moore; and he knew how to keep the people who reported to him in line. But Grove seemed unable to work cooperatively with a manager who was one of his peers. Graham resented what he saw as Grove’s constant attempts to expand his own empire by ordering Graham’s sales and marketing people around – and sometimes by trying to force Graham himself to take his advice.

‘I’m not that kind of person,’ Graham recalled afterwards. ‘You can’t tell me what to do. You can tell me what you’d like me to do. You can tell me what you’re going to do. But you can’t tell me what to do unless I agree to it … [Yet] Andy’s tendency was to tell you what the ads ought to be like and what ought to be in them, when they ought to run, how much they ought to cost and who the reps ought to be. He was delving into areas where he had no expertise.’

To make matters worse, Grove had a strong tongue. For Noyce and Moore, both brought up in God-fearing households where cursing was very much frowned upon, it was little more than a tease when Grove got up from the meeting table and announced his intention to ‘go for a piss’. But Graham’s resentment burned red-hot when Grove told him day after day that he was a ‘stupid sonofabitch’ or a ‘bastard’ or a ‘dumb shit’.

Grove also knew how to make the best of a disability. In his early years at Intel, the hearing problem he suffered from required him to wear an ungainly hearing-aid that wrapped around his ears and over his crown like a pair of headphones. Just at the point when he was being told something he did not want to hear, the device would seem to fail – and Grove would interrupt loudly, bellowing ‘Huh? Huh?’ He was also able to use the device as a weapon of battle. When a speaker at a meeting was running over his allotted time or straying from the point, Grove would take the hearing-aid off his head and, with an eloquent gesture that said more than any complaint or expression of boredom, thump it down on the table to indicate that he would listen no further.

The last straw for Graham was a dispute over something apparently trivial: the data sheets on its products that Intel sent to engineers, providing technical specifications and performance information. When the 1103 chip was ready to ship, Graham discovered that the bipolar driver circuits that went with the 1103 wouldn’t work with the chip over the full temperature range that the memory chip itself could tolerate. He immediately took up the problem with Grove.

‘I could not get Andy to understand that the operating range over which the memory would work, the drivers needed to work too.’

Soon it became clear that modifying the drivers to make them more tolerant of extremes of temperature would be technically unmanageable: it would raise the cost and make the devices clunkier. Graham then decided to aim for the next best thing. He told Grove that if the drivers and the memory chip would not work together across the memory chip’s full temperature range, then the data sheets would simply have to be frank about the narrower temperature range in which they worked.

Grove disagreed. Few customers would care about the issue, he said, since their computers were already adequately air-conditioned and they would be unlikely to test the performance of the Intel parts below freezing or close to boiling point. Even if they did, the problems with temperature range would never be visible to the outside world. Tested on their own, the drivers worked across the full range; it was only when running together with the 1103 that their range narrowed.

Graham insisted that the honest thing to do was to make the problem clear on the data sheets. He wanted to eliminate even the tiniest risk that Intel might have to compensate a customer who discovered that its engineers had been misinformed. But Grove would not be moved.

Day by day the dispute escalated, until finally it reached the point where the two men were no longer on speaking terms. Noyce was called in to adjudicate. Late one morning he ordered Graham to publish the data sheets as Grove had specified.

‘Bob, we can’t do this,’ Graham replied.

‘Do it anyway,’ snarled the Intel founder.

Thirty minutes later Graham was sitting at a restaurant table facing his wife, Nan. He had got into the habit of calling her and inviting her out to lunch when the internal battles got too much for him, so she knew that this morning there must have been a bad one.

Graham looked into his drink. ‘Nan,’ he said, ‘I’m not working at Intel any more.’

His wife’s eyes filled with tears as she recalled the pressure they had been under during recent months, and realized that it would now all be behind them. ‘Thank God!’ she said.

On his way from the confrontation with Noyce to his lunch date with his wife, Graham had cleared his desk. He wasn’t fired; he had simply been presented with an ultimatum: either leave or publish the data sheet the way Grove wanted. At that moment he’d recalled that it was precisely in order to get away from issues of politics and style that he had left Fairchild. If Intel, for all its fine words, was going to revert to the characteristics of his old employer that had so dismayed him, then he wanted no further part in the venture.

For Noyce and Moore, the decision was tough but straightforward. It had increasingly become clear that Intel was too small a company for both Graham and Grove. They would therefore have to choose which one to lose. Whatever the rights and wrongs of the specific dispute that had precipitated the decision, Grove was clearly more vital to Intel’s success then Graham. The marketing vice-president was talented, experienced, focused, well-connected and energetic. But the operations vice-president was more than that. He was a figure of towering intellect, an obsessive chaser of details, a fearless fighter who would allow no personal friendships or loyalties to get in the way of what he believed to be best for the company. He also possessed willpower, self-discipline and determination in quantities that were given only to one person in a million.

If Noyce and Moore wanted their young company to fulfil all their ambitions, they could not rely on their own combination of inspiration and scientific insight. It was inevitable that tough decisions lay ahead. Research projects would have to be axed. Talented engineers would have to be fired. Requests for pay rises would have to be refused. Troops would have to be marshalled to drive imitators and competitors from the market. Noyce and Moore could not do these things on their own; in many cases, they shrank from the force that was needed. They needed a really tough manager. And neither of them had ever come across anyone tougher than Andy Grove.

There was just one aspect of the Graham affair that was surprising. Bob Graham and Gordon Moore had been great friends. Not only had they fished together, Graham and his wife often went to have dinner with ‘Gordon and Betty’, and Graham considered Moore his most important mentor inside the company. But after Graham’s encounter with Noyce, Gordon Moore was nowhere to be found. Graham was of course too proud to ask him to intervene. Staying out of the dispute altogether, Moore had nothing to do with the negotiation of the departing vice-president’s severance terms. From the moment of Graham’s fateful conversation with Bob Noyce, he neither saw Gordon Moore nor spoke to him for the next five years.

The departure of Bob Graham marked a turning point in the structure of Intel’s leadership. Although Bob Noyce and Gordon Moore were still CEO and executive vice-president, while Grove was merely one of three other vice-presidents, it was clear that the power of daily decision-making was passing – less than three years after the company’s foundation – into Grove’s hands. It would be another fifteen years before Grove received the title of chief executive, but from 1971 onwards he was to be the dominant influence over the company and its culture.

Shortly after Bob Graham left Intel the company began preparations for an initial public offering. There were just enough weeks left before the IPO for the company to hire in a talented new marketing vice-president whose name could appear in place of Graham’s on the prospectus.

A headhunter hired to carry out a search came up with a number of names, of whom the most promising was Ed Gelbach, an intolerant but charismatic figure who ran the national sales operation for Texas Instruments, and was suffering homesickness for the beaches of California. Asked by the search agent whether he would consider a move from TI to another firm, Gelbach replied that he had no interest in joining any other company – unless it was Intel.

Gelbach seemed a perfect match, but Noyce and Moore were determined to avoid a repeat of the Graham episode. Before hiring him, they handed the candidate over to Grove, making it clear that since Grove would have to get along with the new vice-president of sales and marketing, he might as well have a veto over his appointment. The outcome of the meeting was that Gelbach and Grove came to an accommodation, and the TI manager agreed to join Intel.

By midsummer 1971 Ed Gelbach was back in California with his feet under the desk. Every bit the expert negotiator, he exacted a price from Intel that was fully commensurate with his talents – including options to buy 20,000 shares, almost 1% of the company, at $5 apiece. Since it was clear that Grove did not believe that sales or marketing really mattered, he also took care that he would have the powers and the budgets to carry out his responsibilities.

When the IPO took place in October 1971, there were few surprises in the prospectus. Gelbach’s name appeared in the management section above Grove’s. Many of the backers who had put up the money to fund Intel’s creation were revealed as members of the Traitorous Eight, the crew that had left Shockley’s operation en masse to set up Fairchild in 1957. The company’s two biggest shareholders were still Noyce and Moore, with a combined holding of over 37% worth nearly $20m at the flotation price of $23.50. Although the prospectus showed that both men were paying themselves less than $30,000 a year, neither of the two founders was proposing to sell a single share from their holdings. Both men clearly believed that Intel still had far to go.