The world's fastest growing manufacturer of semiconductors wants to go even faster. Nick Hasell examines the cost of staying ahead.
Andrew S Grove, president and CEO of Intel Corporation, lays three small, metallic squares on the desk in front of him. For those who like their corporate histories concise, the squares offer a neat three-step guide to Intel's recent fortunes. The first and smallest, Grove explains, is the 386 microprocessor, Intel's cash cow. Since its introduction in 1985, the 386 has proved the highest selling chip in the history of semiconductors. Next, in the middle, is its younger, more powerful relative, the 486, Intel's state-of-the-art offering and current revenue-earner. Lastly, Grove comes to the largest square, a microprocessor simply code-named P5. With the ability to process 100 million instructions per second - five times the speed of the 486 and 20 that of the 386 - its revolutionary performance is tipped to change the course of personal computing. All the signs are that, when launched in the first quarter of next year, it will duplicate the success of its predecessors and take Intel effortlessly and lucratively into the silicon future.
If all this seems impossibly neat, then rightly so. Grove, for one, is only too aware of what lies behind and, potentially, what lies ahead. At 56, he has been immersed in semiconductors all his working life. He arrived in the US from his native Hungary in 1956, enrolled in college and emerged several years later with a doctorate in chemical engineering from Berkeley. From there he joined Fairchild Semiconductors, the parent of the modern semiconductor industry. While working in R and D he met Gordon Moore and Robert Noyce, two of Fairchild's founders who had already pioneered a string of innovations in microelectronics. Finding that they were spending less and less time in the laboratory and becoming increasingly absorbed in administrative work, Moore and Noyce left in 1968 to form Intel (short for Integrated Electronics), one of 50 or so such start-ups to emerge from Fairchild. Within a couple of weeks Grove had joined them. From a one-storey building with just 12 employees and first year revenues of $2,672 Intel has become the world's fastest growing and most consistently profitable manufacturer of semiconductors.
As recently as the mid-'80s, Grove notes, things were rather different - "Just six years ago I was going round the company giving speeches to our engineers, asking them to stay and reassuring them that we were going to survive". The cause for the sermonizing was the sharp slump in morale following an unprecedented round of salary cuts, lay-offs and plant closures. Intel, it seems, had hit bottom. The trouble started in 1985 when the dynamic random access memory (DRAM), Intel's proud invention and main product line, began to lose money in the face of slumping domestic demand and predatory pricing from Japanese competitors. That year's profits tumbled from $198 million to $2 million, to be followed the next year by a $203 million loss - the first and only in Intel's history. Austerity inevitably came in its wake.
More critically, such losses forced Intel to change direction abruptly and abandon DRAM production altogether. The decision was all the more dramatic given that this was a technology that Intel had so conspicuously pioneered 15 years earlier. Accordingly, the resulting structural changes were severe. "We really had to tear the company apart and then put it back together again," recalls Grove. The switch away from memory chips to an all-out focus on logic processors required both a strong nerve and, in the long term, high levels of capital investment. The gamble paid off.
After low initial orders, the 386 was embraced by the majority of computer manufacturers, lead by IBM, and installed as the motor for their growing lines of PCs. Apart from the 10% of the market claimed by Apple, who opted for Motorola chips, Intel was left to clean up. Since the dark days of 1986, revenues have climbed steadily to last year's $4.8 billion high. Intel's pre-eminence was recently confirmed by Dataquest, whose research found Intel's average annual compound growth rate in revenues of 32.6% over the last five years to be the highest in the world among semiconductor companies.
Such colossal growth is matched only by the vast outlay required just to stay ahead. Advances in silicon technology do not come cheap and for any smaller player the economics can soon look unattractive, if not prohibitive. This year alone Intel is putting $2 billion (or some 30% of total projected revenues) into R and D and capital spending. As well as expanding an existing plant in New Mexico and opening a $300 million development facility in Oregon, a further $800 million is committed to Intel's first European manufacturing complex at Leixlip, just outside Dublin. Of this, the largest single outlay is for a silicon wafer fabrication plant. Gerry Parker, general manager of the manufacturing and technology group, points out that silicon fabrication demands not only fantastically expensive process equipment but also a sterile environment that is typically thousands of times cleaner than a surgical operating theatre. As a result, notes Parker, "$500 million will have been put in before the first wafer is produced. The actual building is less than 10% of the overall cost".
Success in semiconductors, it is held, is only achieved through high-volume production and continual product development. Yet beyond the innovations of any one company there seems to be an inexorable driving force at work, first identified in the late '60s by Gordon Moore, Intel's current chairman. Now enshrined as Moore's Law, it dictates that the number of transistors that can be built on the same piece of silicon will double every 18 months. Moore's Law has held good. In 1970 Intel introduced the world's first general purpose microprocessor, the 4004. It contained 2300 transistors. The keenly-awaited P5 contains over 3 million.
Despite such technical progress, Intel has found its competitive breathing space increasingly reduced. On one side, rival US chipmakers such as Advanced Micro Devices and Cyrix have steadily encroached on its valuable market for the 386 and 486 by offering lower-priced alternatives. On the other, its family of logic processors has come under threat from competitors using a different - and, in some applications, superior - chip architecture known by the acronym RISC (reduced instruction set computing). To counter the threat Intel is taking up Moore's Law with a vengeance, accelerating the introduction of its next generation of chips by developing more than one at a time. "Speed", reflects Grove, "is the ultimate competitive weapon. It's a cliche, I know, but true." Already the P6, projected to contain seven million transistors, is halfway through development at Intel's Oregon facility. Members of the P5 team, their work almost through, are now switching their sights to the P7. The only potential brake on this headlong rush, say experts, is the ability of computer manufacturers and software developers to keep pace. Indeed, some attribute the delay in the P5's launch, originally scheduled for this autumn, to the fact that Intel's principal customers aren't yet ready for it.
Intel's savage entrepreneurial appetite has found favour with Wall Street. Analysts there hail Intel as a "model cannibal", a relentlessly aggressive company willing to re-invent its past in order to feed the future. Industry observers point to the way it has taken on rivals through cutting prices and spinning off a host of "second wave" products from its existing lines. Intel has also sought to blur the distinction between RISC architecture and its own designs by incorporating its most innovative features - those responsible for its phenomenal speed - into the P5 and its successors. Through self-renewal and an ability to both out-spend and out-engineer its rivals, Intel has kept its edge. "Other companies still wait for Intel to lay out the roadmap," notes Brett Hodess, a semiconductor analyst with San Fransisco's Montgomery Securities.
Yet if others wait on you they also tend to imitate. Intel seems to find it less than flattering, as witnessed by its frequent pursuit of copycats through the courts. The courts, however, have not always ruled in its favour. Last year Intel failed in its attempts to prevent a rival using the numbers 386 to designate one of its microprocessors. Since then others have followed. The result in the marketplace has been an understandable confusion. "We don't own the numbers any more so we can't assure customers that the numbers mean compatibility", laments Dennis Carter, general manager of the corporate marketing group.
As part of its drive to fend off rivals, imitators or otherwise, Intel has sought to raise the end-user's awareness of its products. This need is seen to be all the greater given that it has also had to adjust to a new set of customers. In the past Intel sold only to a narrow base of computer manufacturers. Now, with the advent of retail products such as the 486 OverDrive processor, it finds itself addressing millions of PC users. Hence PC buyers are urged to ask for an Intel chip, "the computer inside", while for the first time TV viewers encounter advertisements encouraging them to upgrade their microprocessors. Perhaps most boldly, not to say wryly, the roof of Intel's new HQ, directly under the flight path of San Jose International Airport, is emblazoned with outsize letters announcing "Intel Inside".
While Intel has undoubtedly grown more focused (its core family of logic chips is now estimated to account for 57% of revenues, compared to 35% two years ago), this has not prevented it from exploring other market niches. Divisions concentrating on networking, supercomputing and "flash memory" all contribute significantly to the balance sheet. The latter, a credit-card sized alternative to the traditional hard disk, has assumed a particular importance with the growing trend in PCs towards portability; Intel predicts that by 1995 mobile computers will constitute 50% of all PC purchases. It has already shown its agility by cornering 85% of the worldwide market for flash memory and last year consolidated its position by signing a 10-year deal with Sharp.
Grove is a firm believer in the future of multimedia computing, where electronic mail, text, graphics, video and sound are incorporated into the PC. To this end Intel recently snapped up the rights to an image compression system, Digital Video Interactive (DVI), for an estimated $20 million from General Electric. "All of the technologies already exist." explains Grove. "It's just a question of packaging them together in a PC network." Grove's concern, however, is not just with the technology but with its implementation. Whereas the first decade of the PC improved personal productivity, Grove predicts that the second decade will be characterised by improved workgroup productivity through "computer-supported collaboration". Accordingly, Grove's vision of Intel in the '90s is of a company that enables such transitions to take place. And if multimedia computing truly takes off, Intel could stand to gain as much as it did from the microprocessor. The irony of all this is that Grove himself admits to having not recognised the potential of the PC early enough. "Here we were just two miles away from Apple and we didn't take it seriously. It set us back in a big way." Not, of course, that you'd notice it today.
For reprints of this article, contact Anne Oakley (071) 413 4336.