(FORTUNE Magazine) – It is worth remembering, now that mighty Intel has fallen from grace, that between 1985 and the turn of the century, this company pulled off one of the most amazing extended runs of technological, manufacturing, and financial success in business history. During that time, the company that invented the microprocessor--the fingernail-sized dynamo providing the intelligence of most computers--increased the calculating power of that chip more than 10,000-fold. Its pristine clean rooms pounded out more than half a billion such chips, as well as hundreds of millions of other kinds, with few glitches. In 1985, it was the world's tenth-largest semiconductor maker; by the early '90s it had vaulted to No. 1; and last year it rang up profits of $10.5 billion on sales of $33.7 billion, thirty times the sales it posted in 1985. Investors were well rewarded--$1,000 invested in Intel shares in 1985 had become a stake worth nearly $250,000 at the stock's high-water mark on Sept. 1, 2000.

Now that stake has shrunk to $65,000. Intel's stock has been pummeled in the last 12 months, as the Internet bubble burst, consumer spending on computers slowed, and corporate IT managers pulled back as well. After September's terrorist attacks on New York and Washington, D.C., hoping for a significant tech rebound seems naive at best. To make matters worse, for nearly a year Intel has been in the uncomfortable and unfamiliar position of playing catch-up to its longtime Silicon Valley foe, Advanced Micro Devices, whose Athlon microprocessor started consistently beating Intel's Pentium III in benchmark tests, and is considerably cheaper to boot. Partly as a result, Intel's traditional, Microsoftean 90%-plus market share for PC microprocessors slipped into the high 70s earlier this year, leading the company to slash the premium prices it has always been able to charge. The result: Intel's annual revenues are expected to decline 20% this year, to about $27 billion, its first contraction since it pulled out of the memory-chip business back in 1985.

The facile explanation for Intel's swoon is that the company has lost a step in the 3 1/2 years since legendary CEO Andy Grove yielded his title to Craig Barrett, formerly Intel's chief operating officer. Barrett, 62 years old, is a tall, laconic former Stanford professor, who gets much of the credit for masterminding the company's manufacturing and technological prowess during the '80s and '90s. He seemed a promising successor, yet the fact that the company's dive has come on his watch makes him temptingly easy to blame.

The truth is a lot more complicated. Intel's fall is the result of an unholy combination of operational gaffes and strategic misjudgments--several originating during Grove's tenure. Together they provided AMD the opening it needed to make its first serious attack on the most lucrative part of Intel's business: high-performance PC microprocessors and the ancillary chipsets that support them. All it took was the flaky post-Internet bubble economy to amplify Intel's foibles into a full-blown mess.

That story has never been fully explained, and to do so we're going to have to take you back some 15 years, to when Intel was just another struggling American chipmaker. But going that far back helps set the stage for another story as well, the tale of Intel's imminent comeback attempt. As battered as its share price is, and as humbling as the last 18 months have been, Intel--the other beast of Wintel PCdom--is finally rebooting itself to come roaring back. With the recent announcement that it would begin shipping its ripsnorting new two-gigahertz Pentium 4 microprocessor later this year, Intel has reclaimed the performance crown, outclocking AMD's fastest Athlon microprocessor by more than 40%. Just as important, Intel seems to have rediscovered its inner Attila, announcing even more ruthless price cuts, unleashing its lawyers on competing chipset makers whose circuits aid and abet AMD's microprocessors, and accelerating and diversifying its own line of chipsets. (As we will see, so-called chipsets are almost as crucial as the microprocessors themselves in this high-stakes game.) Says Lehman Brothers analyst Daniel Niles: "Intel was losing out in its knife fight with AMD, and so they decided to pull out a gun." That's a creepy metaphor these days, but it's accurate: Intel has never been a nice company, and it's about to get a whole lot meaner.

The monopoly that never was

Ever since the IBM PC was introduced 20 years ago, Intel, headquartered in Santa Clara, Calif., has tacitly encouraged a grand illusion: namely, that it owns a monopoly in the market for PC microprocessors, just as Microsoft dominates operating system software. That's the main reason why its stock, like that of its counterpart in Redmond, Wash., is so widely held and commands a premium price/earnings multiple. But in reality, even though Intel got the "design win" for the 8086 microprocessor that powered the first PCs, it has had competition of some sort every step of the way. One of IBM's conditions before awarding that 1978 contract was that Intel must license to several other chipmakers (including IBM) the right to make the processors. Big Blue wanted to be assured of a constant supply of chips--its goal back then was to sell perhaps 100,000 PCs a year. Other licensees included National Semiconductor, Zilog, NEC, Fujitsu, and, just down the road in Sunnyvale, AMD.

Over the years, obviously, Intel demonstrated its ability to supply by itself virtually all of the processors that IBM and the growing swarm of PC "clonemakers" needed. Many of those "second-source" licenses expired, while others were amended to strictly limit how the various manufacturers could use them. More upstarts entered the fray--Cyrix and NexGen among them--but ultimately were absorbed by others.

"There hasn't been a day that we didn't have competition," declares Grove. "The process of competing with all these players made us better, more credible, and earned us more trust. The names of those competitors changed and changed and changed, but the name of Intel has become a lasting brand."

The one competitor that stubbornly stuck with it was AMD. Like Intel, the company was founded in the late 1960s by an alumnus of the original Silicon Valley chipmaker, Fairchild Semiconductor. And ever since, Jerry Sanders has behaved like an envious and quarrelsome younger sibling. Flamboyant, combative, and controversial, founder Sanders, who is still AMD's chairman, aped every move of the Intel brain trust. As Intel changed the focus of its business from discrete logic circuits to memory chips and microcontrollers, and finally to microprocessors and "flash" memory chips (devices used in digital cameras, handheld computers, and cell phones), AMD followed right behind.

Occasionally AMD would beat Intel at this game--AMD is today the leading supplier of flash memory. But in the extraordinarily lucrative microprocessor business, the smaller company barely made a dent. And whenever it did make headway, Intel's zealous marketers found some way to thwart AMD, often by putting the squeeze on customers or by offering them advertising subsidies that they couldn't refuse in return for exclusive supply deals. Fed up, Sanders in 1987 filed an antitrust lawsuit charging that Intel was stifling competition and unfairly keeping key chip technology out of AMD's hands. The suit triggered an eight-year legal donnybrook that ended in arbitration. And during those eight years, Intel's annual sales grew eightfold, from $1.9 billion to $16.2 billion. AMD struggled to increase sales, from just under $1 billion to $2.5 billion, a mere 150%. At times, AMD was rumored to be heading for bankruptcy.

In the 1995 settlement, Intel agreed not to challenge AMD's license as long as the company refrained from developing microprocessors that could simply plug into the socket on the PC "motherboard" where Intel's Pentium chips neatly fit. The net result was that AMD was free to stay in the microprocessor business--provided it could not only design and build a competitive chip, but also come up with its own design for the rest of the circuitry that makes up a PC.

"The settlement upped the ante for AMD," says Mike Feibus, a chip-industry analyst for Mercury Research in Phoenix, "but it also upped the potential reward. To compete, AMD would have to find some way to design a whole new platform--namely unique chipsets and motherboards that would be put together quite differently, yet that were essentially Windows-compatible. If they could pull that off, they could develop an alternative PC architecture that was effectively their own." That's just what AMD did, with a little help from some friends.

Of chipsets and gigahertz

The computer on your desk works as well as it does because of two things: a microprocessor, which typically costs the PC maker several hundreds of dollars; and a chipset, which is usually a pair of semiconductors that together cost PC makers between $20 and $50. The microprocessor gets all the ink, both in the financial press and in the computer makers' ads, which proudly trumpet the speed of their expensive chip. But in the case of Intel and AMD, the lowly chipset is what finally transformed their rivalry.

The chipset directs the flow of data through the various components of a PC. It shuttles information from a hard disk or from memory to the microprocessor and back again, coordinates how graphic images are displayed on the screen, manages instructions entered via keyboard, mouse, or network connection, and even routes digital images from a camera to a printer or music downloaded from the Net to an MP3 player. Think of the chipset as the undercarriage, transmission, and drive train of a car. Without it, a microprocessor is like a high-performance engine sitting on a couple of sawhorses.

Chipsets contribute just a fraction of Intel's revenues and profits, but for years they have been the key to the company's domination of the PC hardware business. By ruling the chipset market, Intel could enforce a standard hardware design--which it calls the "Intel architecture"--for all PCs from all vendors. For PC makers, the benefits are fairly obvious--using Intel's chipsets and architecture guaranteed absolute software compatibility, and saved PC makers the redundant cost of designing, testing, and making their own chipsets. It also gave add-on hardware makers a single standard set of interfaces for plugging their circuit cards and multimedia devices into a PC. Intel also tailored chipsets to create uniform internal architectures for other PCs besides conventional desktop models--high-powered Wintel workstations, servers, and mobile computers. (The company would like to impose the same uniformity on cell phones, handheld computers, and networking gear. See "Platforms 'R' Us.")

Intel played the chipset game masterfully, but its original goal was not to become the de facto R&D lab for the PC industry. "Truthfully, we first entered the chipset business in the late '80s thinking 'Let's get some extra value,' " says board member Les Vadasz, who is in charge of Intel's venture investments. That meant having its world-class designers create chips that could be profitably made in the company's older, fully depreciated chip factories. Those "fabs" weren't advanced enough to make Intel's latest, greatest microprocessors. "But over time it became more strategic, as we got economies of scale the individual PC makers couldn't match, and as it became clear that we, better than anyone else, could manage and drive a consistent, steadily improving standard for the industry," Vadasz says. Indeed, in the early '90s, Intel even manufactured most of the "motherboards," the circuit boards that held practically all the innards of a PC, and then shipped them to PC makers.

Needless to say, that didn't leave much room for AMD. But then Intel goofed.

By the end of the '90s, Intel's fastest Pentiums had become so adept at crunching data that they often had to sit, like an idling motor, waiting for data and instructions to trickle in from memory chips. The reasons for the bottleneck are arcane, but Intel thought it had found a solution in technology developed by a Silicon Valley startup called Rambus.

The Rambus design did dramatically accelerate the speed of memory. But it required memory makers to incorporate expensive alterations into their chip designs, improvements that increased the manufacturing cost of a typical PC by about $50. What's more, the Rambus technology required special chipsets, and the early ones Intel delivered were buggy enough that some had to be recalled. Consequently, both PC makers and PC buyers shied away from the new systems.

This left an opening for AMD. In the years after the 1995 settlement of the old lawsuit, AMD had been working with partners to develop chipsets that both enhanced performance and reduced the costs of other parts of the computer. "We call it our 'Virtual Gorilla Strategy,' " says founder Sanders. "We're not big enough to be a gorilla like Intel ourselves, so we work with partners like VIA Technologies in Taiwan to design and manufacture our chipset platform." When Intel stumbled, AMD was ready with a chipset technology called DDRAM that basically doubled the transfer speed of data from conventional memory chips. Not only was the DDRAM platform cheaper, but it also enabled AMD's Athlon processors to outperform Pentiums with higher speed ratings. Suddenly AMD, which historically had made mainly chipsets for budget computers, had a shot at the richest part of Intel's market--premium-priced, high-performance PCs.

Intel, not wanting to undercut its Rambus strategy, stubbornly refused to build DDRAM chipsets for its Pentiums. That left its customers with three choices: They could buy Intel microprocessors and chipsets with a memory bottleneck, they could absorb the risk and higher cost of Rambus chipsets, or they could switch to AMD. Many chose the latter: AMD gleefully saw its market share increase from about 13% in December 1998 to 20% by the end of 2000.

What made the flight of customers all the more galling and costly for Intel was that in March 2000 the first PC microprocessor whose speed measured one gigahertz--one billion clock cycles per second, as opposed to the millions that were the measurement norm before--was introduced...by AMD. That ended Intel's 19-year reign as the maker of the world's fastest PC microprocessors, compounded Intel's Rambus chipset miscue, and made AMD's Athlon processors the favorite of speed-freak geeks, a group of customers willing to pay for performance.

How could Intel have let this happen? After all, it has far greater resources than AMD--armies of engineers, flotillas of fabs, and very deep pockets. "We got caught between chip generations, it's as simple as that," says Paul Otellini, executive vice president and general manager of the Intel Architecture Group. "Each design generation can generally be improved by a factor of ten in its lifetime, and the Pentium III had reached the limit of what we could get out of it. But we just weren't ready to make the shift to manufacturing the Pentium 4, which is a completely different chip design that we think ultimately could run at ten gigahertz. One reason was that in the last quarters of 1999 and in early 2000, we were selling everything we could make. The whole industry was sold out. So we didn't have the fab capacity to start ramping up production of the new architecture, and others exploited that."

"I really don't think there was was a screwup on Pentium 4," says Lehman Brothers' Niles. "Intel always intended to bring it out in volume later this year. What they didn't anticipate, and didn't seem to accept once it happened, was that AMD could execute so well and hold its lead for so long."

Gestation periods for new chip generations can't be speeded up all that much, so Intel stuck to its original timetable for the Pentium 4, introducing its first models, rated at 1.5 gigahertz, in limited quantities and with limited chipset choices last November. But Intel was able to accelerate the development of chipsets for the Pentium 4, shaving six weeks off the timetable for an especially important version. Called the i845, this chipset, which will become available in the fourth quarter, is designed to couple cheaper DDRAM memory with a Pentium 4 running at two gigahertz and costing a mere $562. This is the chip that Intel announced in September, and is the version that will entrench the Pentium 4 in the broad consumer market. Otellini claims that PC makers that choose this combination will be able to offer consumers truly high-powered PCs that will cost under $900. That's right in the heart of the "value-priced" consumer PC market that AMD had always claimed as its stronghold. And that is how this knife fight has become a gun battle.

Attila reborn

So here's where everything stands today: The PC business, which is projected to shrink this year for the first time ever, is locked in a price war that has made money-losers of nearly everyone, and that is in large part responsible for driving Hewlett-Packard and Compaq into a shotgun marriage. That price war reached the microprocessor business on Sept. 4, the day Intel introduced its two-gigahertz Pentium 4 at what Mercury Research analyst Feibus calls "Crazy Eddie prices," and announced accompanying price cuts of up to 40% across the rest of its Pentium line. AMD was forced to counter, slashing prices of its Athlon chips by as much as 49%.

Price cuts are only part of Intel's new scorched-earth policy. The company has also unleashed its lawyers on VIA Technologies, AMD's primary chipset partner. Intel sued to prevent the $1 billion-a-year company from making DDRAM chipsets for Pentium 4 chips. VIA CEO Wen-chi Chen, who turned around and sued Intel right back, was not surprised by Intel's belligerence. "Intel has always been tough, and sometimes they're too tough," he says. "They've done this to companies like mine before. But this is a $200 billion industry, and we think the market is big enough for all of us now. And we also think the courts will take our side."

Perhaps more significant, Intel has dispatched its salespeople to put the hard sell on PC makers, offering special advertising subsidies and volume discounts for exclusive deals. Early this year, Dell Computer was the only major PC maker that shunned AMD's chips. But over the summer IBM phased out AMD-powered PCs in North America, and MicronPC announced its intention to begin selling only "Intel Inside" models. Then, in late September, Gateway announced that it, too, would use only Intel microprocessors. The next day, AMD, which analysts expect to post at least a 10% drop in sales for the third quarter, disclosed that it would fire 15% of its 14,000 employees and shutter two chip fabs.

AMD's Sanders can barely contain his emotions when he talks about Intel. "This is a fascist organization that uses intimidation of their customers and the rest of the infrastructure guys to make our lives extremely difficult," he says. "This is their 'cut off the air supply' strategy. No matter what they say, they want to be an out-and-out monopoly. If we're gone, there is no competition, and if you believe, as I do, that competition is the key to innovation, we'll lose that too."

Hold on, he's not finished yet. "There's another reason we're attacking their business model, and that is that the entire PC hardware industry, thanks to Intel, has become an unstable molecule. It's unreal, it's unbelievable, that a company is making as much money as they are when all their customers are suffering huge losses, and have delegated most of their R&D to them."

Not surprisingly, Intel CEO Craig Barrett, who by contrast exudes the unexcitable demeanor of a Presbyterian minister, doesn't buy a whit of Sanders' argument. "It's true that computer makers invest a very low number in R&D, relative to the amount of technology that goes into a PC. But if everyone was doing different variants, the industry would move forward more slowly. Someone has to play the role of organizer." Moreover, he adds, "The health of the PC industry has to be driven primarily by their own efforts--their channels of distribution, their manufacturing execution, their marketing. All these things are under their control. We provide a level playing field, and help with promoting the platform. We provide air cover, if you will, and that applies to everybody."

There's a problem with Barrett's metaphor, of course, which is that air cover tends to devastate the terrain below. The PC industry today is, in fact, less a "level" playing field than a playing field that has been leveled. Like Dell, Intel seems to have decided to savage its weaker competitors at a time when the whole industry has been starved by a recession. Says Dan Niles of Lehman: "Intel was faced with a decision between maintaining its historically high profit margins at the cost of a lower market share, or to take a beating on margins and take back market share. A down market is a good time for the strong to squeeze the weak, so they went for market share." In the short term, that's a problem for investors, as Dell backers have learned. But this is clearly a long-term bet.

So, can Intel pull it off? It's a fair question, given that the company seemed to have fallen asleep during the transition from Grove to Barrett. The answer may well be revealed by the way the company has reexamined itself after the failures of the last two years.

Intel is a company of, by, and for engineers, and hence a company that tries to measure and quantify just about everything. (When its human-resources managers set out to design a companywide sabbatical program, they went so far as to analyze career tracks to try to determine if there was any regularity to employee ennui. Sure enough, every seven years, regardless of job, there appeared to be a productivity lull.) So the problems that gave hope to AMD have not escaped scrutiny.

"We realized, about 18 months ago, that we had some problems making timely strategic decisions about the various PC chipsets and architectures. Back then, the proposals for a new chipset or microprocessor strategy would come up independently through separate organizations, and got the final go-ahead from the CEO," says Otellini. "We didn't make a big deal about it at the time to the outside world, but we reorganized management so that those decisions are now made three layers down the chart from the CEO, by the managers of the different platform groups, the guys who are living in those markets."

Then, about one year ago, says Otellini, he "ordered up a study of how Intel made the Rambus chipset decision to see if we might have made a mistake somewhere." The conclusion: The Rambus technology is still worth pursuing, but the company should have been more open to other approaches too. "From this point on," says Louis Burns, who heads Intel's desktop PC platform group, "we'll offer chipsets that support whatever type of memory our customers want."

Grove's explanation for the missteps sounds a whole lot less geeky: "The past two years have been rough, no question, and there's a whole litany of factors, but there's one you probably haven't heard: The disappearance of our COO. When Craig got kicked upstairs he could no longer play that old role, and it took a while before people spread their attention, capabilities, power, influence, and competence into the vacuum he left behind. Only now have we gotten back to executing like we ought to." In other words, it was a human failure, not a design flaw.

All this helps explain why Barrett seemed such a promising successor: Like Grove, he is an engineer who has shown he can excel at the personal side of the business. That duality is reflected in the way he approaches his job as CEO. Barrett still logs thousands more miles each year than Andy Grove ever did, visiting customers and Intel facilities around the world. And he still occasionally holds what some Intel engineers call "prayer meetings" to inspire technology-development groups he thinks are falling behind. Says he: "If you or anyone at Intel asks me what keeps me awake at night, I'll say the same thing I've been saying for more than ten years: I worry about the internal execution of our product road maps. If we do that, we win. If we stumble there, we give the competition a chance."

They did that once, of course, and AMD stepped right in. But Intel seems to have learned its lesson. And that may spell big trouble for its little brother in Sunnyvale.

FEEDBACK: bschlender@fortunemail.com