(FORTUNE Magazine) – Shortly before Jerry Sanders stepped down as chairman of Advanced Micro Devices last spring, Hector Ruiz, his handpicked successor as CEO, took the company co-founder aside for what Ruiz hoped would be a heart-to-heart talk. A legend in Silicon Valley, Sanders was a flamboyant engineer who over 35 years had built a business known for technical excellence, marketing spunk, fluctuating financial results--and for being a rare survivor among the dozen or so microprocessor outfits that have tried to compete against Intel. Sanders had willingly relinquished the top job to Ruiz, but he was having trouble letting go.

Ruiz recalls telling Sanders: "Jerry, when I was a boy, my father told me, 'Son, you need to be a better husband and father than I was.' I hope you're like that too, and want me to be a better CEO." All the earnest Ruiz got for his gentle approach was a snarl. Replied Sanders: "No one can be a better CEO."

With AMD flourishing as never before, Sanders may want to rethink that. Though it had a mere $3.5 billion in sales last year, barely more than one-tenth of Intel's, Ruiz's AMD seems to have proven once and for all that it can sustain a challenge against its giant adversary. By employing its own chip-design innovations and exploiting strategic missteps by Intel, AMD has built alliances with the likes of Microsoft, Hewlett-Packard, Sun, Fujitsu, and IBM. These tech powers mostly ignored AMD before, but now they see the chipmaker as a means to build market share by helping customers lower the cost of their IT operations. Almost overnight, AMD has become a major supplier of chips in the high-priced and high-margin world of servers, the big machines that power the Internet and corporate networks. In a grudging acknowledgment of AMD's prowess, Intel this year changed its technology strategy to copy the design of AMD's hot new server chip.

And for once, the ever volatile AMD seems to be showing that it can translate its gains into the kind of steady numbers that Wall Street looks for. After three years of losses, the company has racked up four consecutive quarters of profit. Revenues this year will probably top $5 billion, up more than 40% from 2003. Profits will exceed $150 million, compared with a loss of $274 million last year. In microprocessors, which account for slightly more than half of sales, revenues leaped 21% from the second quarter to the third, and average prices and profit margins surged. In the much more commodity-like memory business, AMD has become the global leader in a critical type of flash memory, the most expensive component in the world's hottest technology product, the cellphone. AMD's flash-producing Spansion subsidiary (created in mid-2003 when AMD combined its flash unit with Fujitsu's and took a 60% stake) has grown 27% in a year. Says Ruiz of AMD: "This company is in the strongest position we've ever been in."

So far, though, the jury on Wall Street is deadlocked: The company has as many negative analysts as positive ones. The stock is down 19% since Ruiz arrived, and at a recent $14, still trades slightly below its price of 20 years ago. But chip analysts--those with laser focus on this previously one-pony world--are wowed. Says Nathan Brookwood of the Insight 64 tech research firm: "AMD has some serious performance advantages over Intel."

More than just taking on Intel and winning, Ruiz, 58, is rebuilding AMD in his image--a tough job at a company that has long been seen as an extension of Sanders's outsized personality. Sanders started in sales at Fairchild Semiconductor in the late '60s before founding AMD, where he quickly set himself apart from his Silicon Valley peers. The gold-jewelry-sporting Sanders drove a white Rolls-Royce and commuted from his home in Bel Air, Calif. AMD executives say he had an answer for every question, whether or not he knew anything about the subject, and that he managed strategy single-handedly, making all critical decisions. The result was a company that Wall Street viewed as prone to inordinate risks and, too often, to losing its bets.

Ruiz is trying a different course. He worked his way up from a poor border town in Mexico to his current position by relying on his smarts, an almost photographic memory, and the help of others. And he's still inclined to listen where his old boss might have talked. Sanders used to host a regular quarterly meeting with executives called Breakfast With Jerry. Ruiz maintained the tradition, but he took his name off what's now known as the management committee meeting. Sanders would talk for the entire 45 minutes; Ruiz speaks for 20. "Under Jerry, frankly, the company was very autocratic and power-centric," says the candid and softspoken Ruiz. He's sitting in his Austin office, his back to a panoramic view of the city's skyline; next to him rest two gleaming electric guitars he plays to relieve stress. "But I said we're going to die here if we don't do something different." Says CFO Bob Rivet, who worked with both men: "Jerry's style was home run or strikeout, with nothing in between. Either you had a great year or it was a flaming disaster. Hector's more process-driven. Now we worry more about getting men on base."

Still, even though Sanders has gone into retirement, his presence is inescapable at AMD. The new AMD is neither a Sanders nor a Ruiz production, but a strange amalgam--both risk taking and partner seeking, forward looking yet quarterly driven, still superb in technology but now managed for predictability and solid execution. It just might be the formula that helps AMD not simply survive but thrive. To see how that can be, it helps to understand the inner workings--the history, the battles for market share, and the tech--behind the breakthrough chips that have enabled the business to come from behind.

Jerry Sanders hasn't been involved full-time with AMD since 2002, but he still knows what's driving the place: him. "AMD is in a better position than it's ever been in its history, and at the risk of sounding self-aggrandizing, it's because we're still following all the principles that were put in place in the first 35 years." Self-aggrandizing or not, he's right. Ruiz would never have ended up at AMD had Sanders not made a bet-the- company technology decision in 1998.

Back then, everything seemed possible. The Internet boom was accelerating and growth looked limitless. Intel had realized that with its vast size and huge infrastructure of chip- manufacturing plants, it could expand its reach from the PC desktop into companies' all-important data centers--the heart of their computing operations. With its eyes mostly on territory controlled by Sun and IBM, Intel had teamed up with Hewlett-Packard several years earlier to jointly design an entirely new chip architecture, ultimately named Itanium. Two big differences distinguished Itanium from Intel's so-called x86 chips, which powered almost all PCs then in use. Instead of processing data in chunks of 32 bits of information, the new chip would analyze 64 bits at a time, greatly increasing computers' speed and their ability to tackle complex tasks. But this new architecture would not use the same software applications as 32-bit x86 chips, except in a clunky "emulation mode." As the world switched to it and as it made its way onto the desktop, Intel customers would eventually have to buy all new software.

The chip colossus didn't worry much about its wee rival. For most of its history, AMD had defined itself as a "second source" of microprocessors. After a bitter legal battle in the '90s, it won the permanent right to build chips to Intel's x86 standard. It sold chips compatible with Intel's but at lower cost. Occasionally the performance of AMD's chips exceeded that of Intel's own, and the company's financial results and stock price would jump. Intel inevitably retaliated with lower prices or superior designs, though, making AMD's booms short-lived.

Sanders realized Itanium would put AMD in a bind. He would have no legal right to the new chip's design, and if customers switched to Itanium, AMD's license to make x86 would eventually lose its value. Yet Sanders also realized that Itanium could be a tar pit for Intel--if companies balked at buying all new software in order to get their hands on the latest hardware, they just might welcome an alternative.

In a meeting in 1998, AMD's technologists presented Sanders with the play he was looking for: a way to build a chip that would enable a move to 64-bit computing yet stay true to the x86 architecture. That meant new computers would be able to use both new 64-bit software and all the many thousands of existing 32-bit applications. AMD's chip design also adopted a more efficient way of connecting computer memory to the processor circuits and presumed that, down the road, each chip would incorporate more than one processor. Sanders began calling the design AMD's "Archimedes' lever." ("Give me a place to stand, and I will move the earth," the ancient Greek mathematician is said to have remarked.) He instructed the company that this was the future of AMD--everything else would take a back seat. Intel yawned.

Sanders, then in his early 60s, knew he might not be able to see his plan through. He was getting pressure from the board to hire a potential successor. So he made a second big bet, this time not on technology but on talent. He started courting Ruiz, who at 53 was running Motorola's $7.4-billion-a-year semiconductor division. Ruiz had decades of experience in the chip world and a much steadier personality; Sanders didn't pick a younger version of himself, but someone who complemented him. "I'm an impact guy; Hector's a process guy," says Sanders. "I got to know Hector and realized that he was a corporate kind of guy--he knew the details of inventory and supply-chain management, things that were not my thing."

Sanders says he first tried to buy Motorola's entire semiconductor group and merge it with AMD. When Motorola CEO Chris Galvin nixed that plan, Sanders went after just the group's boss. Galvin "threw tons of things at Hector to convince him to stay," says a colleague--"promises, money, and restricted stock." But Ruiz went to AMD for a $750,000 salary, the title of COO, and one million AMD options at $17.25 per share (they are still underwater). The most important thing Ruiz got was a verbal guarantee: Sanders promised he would turn over the CEO slot within two years.

For Ruiz, this represented the ultimate achievement in an amazing personal odyssey. He was born in Piedras Negras, Mexico, just across the Rio Grande River from Eagle Pass, Texas. His mother was 17. His parents named him Hector after the doctor who delivered him and gave him the middle name de Jesus because he was born on Christmas. Methodists in an overwhelmingly Roman Catholic country, they opened a religious bookstore, which his mother, now 75, still runs. In his early teens, Ruiz ran errands and cleaned house for an American Methodist missionary, Olive Givin, who lived nearby. She saw he was gifted, and one day asked what he wanted to do in life. "I said, 'I would love to be an auto mechanic,' " Ruiz recalls. " 'Well,' she told me, 'You're never going to be one unless you learn English.' That had me thinking, 'Oh, shit.' "

Givin worked out a plan for Ruiz to attend high school in Eagle Pass, an unprecedented arrangement. Since his family didn't pay property taxes, the school board decided it should charge him a fee to attend--$25 a month. Givin agreed to pay. Still knowing almost no English, Ruiz entered in tenth grade. Three years later he graduated as valedictorian. That led to a full scholarship to the University of Texas at Austin, where Givin paid his room and board. "Auto mechanics began to shift down in my priorities," deadpans Ruiz. He stayed and earned a masters degree, went on to a Ph.D. in electrical engineering at Rice University, and joined Texas Instruments in 1972, where he worked on the team that developed the first single-chip calculator.

Ruiz continued his way up, moving in 1977 to Motorola and rising to head the chip business by the late 1990s. He won a loyal following with a style of asking first and acting later; the initial question he almost always asked of peers and underlings was, "What do you think?" He also dazzled colleagues with his incredible memory; one AMD exec likens him to Star Trek's Mr. Spock, who could watch the ship's monitors and absorb everything. Says chief administrative officer Tom McCoy, a longtime Sanders confidant whom Ruiz persuaded to stay: "When Hector joined this company it was amazing how many people were pounding on the door--'Hector's at AMD, so I want to be at AMD. I don't care what the job is.' " Several top executives worked with Ruiz at Motorola, including CFO Rivet and Bertrand Cambou, now CEO of Spansion.

Things got rough not long after Ruiz arrived. As the chip project known as AMD64 progressed, the dot-com bubble burst and the bottom fell out of the tech market. AMD was hurt even more than most tech companies because the downturn coincided with a production problem that prevented it from churning out as many of its then-hot K6 microprocessors as customers wanted. Ruiz laid off 4,500 employees, mostly by closing two factories and axing unnecessary internal tasks--in one instance he cut AMD's 15-person travel department and told employees to book trips online instead. One area that was spared the rod was R&D. In 2002, AMD lost a stunning $1.3 billion on revenues of $2.7 billion; research expenses that year amounted to an equally stunning 30% of revenues. Ruiz kept hiring the chip architects he felt were critical to long-term growth--including 70 laid off from Sun--and acquired two small chip companies full of technical talent.

Sanders turned over the CEO job to Ruiz as promised in early 2002, but remained as chairman. This year the board decided Ruiz should hold both titles, so Sanders resigned entirely. "I didn't want to be co-pilot," he says.

Ruiz has already put his own stamp on the business, instituting a strategic shift that he calls "flipping the company upside down." While AMD had long focused on microprocessors for desktop computers, laptops, and servers, in that order, now it has reversed the list. It also now targets corporate customers. Characteristically, Ruiz refuses to take full credit for the change. Instead, he says, "I recognized the people inside the company who saw that as important and just gave them the encouragement." CFO Rivet explains why the flip was necessary: "As hard as we tried in the past to win the hearts and minds of CIOs, with the desktop as our focus we were going to fail. They made their decisions from the server on down. When Intel had 100% of the x86 server market, it could charge whatever it wanted and use that money to beat us on desktops. We had to be in the profit haven." Ruiz calls the server-led approach "do or die" for AMD: "If we hadn't pulled this off I would have shut the door."

The AMD64 design lent itself to the change because the new chips are much more powerful than any the company had sold before. So the first new chip launched, in April 2003, was one just for servers, called Opteron. In September 2003, AMD started selling a chip for desktop computers called Athlon 64.

Reviews, for both product and company, have been ecstatic. "AMD's top-end desktop processors are faster than Intel's for most applications," began an article this August from tech reviewer ZDNet. Says Rick Whittington, a longtime semiconductor-stock analyst now at Caris & Co.: "AMD has a better product and lower-cost manufacturing. In the history of technology that's a recipe for a major disruption." Adds Brookwood of Insight 64: "AMD has a far superior technical approach, at least near term, than Intel, and that advantage should be amplified over the next 18 months."

And this is all for a chip that has yet to fully flex its muscles. The praise today is almost entirely for AMD64's performance on 32-bit applications. Much of AMD's early success with its 64-bit chips has not been because they are 64-bit at all, but rather because they are so well designed in other ways. For instance, Opteron uses much less electricity and thus generates less heat than Intel's competing Xeon server chips--an important difference in corporate data centers. As software emerges to take advantage of AMD's 64-bit capability, the chips may deliver even better results. For now, only the Linux operating system runs 64-bit applications on AMD64. But by the end of the year Sun Microsystems will release a version of its industrial-strength Solaris operating system for AMD64; Microsoft has promised a 64-bit version of Windows by early 2005.

Ruiz has also started demanding that the company become obsessed with pleasing customers; he has spent much of his energy building new alliances with the biggest ones. While AMD in the past had periodically sold major PC makers its desktop chips, it never won a major server maker until last year. Now it counts as customers not just IBM and Sun but Intel's Itanium partner HP. For certain data-intensive applications like Wall Street trading, HP's tests showed Opteron brought performance improvements of "anywhere from 40% all the way up to several hundred percent" over Intel's latest Xeon line of 32-bit server chips, says HP server marketing manager Mark Hudson. He insists that HP remains committed to Itanium and believes sales won't be hurt by Opteron but also says, "Opteron has sustained legs in the marketplace." HP recently announced it would no longer sell Itanium-based workstations.

A major Wall Street firm told computer companies it intended to standardize on Opteron servers, and if they wanted to be its supplier, they had to offer them. AMD says 25 of the 100 largest firms on FORTUNE's Global 500 list now use Opteron servers in some substantial way. The chip really does perform better for many applications: Microsoft adopted Opteron for highly complex risk-management software it uses to manage its $60 billion in cash. The time to run the program dropped from 40 hours on Intel machines to 13 hours with Opteron.

AMD is also gaining critical PC-maker converts around the globe, notably in the world's fastest-growing PC market, China. Recently it sold to both the top local server maker, Dawning Information Industry, as well as the No. 1 desktop computer company, Lenovo. The deal helped Lenovo lower the price of its cheapest PC. Winning this one was a particular coup. Lenovo has been so close to Intel that then-CEO Andy Grove himself traveled to Beijing to be presented with the company's one millionth PC when it came off the assembly line in 1998. There are issues here that go beyond chip performance and price. China's government, a part owner of both Dawning and Lenovo, is acutely concerned about avoiding the grip of foreign monopolies in key industries. Talk about Intel's power and prices has been rising in China. Wrote a national Chinese tech magazine last year: "AMD for the time being is the most useful force ... [for] a very significant war of liberation." Says a pleased Ruiz: "The only deal bigger than Lenovo we could have done is Dell."

Dell, in fact, is the only major PC maker not buying from AMD. Last year Dell accounted for $5.7 billion of Intel's $30 billion in sales, and analysts assume that the PC maker gets some sweet deals from Intel as a result. Dell senior vice president Joe Marengi, though, says it's just a matter of what customers want: "We're always asking what best meets customer needs. In microprocessors, the answer has been Intel products."

This past summer Intel finally conceded defeat--by all accounts except its own. The company, which recorded just an estimated $250 million of Itanium sales in 2003, announced a 64-bit chip that, unlike Itanium, could truly run 32-bit software. Intel now says it always intended to eventually launch such a chip. But it was the first time Intel ever designed a microprocessor that, in effect, mimicked one invented by AMD. Even so, the Intel chip shows inferior overall performance to Opteron. Now Intel is smothering PC makers who chose Opteron with consideration, trying to woo them back. Says HP's Hudson: "Our interactions with Intel since we announced our Opteron products have actually become better. They realize we have an alternative. That's something they really need to take seriously."

So will Intel again squash the smaller chipmaker? That can't be ruled out, but both inside AMD and among chip analysts the feeling is that if the smaller company can get 30% of the overall x86 chip market, AMD will be too big for Intel to take out. Today AMD has about 17% overall. Sanders thinks the only risk is that Ruiz might soft-heartedly screw it up. "We have to move even faster; I don't think we're aggressive enough," he says. "I think maybe we're too focused on, you know, profit. The knock against me was always that I had too long-term a view of the business."

Ruiz says he's pressing plenty hard already. Now that Opteron is taking off, AMD got to 6.9% of the crucial x86 server market for the third quarter, more than double what it had a year ago, according to research firm IDC. Executives predict it will reach 10% by year-end and 20% by late 2005. HP, Sun, and IBM are selling AMD servers to giant global corporations.

Intel tries to avoid even acknowledging AMD's existence, though spokesman Howard High agreed to speak gingerly about it. "We have great technology and AMD has great technology, but I don't see them as being ahead of us," he says. He raises the specter of the old cycle repeating itself. "Right at this moment AMD's better than they've been in a long time. [But that's after] a three-year run of nonprofitable years. They tend to have point periods in their history when they are competitive."

To stop that boom and bust--a legacy of past overreaching--Ruiz has spent billions building up AMD's manufacturing skills. Its main processor fabrication plant, Fab 30 in Dresden, Germany, is at least as good as any of Intel's 11 state-of-the-art fabs. Chip-manufacturing expert Dan Hutcheson of VLSI Research wrote that it "holds many of the secrets to AMD's competitiveness." Indeed, Fab 30 helped AMD beat Intel in incorporating a number of chipmaking techniques, notably including copper in the process and using a technology called "silicon on insulator," which enables a chip to use less electricity and run cooler. AMD was less than two quarters behind its giant rival this year in starting to make chips in quantity with circuits only 90 nanometers across, the current state of the art. That's half the historical lag for AMD. Writes Hutcheson: "AMD has systematically matched and even exceeded the best with as little as a tenth the resources."

By early 2006, AMD expects to have yet another Dresden fab operating that will enable it to build far more chips at a lower cost. IBM is working with AMD to ensure the challenging manufacturing processes work as expected.

Now Ruiz wants AMD to be the company that helps the world's poor connect to the Internet (see box). It reflects another major change he says he's brought to the company: "I've helped everybody at AMD realize that there's actually a world outside the U.S. that's very important."

Sanders, not surprisingly, thinks the focus should be on battling Intel. "We've got these guys by the short hairs if we move quickly to 90 nanometers," says Sanders. "We've got to drive 90 nanometers hard."

But then, this isn't Sanders's game anymore. Hector Ruiz isn't obsessed with Intel, but rather with earning AMD some solid respect. He's already won it from major computer makers and a growing number of enterprise CIOs. Wall Street may be next. Says security analyst Whittington: "AMD is poised to take tremendous market share from Intel over the next four to seven years. But AMD's market capitalization is only $5 billion compared to Intel's $130 billion. That's too big a gap."

Ruiz certainly has won plenty of respect for himself. Not only has he taken over from a legendary, heavy-handed CEO, but also he's done it his way. Now the plain-speaking CEO is the most prominent Mexican-American businessperson in the country. When Mexican President Vicente Fox wanted to talk to Mexican-American business leaders, Ruiz organized and moderated a meeting of about 100 of them. Recently he received an invitation to the Quinceanera for the daughter of a Mexican-American AMD factory employee. In Mexican culture this celebration when a girl turns 15 is comparable in scale and importance to a wedding, he explains. "I have to find a way to go," Ruiz says. "It's important that people know that if they want to aspire to be in my shoes, they should be able to, because I'm no different."

To believe in AMD, you have to believe in Hector Ruiz. An increasing number of people do. Says Tom McCoy: "Don't underestimate the ambition of a poor kid from Mexico to change the world." Even Sanders now admits he's a pretty good CEO.