Posted on 11/09/2002 11:46:53 AM PST by Ernest_at_the_Beach
AMD's new CEO, Hector Ruiz, has one hope to save his company, a new microprocessor technology called Hammer. But first he has to shake AMD's spotted history -- and a ferocious competitor named Intel.
Life as the No. 2 maker of PC microprocessors in Silicon Valley has always been precarious, considering that No. 1 is the hypercompetitive Intel Corp. (INTC). But AMD is now as close to imploding as at any time in its 33-year history. And yet not only does the new boss appear unruffled, he even occasionally unleashes, seemingly from nowhere, a disarming, booming laugh.
When Ruiz joined AMD in early 2000, leaving a triumphant 22-year career at Motorola (MOT), his timing seemed exquisite. As the company's flamboyant founder, Jerry Sanders, groomed the quiet engineer and physics Ph.D. to be his successor as CEO, AMD was having its best year ever. But by the time the 66-year-old Sanders stepped aside (he remains chairman), the tech recession had utterly changed AMD's fortunes. "I think if Jerry had any thoughts of sticking around, they were taken away when he saw this coming," Ruiz says. "OK, here's your company!" he cries, parodying a jovial Sanders. "Good luck with it!" Cue the booming laugh.
Sanders did, however, leave Ruiz an intriguing legacy -- a project that might just bring AMD back from the edge and, in the process, embarrass Intel and fundamentally transform the industry. Code-named Hammer, it's a new microprocessor technology that promises to process data in bundles of 64 bits rather than the current 32, allowing one high-end server to do the work of hundreds at lightning speed. Slated for a debut early in 2003, it's AMD's alternative to Intel's next-generation superchip, Itanium, which some are already calling one of the biggest debacles in the history of computing. Both chips promise 64-bit processing, a technology found mainly in heavy-duty servers made by the likes of Sun and IBM. But while Hammer has received early industry support, Itanium has bombed since its release last year, giving Ruiz's beleaguered company what may be its last, best shot at survival.
So far, however, Hammer remains little more than a promise from a company known more for its salesmanship than its engineering and run by a CEO who's been on the job only six months. But if Ruiz can't deliver Hammer as promised -- and we'll know by this time next year -- it could mean more than just another vacant building in Silicon Valley. Without AMD, there would be nobody left to sell semiconductors to the mass market besides a monopoly named Intel. Those who know Ruiz well say it would be unwise to underestimate him. He grew up poor in the Mexican village of Piedras Negras, where he traded housework for English lessons from a Methodist missionary who later helped him attend high school across the U.S. border and paid part of his tuition at the University of Texas. Shortly before he joined Motorola's semiconductor division in 1978, he lost his wife to cancer, and he was raising their son alone as he began climbing through the ranks. When he took over Motorola's semiconductor business in the late '90s, Ruiz sold off a lower-margin unit and laid off thousands, while maintaining sales at the same level. "I've had some tragedies in my life, so you put it in perspective," Ruiz says. "This is not the most difficult thing that I've ever done."
But fixing AMD certainly won't be easy. Before he can expect corporate customers to commit millions of dollars to Hammer technology, Ruiz must, in his quiet, diplomatic way, overcome AMD's reputation for second-rate engineering backed by first-rate hype. Within AMD, he has to wring out an inferiority complex born of years of working in Intel's shadow. And most of all, perhaps, he must exorcise the ghost of Jerry Sanders.
All three elements of AMD's identity date back to the company's founding in 1969. Intel had been created a year earlier by Fairchild Semiconductor veterans -- and soon-to-be Silicon Valley legends -- Andy Grove, Gordon Moore, and Bob Noyce. Sanders also came out of Fairchild, but by the time he founded AMD, all the best minds were already at Intel. He scraped together his company from what was left. "AMD was a ragtag group of engineers and a couple of hustlers," says Steve Zelencik, a vice president who joined AMD in 1970. "We weren't inventing shit. We were copying Intel." At times, it seemed, all that held it together were Sanders's outsize chutzpah and salesmanship. The tech-industry chronicler Michael Malone dubbed him the "clown prince of Silicon Valley" for stunts like showing off his convertible Bentley and bragging about his wealth on 60 Minutes.
While it didn't earn the company much respect, copying Intel proved to be one way to make a living. As a licensed "second-source" supplier of Intel chips, AMD got access to the fruits of Intel's bright engineering minds; Intel, in turn, could assure its customers that AMD would step in whenever demand outstripped supply. Elaborate cross-licensing deals were struck, and AMD became the industry's top second-source supplier of processors found in IBM PCs. But as the PC market grew into a multibillion-dollar business, the amiable partnership between Intel and AMD soured. In 1988, Intel tried to terminate an agreement that allowed AMD to manufacture chips using its popular x86 processor technology, claiming that AMD had failed to live up to parts of the contract. Sanders says Intel simply got greedy and wanted the whole market to itself. While Grove disparaged AMD as "the Milli Vanilli of semiconductors," Sanders accused Grove of trying to run him out of business.
The court battles ended in 1994, and AMD was allowed to continue using Intel's x86 technology to make its chips. But the experience shook Sanders up. He decided there was only one way to survive against Intel: Stop lip-syncing Grove's tunes and try to create some monster hits of his own.
From that point on, AMD began to bet on its own brand of technology -- and the events that led to the showdown between Itanium and Hammer were put in place. Microprocessor development is a high-risk, slow-motion gamble, requiring multiyear lead times and bet-the-company investments in research and equipment. Robert Palmer, former CEO of Digital Equipment and an AMD board member, once likened the process to Russian roulette, but with a twist: "You put a gun to your head and pull the trigger," he says, "and four years later you find out if you blew your brains out."
AMD misfired more than its share during its early years as a microprocessor designer. Its first product, the K5, was an outright bomb when introduced in 1996. The next, the K6, was a near-miss that was faster than Intel's Pentium Pro but suffered from manufacturing glitches. AMD finally scored a huge hit in 1999 with the Athlon, the world's first PC processor to reach a clock speed of 1 gigahertz. On the strength of Athlon, AMD's stock price hit a high of $47.50 shortly after Ruiz joined the company.
When Intel began working on Itanium in 1994, however, AMD wasn't even on its radar screen. Having largely sewn up the PC processor market, Intel took aim at the lucrative market for servers that powered corporate networks and that would power the emerging Internet. Just as Intel chips had helped IBM-compatible PCs trounce Apple, Intel wanted to develop a "Sun-killer" processor that would dominate Sun Microsystems in the server business. A 64-bit chip seemed like just the ticket, since it would be able to process data in chunks twice as large as existing PC processors could. But there was one imponderable: In creating such a chip, was it wiser to build on existing x86 technology or take a leap into the unknown and create a whole new architecture? Despite the greater risk, Intel chose the latter. The decision wasn't driven entirely by intellectual curiosity. Since Itanium would have no real competition, customers would have to buy Intel whenever the inevitable transition to 64-bit computing took place. When Hewlett-Packard (HPQ) agreed to help Intel develop Itanium in 1994, heads swelled even larger. "If I were competitors, I'd be really worried," crowed Albert Yu, then Intel's general manager for microprocessors. "If you think you have a future, you don't."
But hubris had its comeuppance. During the next seven years, Intel became mired in what some consider the longest-running and most expensive computing project in history. As the years dragged on, Intel engineers became overwhelmed by the sheer complexity of the undertaking. When the first Itanium chip was shipped in 2001, it was two years late -- a lifetime in the semiconductor industry. By then the Itanium "superchip" ran software written for 32-bit processors -- meaning most software now in existence -- at a fifth the speed of Intel's newest Pentium 4 chip, which is based on the old x86 technology.
AMD watched with glee. While Intel was reinventing the wheel, AMD quietly decided in 1999 to go ahead with a 64-bit chip of its own. But in part because of greater budget constraints, the company placed the opposite bet -- that the trusty old x86 architecture is, in fact, up to the task. If AMD succeeds, the x86 heritage means that the new chip would easily be able to run existing software at high speeds -- something Itanium has proven it can't do.
Since Hammer isn't even out yet, it's a bit premature to declare victory. But the two longtime combatants have entered the final stage of a relationship that has evolved from business partners to arch competitors to a final duel: Which company will create the hardware platform for 64-bit computing? In the first round of this most dramatic, high-stakes game of Russian roulette, it appears that Intel has blown its brains out. Now it's AMD's turn to pull the trigger.
There couldn't be a worse time for AMD to attempt such a feat. While Intel can easily weather what Ruiz calls "the perfect storm" -- a global recession, low chip demand, and excess inventory walloping the semiconductor industry at the same time -- AMD is looking like a shack on the beach. Sales have dropped from $900 million in the first quarter to about $500 million in the third, and the stock is off more than 80 percent from its 2002 high in early January. A looming cash crunch could force AMD to renegotiate loans connected with its two-year-old chipmaking facility in Dresden, Germany, which is now gearing up to produce Hammer chips. At the rate the company is burning cash, its $1 billion cushion will be gone by the end of next year. Asked whether AMD can survive another year, Ruiz replies, "We can't if we don't do anything different."
Ruiz is determined not to repeat the manufacturing problems that have dogged the company in the past. To that end, he has decentralized the management structures and manufacturing processes so each of his business units -- desktop and server processors, flash memory, and chips for small, mobile devices -- can operate independently of one another. He has also decided to get help with some of the manufacturing from partners with strong track records. In January, AMD signed a joint-venture deal with United Microelectronics of Taiwan, the world's No. 2 contract manufacturer of semiconductors, to build a state-of-the-art facility in Singapore by 2005.
In the meantime, however, AMD will produce Hammer chips itself at its Dresden facility. The version designed for workstations and servers is called Opteron and is due out in the first half of next year. The PC version, sold under the Athlon name, was supposed to make its debut this quarter, but that has been delayed until early 2003. Some analysts have speculated that the company is having trouble with a new manufacturing technique called "silicon on insulator." But AMD says there's no cause for alarm. "Big, complicated engineering programs always slip a little," says Dirk Meyer, senior vice president and general manager of AMD's computation products group, who led the original Athlon project.
If AMD can ultimately deliver on the Hammer, it has one enormous advantage in the marketplace over Itanium -- its ability to natively handle 32-bit software and scale up to 64-bit programs as needed. Analysts say the implications are huge, potentially allowing companies to save billions in IT costs by upgrading at their own pace. Some key players seem to agree. In April, Microsoft (MSFT) announced that it will offer a 64-bit version of Windows to support the new platform -- and IBM (IBM) recently made the same endorsement for its DB2 database software. Even Dell (DELL), which earlier committed to shipping servers with Itanium chips, is now considering using Hammer. Intel, predictably, is not impressed. "Nobody yet has signed on to say 'This is what we'll run our mission-critical apps on," says spokesman Howard High. But no one has said that about Itanium either, and it has been on the market since May of last year. Companies like Google, a natural customer for the chip, have yet to buy it, and market researchers say Itanium would be lucky to crack 10 percent in the server-computing market by 2007. "Itanium is still in the early parts of its life," High responds. "People are still kicking the tires."
Rob Enderle, an analyst with the Giga Information Group, thinks Itanium's situation is more dire. "Intel's arrogance is coming back to haunt them," he says. "Customers don't view Intel as a company that listens to them. Every time Intel changes its technology, an IT department has to change its software, and they don't like that." Sanders, characteristically, puts it more colorfully: "Why would the customers want to scrap billions of dollars of software so that Intel could be the only provider of hardware. Hello?"
Intel tried to hedge its bet with a not-so-secret project, code-named Yamhill, that attempted to build a new 64-bit chip based on the old x86 technology. But it's doubtful that Intel would launch a chip that created yet more competition for itself. Instead the company has upped the ante with the Itanium platform. Three years ago it brought in a veteran Intel manager, Michael J. Fister, to save the bloated and delayed project. By next year, Fister says, a new Itanium chip, code-named Madison, will outperform all the existing 64-bit competition.
But Madison, unlike Hammer, won't be backward-compatible, and that crucial distinction could give AMD the biggest hit of its tumultuous history. Which is why Ruiz can afford to be so calm while the world seems to be shattering around him. "I'm morbid," he confesses as his shy engineer's mumble disappears into that booming laugh. Seems he may have a little bit of Jerry Sanders in him, after all. "I actually find all this kind of exciting."
From what AMD has shown us up to this point, if they can keep their head above water long enough to get the hammer established in the market, Intel is going to be in for some pain.
We are hoping that Hewlett-Packard who is a perfered supplier of PC servers will find AMD's religion, but given their investment in Itanium I am not optimistic.
The 8x88's instruction set was rather odd from the get-go. I've heard rumors that there was a COBOL compiler that actually used all the instructions, but I wouldn't know. Actually, I'm really quite amazed at how effectively today's processors can manage to run code which has two completely-independent sets of addressing modes, a segment architecture which sometimes uses segment*16+base, sometimes uses segment-descriptors, and sometimes uses neither, an instruction-chunk length which is probably about a quarter the length of the average instruction, and many other fun quirks. I never would have thought it possible that any company could pull such a thing off, much less that two could.
Closing sentence:
Maybe the world really wants a faster x86 instead of a new and different family of processors. Maybe lightning will strike twice.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.