Intel's boss was raised on the inside. Now he must turn it inside out.
Intel Corp. was mired in misery early last year. After two decades dominating the microchip market-- one of the fastest-moving and most unforgiving businesses in the world--
Moreover, forays for new growth were fizzling: A big move into new chips for cell phones flopped; a plan to create a business running server farms for corporate clients faded; a billion-dollar gamble on Itanium, a new-generation chip for big servers, failed to pay off. The efforts had been plotted by Craig Barrett, the materials-science engineer who in 1998 succeeded Andrew S. Grove, the salty leader who had helped start the company. Barrett was helped by a successor of his own: Paul S. Otellini, an Intel lifer who in mid-2005 became the first non-Ph.D. to run Intel.
Unaccustomed to losing, Intel's ranks pelted their new chief with bitter e-mails: Intel had lost it; management was incompetent. Some likened Intel to a lumbering Detroit carmaker. Today Otellini, 56, is reluctant to talk about the backlash, though his chagrin is apparent. His de facto number two, Sean Maloney, is more blunt: "It was a swift kick in the gut. We were angry and disappointed in ourselves." He adds: "We just had a visceral emotion: We're gonna fix it."
That frustration pushed Otellini to wage the most sweeping overhaul at Intel in 20 years. Gone are plans for diversifying away from Intel's chips. Gone is 10% of its workforce. Otellini also bailed Intel out of cell phones, selling off the XScale mobile-chip line. In a first for the company, he has put one chip factory--so far--up for sale. And Otellini is reorienting Intel's focus to look beyond its slow-growth mainstay, processor chips for desktops, to what he hopes will be the Next Wave. This is a world of lightweight notebook PCs and a gaggle of ultramobile machines smaller than a laptop but bigger than a BlackBerry. "The PC market has been very good to us. It's near 300 million units [a year]. It's going to grow to 500 million units," he says. "But how do we sell a billion of something? Can we create a multihundred-million-unit market, per year, in handhelds?"
Intel already had in place a growth engine that could fuel its comeback: Centrino, simpler, faster and cheaper than the Pentium 4. The design began with Intel engineers in Haifa, Israel, far from Silicon Valley. Centrino would power notebook computers--but the "core" processor at Centrino's heart would become the inspiration for all of Intel and lead to a starkly different design than the Pentium, which had reigned since 1993.
In Silicon Valley Intel engineers thrived on using every available transistor to get more speed and power from a chip. The adverse side effect: lots of waste heat. The 178 million transistors in Intel's top Pentium 4 give off enough heat to fry an egg.
In the new-gadget era envisioned by Otellini the chips must be the antithesis of a hefty Pentium--sleeker, simpler, far cheaper and, above all, cooler. Lash a couple of processor cores together and bundle in specialized parts for, say, wireless linkups or video graphics, and this system can power everything from a palmtop to a server. Intel says a new family of cores, aimed at mobile devices, will be ready next year.
"How do we fit inside of something that sells for $100 and make some money?" Otellini says. "Costs become essential. Architecture becomes essential. Integration becomes essential. And the culture of the company has to wrap itself around that." This threatens "the ego of the Intel engineering community," says Maloney, executive vice president. "Their whole notion of self-worth was based around bigger and faster. That aspiration needed to change to cooler, sleeker, smaller. That's a big deal."
Thus Intel has abandoned what may be the most prodigious platform--the cell phone, with 1 billion units sold last year, four times the number of PCs--in favor of a new gadget that barely exists. Succeeding requires Intel to do two things it never has done particularly well: make chips at the lowest cost possible and let customers' demands shape development.
The Intel of old held 85% of the PC microprocessor market and routinely dictated upgrades and designs with little input from the clientele. AMD Chief Hector Ruiz tacitly goads Intel for this: "We did something that, unfortunately, is all too rare in the semiconductor industry--we went out and talked with [customers] about what [they] needed," he said in an industry speech in October 2006.
The new Intel must undergo a personality transplant. The Intel that Andy Grove built had enshrined sharp confrontation as constructive engagement, in the imperious and emphatic style of its chairman, for whom decisions were crisp, choices were binary and markets were won or lost.
Otellini, who scoffs privately at the "cult" that can surround a company's founders, can deliberate something to death. He deploys a reserved manner and prefers persuasion over fiats, consensus over combat. Frustration or embarrassment shows in a red flush to his face. His equanimity is a mixed blessing. It can be seen as indecisiveness.
He was born and bred in San Francisco, and during his college years he spent a summer working with his father, a butcher, in a slaughterhouse. ("I think he did that on purpose, because he didn't want me to ever think of that as a career," Otellini has said.) He attended the University of San Francisco, and in 1974 he landed his M.B.A. at the University of California, Berkeley, joining Intel as an analyst. He hasn't missed an Intel paycheck since. He rose in marketing and management--"I'm a product guy"--and spent a year in 1990 as an aide to Grove.
Barrett succeeded Grove in 1998 and began looking beyond microprocessors, a business he derided as a "creosote bush." (In the desert a creosote bush poisons the ground around it to ward off other vegetation.) He had Intel spend $10 billion buying communications and networking firms, even as it invested hundreds of millions more in the Itanium chip project with HP.
By 2002 the dot-com crash and the collapse of telecom had devastated Intel's profits and chilled Barrett's plans. Intel's move into chips for mobile phones had become a quagmire. Although the company had grabbed a promising chip line in a legal settlement with the old Digital Equipment in 1998 and renamed it XScale, the chip wasn't enough. Unlike the PC world, software for such chips was patchy. Even making the chips proved more costly than expected as Intel had to rejigger manufacturing processes.
"In hindsight, phones--even the smart phones we targeted--was not an area in which we had 20 or 30 years of expertise," says Otellini, who became president in early 2002. "It didn't play to any of our strengths. We didn't have the software or the architecture." Nor did Intel have many customers.
Meanwhile Intel was getting into trouble in microprocessors. The Itanium, in gestation since 1994 and a few years behind schedule, faltered when customers balked at the hassle and the cost of rewriting old Intel-based software for the new chip. Worse, the Pentium 4 was a hothead and a power guzzler, at a time when corporate customers eyed even electricity bills in a bid to reduce their tech spending.
AMD, Intel's plucky rival, was poised to benefit. Its engineers had been working on a homegrown chip that rivaled Intel's high-end Itanium for power but easily ran existing software. And it was cool--generating less heat than Intel's big chips. AMD debuted its Opteron for high-end servers in April 2003 and rocked Intel's world. The competition would knock the average selling price for high-end chips from more than $600 apiece in 2003 to half of that today, says IDC analyst Shane Rau.
Intel, meanwhile, had glitches. It canceled one new version of Pentium 4, ran a year late on another, delayed several other products and ran short of chips because of bad forecasting. Only the transition of the chief executive job from Barrett to Otellini, in May 2005, went smoothly. Intel stock rose 8% in calendar 2005; AMD's rose 43%.
Then Intel stumbled in a spectacular way: It missed sales forecasts on Wall Street two quarters in a row, through the first quarter of 2006. And Intel was bloated. In 2000 it had 86,000 people producing $34 billion of revenue; by 2006 it had added 17,000, though the top line had grown only 5%.
Otellini spent much of last year handling the fallout--disillusioned employees, the board demanding to know why Intel had slipped so badly, a huge round of layoffs. Yet Intel already had a key element in place for a dramatic comeback--the processor core inside Centrino.
In 2000 Otellini, then head of Intel's microprocessor business, had realized slim notebooks would need a cooler, less power-hungry processor than the Pentium 4. So he set engineers in Israel to the task. They approached it in a non-Intel way, sacrificing some raw power to get a chip that ran cooler. The idea was scorned inside Intel. "The company had been so successful in the 1990s it was hard to talk about doing things differently," says David Perlmutter, who led the project. "It was easier to be remote and question the basic religion of the company."
In 2002 their work was all but finished, when Otellini had an epiphany: Notebooks and laptops had to be able to connect wirelessly to the Internet. So Otellini decreed that the new chip should wait until the engineers could fuse their core to a homegrown Wi-Fi component. "Making that decision was tumultuous inside of Intel, to say the least," he says. "It was a cultural issue. We're a microprocessor company." The Intel faithful disliked delaying a new chip to wait for adjunct technology. One computer maker jeered at the project, calling it "Latrino."
Intel rolled out its Wi-Fi-ready Centrino in March 2003. Six months later the new chip was a much-needed hit. Intel's Perlmutter was convinced he could see the next horizon. "The first Centrino wasn't bad," he says. "But could we evolve the architecture to be better than the Pentium 4?"
It could. A Centrino-like core was anointed as Intel's flagship for notebook and desktop PCs. In October 2004 Otellini canceled future Pentium 4 efforts. He signed on for a big test of whether Intel's engineers could shed their dictatorial ways to work closely with a most demanding customer: Steve Jobs of
Subscribe to Forbes and Save. Click Here.
After much wooing by Otellini, Jobs had agreed to consider using the next core, so long as Apple engineers could work hand in hand with Intel's designers. And in mid-2005 Jobs took the stage at his annual powwow with developers to announce that the Apple Macintosh would start using Intel chips. Since the Mac's debut in 1984, it always had run on chips from
In spite of Centrino's success, Intel overall was sagging. Early last year Otellini hired consultants from Bain & Co., who huddled with some of Intel's smartest managers to take a snapshot of the company. Their report, two months later, stung Otellini and staff: Intel was fat and inefficient, hampered by high costs and a swollen marketing department. One example: Intel had too many "two-in-a-box" managers (a pair who share title and job). One set even jointly oversaw a staff of one.
"We weren't used to the sniping. No one had questioned us for years," says marketing chief Sean Maloney. But instead of demurring, senior execs focused on a fix. "How could you let Intel fail?" he says. "We're the company that's famous for technological change. The sense of personal shame would be overwhelming."
But Intel stumbled again in launching its rescue mission. Led by the deliberative Otellini, the company imposed job cuts so slowly that the ranks grew ever more angry. Intel first told Wall Street it would analyze the company's structure but didn't cite layoffs. Later it said it would fire a thousand managers. In June 2006 it sold off the xscale mobile chip line to
Now Otellini plies new growth. Intel's sales of various "core" chips (including the Centrino and the "Core 2 Duo" lines, both inspired by that original core approach in Centrino) will exceed sales of its classic design this year. Notebook chips are gaining fast: By 2009 Intel figures it will ship more chips for notebooks than for desktop PCs--happy news because at least for now Intel makes more money on notebooks.
Next: chips for the ultramobile handhelds. These will incorporate, on one piece of silicon, a Centrino-like core plus circuits handling such tasks as Voice over Internet, graphics for games and search. Otellini argues that by 2011 such chips could compete in what he expects will be three newly formed $10-billion-a-year markets--one each in mobile, consumer electronics and supercheap PCs for the Third World. That enhanced core goes for now by the name Silverthorne.
"Silverthorne could really be a thorn in Intel's side," frets Auguste Richard, a senior analyst at First Albany Capital in San Francisco, who nonetheless admires the Intel overhaul and has a "buy" on the stock. Any system-on-a-chip revenues for Intel are a few years away. Wall Street also worries about the inevitably thinner profit margins in chips for cheap palmtops.
Intel never had obsessed over cutting product costs. "You would never have had a discussion with Andy [Grove] or Craig [Barrett] about us being the lowest-cost producer," Otellini says. When a company has products that could command as much as 80% margins, he notes, "costs are important but not critical." But cost will be everything in the handheld market, and Intel is counting on its mind-boggling prowess in manufacturing for an edge.
It makes some of the tiniest chips ever created for a PC, cramming them onto the largest silicon wafers in the world. It now has 5 factories (of a total 16) that use platters 300 millimeters across (12 inches or so). By year-end Intel will produce chips with transistors measuring 45 nanometers (45 billionths of a meter), smaller than most human viruses. That will let it etch 2,500 chips on each 300mm wafer.
Otellini bets Intel can stay so far ahead of rivals that it can make chips as cheaply as any competitor. That includes China, which is emerging as the foundry for the rest of the chip world. "We've been benchmarking them. We don't think we're at a cost disadvantage," he says. In March Intel set plans to build its next chip factory, typically a $4 billion project in the U.S., in Dalian, China.
But Intel will have to prevail over fearsome foes: Samsung and
"Intel talks about being customer-centric, but it's not in their DNA. They've been brought up to rule the world," says Henri Richard, AMD sales chief. When he meets with phonemakers, they tell him what they want, what a chip should do and how much it should cost. In PCs, "Intel tells the customer: 'This is the way it's going to be.'"
At TI, Senior Vice President R. Gregory Delagi says his company has spent a decade learning to coddle clients. It reorganized its supply chain to have TI products ready just across the street from a customer assembly plant. During the tech slump in 2001 TI built a site for a customer within its own factory in France to let the client's engineers work alongside TI staff.
But Intel execs say they learned how to mollify customers as fussy as Steve Jobs. And they vow that their foray into the ultra- mobile market will fare better than their effort to make chips for cell phones, in part because Intel's cores are heirs to the mountains of software written for the 850 million or so Intel-based computers in the world today.
"That's the heart of why it was important for us to make the changes we made last year," Otellini says. A billion computers now link up to the Internet, most of them Intel-based, and it took 12 years to reach that milestone. The next billion machines will come online in only half as much time--and Intel will have to fight hard for every single one of them.
http://members.forbes.com/forbes/2007/0604/092.html
Sphere: Related Content
No comments:
Post a Comment