Has Barrett done a good or bad job at Intel?
01/30 03:02
Intel's Barrett Bets $28 Bln on Chips as Internet Push Flops
By Anthony Effinger
Santa Clara, California, Jan. 30 (Bloomberg) -- Everyone in Silicon Valley got a little crazy during the Internet boom, even Intel Corp., the world's biggest maker of computer chips.
In 1999, Chief Executive Craig Barrett spent more than $6 billion buying companies like IPivot Inc., maker of pizza-box-size devices that sped up the display of Web pages.
He ripped the roof off a building outside his Santa Clara, California, headquarters, installed computers and ran Internet sites for Citigroup Inc. and NEC Corp. He made children's microscopes and digital music recorders.
Intel shares surged to a record $75.81 on Aug. 28, 2000. Yesterday, they closed at $16.66, a 78 percent drop to about where they were when Barrett began his binge. Last year alone, the stock lost half its value.
Now Barrett is making another bet. He plans to spend as much as $8 billion this year on top of $20 billion in 2001 and 2002 to take Intel back to its chipmaking roots.
He's gambling that research and development, new plants and powerful semiconductors will boost sales when demand for personal computers, cell phones and wireless devices picks up.
``Intel will start to print money,'' says John Lau, an analyst at RBC Capital Markets, who expects per-share earnings of 63 cents this year and 84 cents in 2004.
Barrett has two years to persuade investors he can get Intel growing again. He's 63, and Intel mandates retirement at 65. Some shareholders are irked he wasted time and money on investments that didn't pan out.
``The numbers were ridiculous,'' says Christian Koch, senior technology analyst at Trusco Capital Management in Atlanta. Trusco owns 6 million Intel shares and is buying more, speculating that Barrett is on the right track this time, Koch says.
Internet Binge
When Barrett embarked on his Internet bender, his goal was to move beyond microprocessors, the square bits of silicon that provided four-fifths of Intel's sales and almost all of its profit during the past decade. Computer sales had started to slow, rising only 2.4 percent in 1998, according to researcher IDC. Intel looked like it might get stuck inside the PC while Cisco Systems Inc. made billions of dollars wiring the world.
``You don't want to be known as the guy who presided over Intel when it had a slow-growth period,'' Barrett said in a February 2000 interview with Bloomberg News.
Then the boom ended, and the world's biggest economies slid into recession. Intel's sales totaled $26.8 billion in 2002, slightly higher than $26.3 billion in 1998, the year Barrett took over as CEO from Andy Grove. Net income was $3.1 billion, or 46 cents a share -- 30 percent of the $10.5 billion, or $1.51 a share, Intel earned in 2000.
New Plants
None of Barrett's acquisitions picked up the slack: The units that make communications chips, bulked up by his deals, had a combined operating loss of $916 million. And Intel Capital, an in- house venture capital fund, saw assets shrink to $1.7 billion at the end of 2001 from $8 billion in 1999.
Barrett's new spending spree is in full swing. Since 2000, he's built two new plants in Oregon and one in New Mexico and is working on one in Ireland to churn out processors that are smaller, cheaper and faster than those from Advanced Micro Devices Inc.
He wants to horn in on International Business Machines Corp. and Sun Microsystems Inc. with powerful processors for servers, the machines that control computer networks. He's targeting Texas Instruments Inc. with chips for cell phones, organizers and networking gear.
Barrett is placing one of his biggest bets on Wi-Fi, or Wireless Fidelity, which beams files, Web pages and e-mail for short distances over airwaves in coffee shops and airport lounges.
Shrinking Chips
Intel has a new processor for laptop computers that's code- named Banias, for a river north of Haifa, Israel, the city where it was developed. In March, the company plans to start selling a three-piece system called Centrino that contains Banias and several chips for Wi-Fi. Intel plans to spend more than $300 million marketing Centrino, the most in the company's history.
Eventually, Barrett wants to move all functions needed for Wi- Fi onto one chip -- or two at the most -- and manufacture them in ever-shrinking sizes that lead to lower costs, a tricky feat no one has perfected.
Lau says Barrett is wise to pursue Wi-Fi. He predicts Intel will snatch sales from Agere Systems Inc. and Intersil Corp., which sell the Wi-Fi chips that Dell Computer Corp. and Hewlett- Packard Co. put into their notebook computers. The array of chips needed for Wi-Fi costs $20 to $30, Lau says, revenue Intel isn't getting now.
`Savvy'
``This will take them three to four years,'' says Tony Grewe, director of strategic marketing at Agere, which has been making Wi- Fi chips for a decade.
Wi-Fi may also wring more profit from Intel's Pentium line. A 2.2-gigahertz Pentium 4 laptop processor listed for $348 in January. A comparable chip for desktop PCs was $193. ``This strategy is pretty savvy,'' says Trusco's Koch.
Sean Maloney, executive vice president of Intel's communications group, has even higher hopes: He predicts Wi-Fi will create millions of new laptop users, thereby fueling demand the way the Internet did for desktop PCs.
``Every notebook is going to end up on Wi-Fi,'' says Maloney, a 46-year-old London native who looks like actor Anthony Hopkins. ``This is unquestionably enthroned as the next big thing.''
Investors say they've heard such hype before: with the Internet. Companies like Theglobe.com Inc. built businesses around the Net and then wound up trading for pennies.
Expensive Stock
While no one expects Intel to crash and burn, many analysts and investors say its stock is expensive given floundering PC sales, tough times in the communications industry and the unproven nature of Wi-Fi.
Intel trades at 27 times the average earnings estimate for 2003, compared with a price-earnings ratio of 17 for the Standard & Poor's 500 Index. Fellow technology bellwether Microsoft Corp. trades at a multiple of 25; IBM hovers at 19.
``What are people going to pay for a cyclical tech company?'' asks Doug Johanson, head of Vista Capital Partners Inc. in Beaverton, Oregon, who's avoiding Intel shares. ``Not a lot.''
Some investors are casting their lot with Barrett. They don't want to miss a rally should PC sales pick up or initiatives like Wi-Fi pan out.
A few signs point to a revival: Worldwide semiconductor sales jumped 20 percent to $12.7 billion in November, the Semiconductor Industry Association says. Sales are likely to rise a further 12 percent in 2003, according to market researcher Dataquest Inc.
Against the Herd
``We're out of the tech recession,'' says Jerry Dodson, president of Parnassus Investments, which manages $700 million in San Francisco. He bought 1.5 million Intel shares at about $16 in October. ``We're optimistic,'' he says.
Intel has a history of going against the herd -- and succeeding. Barrett says the current slump is nothing like the 1980s, when he was general manager of manufacturing and Intel was a memory-chip maker. Cheap imports forced engineer founders Gordon Moore, Robert Noyce and Grove to abandon the memory market and bet on microprocessors, then an unproven technology.
Moore gambled again when he spent millions of dollars to develop the 386 processor, even though analyses showed Intel would never recoup the cost, says Andy Bryant, who crunched the numbers and is now chief financial officer.
Intel's existing 286 chip was so powerful, Bryant reasoned at the time, why would computer users need anything better? Moore overruled him, and Intel produced the 386, Bryant says. The chip succeeded, eventually spawning the current generation of Pentiums.
Moore's Influence
Intel has been just as aggressive in dealing with recessions, thanks in large part to Moore. Moore's strategy for hard times was based on three beliefs, says Intel spokesman Tom Beermann: Recessions end. You can't stop them by saving money. And some companies emerge better off than competitors.
In his better-known observation, Moore posited that computer chips would double in power every two years or so as engineers figured out how to pack more and smaller transistors onto pieces of silicon. The theory, which has proved to be on target, is known as Moore's law.
Moore, 74, retired as chairman in 1987. In 2000, he and his wife set up the Gordon and Betty Moore Foundation to support education, science and the environment.
Sticking with Moore's philosophy, Barrett says the biggest risk for Intel now lies in investing too little, not too much.
Betting on a Recovery
Intel made that mistake in 1998, when it canceled the construction of a processor plant in Fort Worth, Texas, for which it had broken ground in 1997. When demand picked up in late 1998, Intel couldn't deliver to small customers, like Boldata Systems, which accounted for about a third of PCs shipped worldwide that year. Advanced Micro stole sales.
``We're betting there will be a recovery,'' says the 6-foot-3- inch Barrett after climbing the stairs two at a time to a fifth- floor conference room near his cubicle. He's taken the elevator three times since moving in 10 years ago: twice with guests -- one a Thai princess -- and once when he had the flu.
Barrett's record investments during the chip slump are testing Intel's own stamina. In 2001 alone, he shelled out $7.3 billion for new fabs -- short for fabrication facilities.
It was the most ever for Intel and happened in a year in which sales fell 21 percent to $26.5 billion and earnings tumbled 88 percent to $1.3 billion. Profit margin, or net income as a percentage of sales, shrank to 4.9 percent from 31 percent.
Barrett's bet-the-ranch ambition is clear at the Ronler Acres complex in Hillsboro, Oregon, 13 miles west of Portland. Intel has turned this farming community into a 21st-century mill town.
1,000 Degrees
Ronler has three plants, parking lots and offices covering 300 acres amid subdivisions and malls. Four white cooling pipes snake up the side of one factory, expelling steam from manufacturing processes that run at 1,000 degrees Fahrenheit.
Intel is doing at Ronler what it does best: milking Moore's law. So far, so good. The best Pentium 4 processor has 55 million transistors and runs at 3 gigahertz, meaning electrons race through the chip's microscopic wires and turn the transistors on and off 3 billion times a second. Pentium IIIs from two years ago had 28 million transistors and ran at 1 gigahertz.
Sustaining Moore's law isn't easy -- or cheap. Intel and its rivals make semiconductors by imprinting patterns of circuits on silicon dioxide with ultraviolet light and then removing the bits that aren't exposed. Soon, the circuits will have to be smaller than the wavelength of the light that creates them.
``It's like painting thin lines with a big brush,'' says Sunlin Chou, senior vice president of technology and manufacturing.
$2.5 Billion Plant
The solution is to use smaller wavelengths. The catch is that the higher-intensity light burns the quartz lenses that focus the light on the silicon. Chou's staff is working on mirrors that will reflect light onto silicon instead. The mirrors need an 80-layer coating of materials to withstand the heat. Such solutions will cost millions of dollars.
``We're one of the few companies around that can continue to invest in state-of-the-art manufacturing,'' says Barrett, who spent $4.7 billion on new plants in 2002 and plans to spend as much as $3.9 billion this year. Advanced Micro, by contrast, spent $750 million last year.
Fab D1D, the newest at Ronler, will have cost Intel about $2.5 billion once it's completed this year. Ronler's 5,000 workers will be able to walk through three-quarters of a mile of continuous clean rooms, in which there are only 10 dust particles per square foot.
Monorails will run along the ceiling, delivering beer-cooler- size bins of silicon wafers to million-dollar electron microscopes that workers will use to peer at the maze of wires inside each chip to spot defects.
`Not Justified'
Ashok Kumar, an analyst at U.S. Bancorp Piper Jaffray, says Barrett's spending spree is puzzling. Midrange PCs sell for about $700, and chips for those machines go for less than $200. ``This kind of investment isn't justified if the sweet spot of the market is way below $1,000,'' Kumar says.
Barrett says faster chips will popularize endeavors like digital video editing, which requires lots of computer power. Servers need more juice too, if Intel-based machines made by Hewlett-Packard and Dell are to compete with IBM and Sun, which make their own chips and still rule in sales of million-dollar computers.
Intel hopes to break into the high end with Itanium, a processor it's been working on with Hewlett-Packard since 1994. Like Sun and IBM chips, Itanium processes information in 64-bit chunks compared with 32 bits for Pentiums. Itaniums sell for as much as $4,226 compared with $637 for the best Pentium 4s.
So far, Itanium -- named to invoke the strength of titanium - - has been a bust. The first Itanium-based server, unveiled in May 2001, didn't perform well enough to persuade many customers to switch from tried-and-true equipment from Sun.
`Itanic'
Manufacturers shipped 760 Itanium servers in 2001 and an estimated 3,000 in 2002, according to IDC. Sun CEO Scott McNealy calls the chip ``Itanic,'' meaning that it will sink.
Itanium aside, bulls argue that once computer sales pick up, Intel's earnings will swell because it produces so efficiently.
Right now, Intel cuts most of its chips from disks that are 8 inches in diameter. Two plants -- one in Rio Rancho, New Mexico, and another at Ronler -- use 12-inch wafers, which have twice as much surface on which to layer transistors and the copper wiring that connects them.
Chips cost 30 percent less to make at those two plants as a result, spokesman Beermann says. D1D at Ronler Acres will begin production on the bigger wafers this year.
Wi-Fi Gamble
Barrett is betting that Wi-Fi laptops will help create demand for his millions of new chips. Intel started pursuing Wi-Fi in early 2002. Grove was frustrated by how few homes had high-speed Internet connections. He assembled a team to study the technology as an alternative to cable hookups and digital subscriber lines, Beermann says.
Maloney, who took over Intel's communications chip businesses in April 2001, became interested in Wi-Fi about the same time. He had joined Intel as an engineer in 1982, working on a technology called Ethernet that connects computers into networks.
Top executives at Intel have technical assistants -- aides that do everything from opening mail to leading pet projects. Maloney was Grove's aide from 1992 to 1994.
Talking with Maloney is like stepping back to 1999, except it's not the Internet that's going to change the world; it's Wi- Fi, technically known as 802.11. The number was assigned by the Institute of Electrical and Electronics Engineers, which brings together industry groups to define standards. The 802 part designates technologies that connect computers in places like offices. The 11 specifies wireless networks.
Brave New World
Maloney envisions a day when most cities are wired for Wi-Fi. He predicts that banks will have cameras that will broadcast pictures of robbers to police right after a crime and that parents will be able to send photos of missing children throughout the community.
Intel is pushing Wi-Fi's limits in Oregon, where antennas beam signals to employees living up to 20 miles away -- rather than the 300-foot radius of a typical Wi-Fi system. The system works well when there are no trees or hills in the way, which lends support to Grove's vision of Wi-Fi as an alternative to DSL or cable.
Intel's big stumbling block involves getting all of the Wi-Fi circuitry onto the Banias chip -- or onto Banias and one other chip -- and then mass-producing them efficiently. Transistors that process radio signals such as Wi-Fi are larger and tougher to shrink than the circuits that make up microprocessors, making the benefits of Moore's law harder to capture.
Overrated
``It's a very, very difficult problem,'' says Andreas Stavropoulos, a director at San Francisco-based venture firm Draper Fisher Jurvetson, which made its first Wi-Fi investment in January.
Consolidation of functions will cut costs by eliminating the need to manufacture extra chips. It may also block competition from Wi-Fi-only companies like Intersil that sell chips that work alongside the main processors.
Even if Intel clears the technical hurdles, Barrett may be expecting too much from Wi-Fi. Andrew Seybold, a wireless industry consultant in Boulder Creek, California, says Wi-Fi is overrated. He predicts it will appeal to a small number of businesspeople who want a better way to check e-mail in airports between flights or at Starbucks Corp. shops after sales calls.
The Seattle coffee chain is wiring its stores with Wi-Fi through a partnership with Deutsche Telekom AG's T-Mobile International AG unit and Hewlett-Packard. ``People are looking for the next big thing and are rushing in without really thinking about it,'' Seybold says.
Detail Guy
Barrett's training puts him among the last people investors expect will rush in blindly. A former professor of engineering at Stanford University and operating chief under Grove from 1993 to 1997, Barrett says he likes details and has a penchant for setting goals and meeting them.
Away from Intel's headquarters in Silicon Valley and his home in Paradise Valley near Phoenix, Barrett tries to catch at least one fish every month on his 400-acre ranch outside Darby, Montana. In November, he had to cast beyond a shelf of ice with his fly rod to reach open water. He caught two trout.
The ranch provides another diversion -- a chance to get away on his big horse, a roan-and-white paint called Nasdaq that stands 16 1/2 hands (5 1/2 feet to his shoulder). ``I named him for what paid for him,'' Barrett says.
Barrett invented Intel's copy-exactly manufacturing strategy under which all 13 plants use the same machines and chemicals. Even the Gore-Tex bunny suits workers don in Intel's fabs are identical from Oregon to Ireland.
``Under Andy, I was the detail, inside guy,'' says Barrett. ``As an engineer, you feel comfortable when you have the data. With the world economy, it's a little more difficult.''
Internet Bug
CFO Bryant says Barrett caught the Internet bug because he has absolute faith that technology -- no matter how untested -- can create useful things.
The optimism prompted Barrett to embark on businesses like Intel Internet Media Services, which broadcasts shareholder meetings over the Web for other companies. He planned to spend $200 million on a network to carry the events.
Barrett's splurge started in October 1998, when he agreed to buy networking equipment maker Shiva Corp. for $185 million. A month later, he paid an undisclosed price for ICat Corp., a closely held company that helped businesses set up shop on the Web. He grabbed communications-chip maker Level One Communications Inc. in August 1999 for $2.73 billion.
28 Acquisitions
About the same time, he paid $780 million for Dialogic Corp., whose equipment sent voice signals over computer networks. In September 1999, he agreed to buy closely held networking gear maker NetBoost for an undisclosed price. He shelled out $1.6 billion for DSP Communications Inc. that October.
In all, Barrett bought 28 companies for $9 billion from 1998 through 2000. Investors lauded his strategy, driving Intel shares up 39 percent in 1999. Sales rose to a then record $29 billion, and Intel made a profit of almost $2 billion a quarter. Gross margin -- the percentage of sales left after production costs -- rose to 60 percent from 54 percent in 1998.
Barrett says he saw trouble coming on March 2, 2000. Shares of handheld organizer maker Palm Inc. rocketed as high as $165 the day after its initial public offering, giving the startup a market value of $53 billion -- higher than Ford Motor Co.'s. ``I knew that the situation was unstable,'' says Barrett.
The Internet tide crested a week later, when the Nasdaq Composite Index peaked at 5,132.52. It ended for Intel on Sept. 22, when the company said third-quarter sales would lag forecasts because of weaker demand for PCs in Europe.
Shares Tumbled
Intel shares tumbled 22 percent, their biggest decline ever, erasing $91 billion of market value that day. By the end of 2000, they were trading at $30.06, down 60 percent from their August record.
With the stock market in decline, venture capitalists became reluctant to fund startups. And larger companies reduced spending on Internet technology as they tempered growth projections.
The cutbacks created tension between Intel and its chip customers, Barrett says. Dell, Hewlett-Packard and IBM groused because Barrett was selling machines he got from companies like IPivot that competed with their products. Cisco didn't want to buy chips from a company that was making networking gear, Barrett says.
Shrinking Pie
When sales are growing by leaps and bounds, companies don't care if you get into their markets, Barrett says. ``When the pie starts to shrink, you covet other people's businesses,'' he says.
Eager to keep chip buyers happy, and disillusioned with many of the new businesses, Barrett began cutting his losses. In February 2001, he shuttered ICat. A week later, he closed Intel Internet Media Services.
In June 2002, he said he wanted out of Intel Online Services. Savvis Communications Corp., based in Herndon, Virginia, plans to take over the leases on Intel Online's four computer centers by May, including the one outside Barrett's window in Santa Clara.
By then, Barrett will be mostly done cleaning house. He says he doesn't regret the buying binge, though he wishes he hadn't paid top dollar. Fifty to 70 percent of the purchases were useful, he says.
Ron Smith, senior vice president of Intel's Wireless Communications and Computing Group, says engineers from DSP Communications helped him develop a chip code named Manitoba that will deliver voice and digital information, like e-mail, to mobile phones. It's scheduled to ship this year.
``Would I be happier if we could have bought low? Sure,'' Barrett says. ``That's hindsight.''
Otellini's Role
With his back-to-chips strategy under way, Barrett is turning over more daily duties to Paul Otellini, 52, who became president and chief operating officer in January 2002. Barrett is taking over some responsibilities from Grove, 66, who remains chairman and can serve until he's 72.
``We're transitioning,'' says Barrett, who no longer meets with Grove each week to go through to-do lists like they did for 10 years. Grove still sets the agenda for board meetings and chimes in on strategic decisions. He declined to comment for this article.
If Otellini succeeds Barrett, he'll be the first non-engineer to run Intel. He got an economics degree from the University of San Francisco in 1972 and an MBA from the University of California, Berkeley, in 1974.
Even with his business background, silicon is in his blood. A 28-year Intel veteran, Otellini headed the microprocessor unit for four years starting in 1998.
``The name to bet on is Otellini,'' says Kumar at Piper Jaffray.
Cometa Networks
Barrett is giving his Wi-Fi effort a jump-start. In December, Intel invested an undisclosed amount with AT&T Corp. and IBM to form Cometa Networks, which intends to build a U.S.-wide Wi-Fi network.
Cometa plans to install transmitters and then sell Wi-Fi service to wireless phone operators and Internet providers. They in turn can offer it to their customers. Cometa plans to start service in the 50 largest U.S. cities this year.
Les Vadasz, president of Intel Capital, which made the Cometa investment, likes Wi-Fi because it's growing on its own, without lots of promotion by the industry. ``That's what happened with the PC,'' says Vadasz.
Barrett would like nothing more than to see an Internet revolution go mainstream, this time a wireless one. Without something to spur PC sales, he faces his nightmare of running Intel during slow growth. And investors would face more years of recovering from Barrett's second bet -- at a cost of $28 billion.