By Olga Kharif Like most chip-industry executives, Dick Chang, general manager for semiconductors at Agilent Technologies, isn't particularly looking forward to next year. After the painful of 2001 and early 2002, demand for his chips -- including "filters" that reduce noise in wireless phones -- has inched up by 5% to 6% for a couple of quarters. Chang expects 2003 to be no worse, but not much better, either.
Compared with the industry's glory days, that's quite a comedown. In 2000, Agilent's (A) chip sales -- its No. 2 business after testing and measuring equipment -- reached $2.2 billion, up 28% from 1999. In the fiscal year ended Oct. 31, its chip sales were a shadow of that -- $1.56 billion. Agilent can pin a lot of the blame on its troubled telecom customers. Yet chipmakers whose fortunes are tied to other industries are grumbling about the slow recovery, too. Worldwide, semiconductor sales are hovering at 1995 levels. They should jump 9.2%, to $149 billion, in 2003, and another 15.2% in 2004, to $172 billion, according to tech consultancy IDC. That compares, however, to 36% growth in 2000, when worldwide chip sales reached $204 billion.
WHOLE NEW WORLD. There's an emerging consensus that relatively tepid times are here to stay in chips. The 16% compound annual growth the industry enjoyed between 1976 and 2001 could fall by half over the next two years, according to the Semiconductor Industry Assn. (SIA) -- whose forecasts tend to be optimistic. The PC and cell-phone markets, which account for nearly half of chipmakers' sales, have plateaued. And even deep discounting by the world's largest chipmaker, Intel (INTC) -- which, in November, cut the price of its 1.4 Gigahertz Pentium 4 processor by 27% -- will likely fail to fire up demand. Consumers may be shopping for new computers, but corporations aren't. That, says Chang, means "there are no drivers" for growth.
One minor and ironic blessing of the dot-com bust is that the chip industry's cyclical swings should become shallower, says Ray Burgess, director of strategy for semiconductors at Motorola (MOT). Although the ramp-up in demand that he expects to start in the second half of 2003 will be more modest than the sky-high growth of 2000, it could last three months to six months longer than the semiconductor industry's normal up cycle of 11 to 12 quarters. Moreover, the recovery should be a prosperous one for the strongest players -- thanks to the wounds suffered by weaker rivals during the downturn.
The numbers illustrate this bifurcation. Of the world's 22 largest pure-play chip companies, 17 had operating income in the third quarter that fell 9% short of costs, estimates John Barton, an analyst with Wachovia Securities. With slower growth "we can no longer count on the proverbial rising tide that lifts all boats," Jerry Sanders III, chairman of No. 2 microprocessor producer Advanced Micro Devices (AMD), told a recent industry gathering. Consequently, many chipmakers have embarked on a fresh round of cost-cutting -- eliminating capacity and jobs as they slash research and development -- that could forever separate them from the strongest players.
MORE PRODUCT, LOWER COST. The problem for such outfits -- among them AMD, the industry's No. 2 -- is that the leading chipmakers continue to plunge ahead as though the chip market hadn't tanked. In October, for instance, No. 1 microprocessor maker Intel opened the first mass-production plant for its 300-mm wafer in Rio Rancho, N.M. By stamping more microprocessors on larger wafers -- the silicon sheets from which they're produced -- Rio Rancho will pump out 2.5 times as many chips as Intel's older plants -- at up to 30% lower cost. Intel's main rival, AMD, won't have this technology until 2005, though an AMD spokesman argues, "We're in a good competitive position because...[our chips] deliver equal to better performance."
Translation: From a cost and price standpoint, Intel could make life hell for AMD. Since Intel's costs are lower, it can cut prices and still make money at levels that less-efficient AMD can't match without slipping into the red. AMD has lost money for the past year, and on Nov. 14, announced that it will lay off 2,000 more people, or 15% of its workforce, by mid-2003. In the third quarter, AMD's microprocessor share -- based on the percentage of PCs that use its chips -- dipped 1.1 points, to 16.7% -- and analysts expect it to decline further in the fourth quarter. Intel's growth is likely to come at AMD's expense.
Another way to bolster growth is to buy rivals. That's the current strategy of the leaders in chips for the networking and telecommunications markets, which may not recover until 2004. In networking, for instance, Broadcom (BRCM) and PMC-Sierra (PMCS), which make chips for cable modems and Internet networks respectively, have been snapping up niche players for several years. Broadcom, which had $392 million in cash as of Sept. 30, purchased wireless-chips manufacturer Mobilink Telecom in May. And analysts think it could gobble up or merge with smaller rivals such as GlobespanVirata (GSPN), a player in digital subscriber line chips. "The company isn't for sale," retorts Globespan's chief financial officer, Bob McMullan.
It's often cheaper to develop technology than buy it
Widespread consolidation is unlikely in other parts of the industry, however. While small competitors can be purchased cheaply, they often boost the buying company's operating costs and capacity -- and most chipmakers have too much of both already. Moreover, it's often cheaper for a potential acquirer to develop a new technology in-house instead of buying a company to acquire it. "The cost of consolidating is a distraction [from our own research]," says Tom Engibous, chairman, president, and CEO of the world's largest maker of chips for cell phones, Texas Instruments (TXN). He adds that TI would rather spend as much on research and development this year as in the peak year of 2000, when it spent $1.75 billion.
In fact, second-tier chipmakers that want to continue going it alone probably can because of the recent proliferation of foundries -- basically, contract manufacturers -- many of which are located in the low-cost Far East. Foundries often have the latest gear for making most chips except some microprocessors, which are based on proprietary technology. Thus, a company that sells memory or certain chips used in communications networks can design a product, then delegate the manufacturing to a foundry such as Taiwan Semiconductor Manufacturing (TSM). "This model allows more companies to stay around," says John Dickson, president and CEO of communications chipmaker Agere Systems (AGRA), which plans eventually to outsource all manufacturing. "[Because of that], I would bet that there are more [chipmakers] today than before."
Only the first-tier chipmakers are resisting this trend. Texas Instruments is depending slightly more than usual on foundries this year so as to husband its capital spending dollars, yet it will manufacture 80% to 90% of its own chips, says Engibous. TI does this, he adds, because it helps the company maintain a technological edge over most foundries -- and because it can do more flexible production than a foundry that's trying to serve numerous customers. Bleeding-edge technology customers such as server king Sun Microsystems (SUNW) and the world's largest cell-phone maker, Nokia (NOK), take these competitive advantages into account, says Engibous, who adds: "We believe we will receive more than our share of [business from these] relationships."
Worldwide foundry revenues should jump 40% next year
The foundries, meanwhile, are finding plenty of work elsewhere, says Michael Concannon, vice-president of foundry and manufacturing solutions at the world's third-largest contractor, IBM (IBM). His foundry just recently jumped one spot in the world pecking order, Concannon says, thanks to its design capabilities and cutting-edge production techniques. Price is the other selling point for foundries, especially when the job is to produce huge quantities of the same product. Analog Devices (ADI), a leading supplier of chips for various industries, contracts out 30% of its manufacturing at very low cost, says CEO Jerald Fishman, The company's capital spending, which peaked at $300 million in fiscal 2001, fell to $60 million in fiscal 2002, ended Nov. 2.
The bottom line is that the foundry business is set to flourish, with its worldwide revenues expected to increase 40% next year, to $16.4 billion, according to chip consultancy iSuppli. In October alone, revenues of Taiwan Semiconductor, Analog Device's main foundry, rose 19.3% vs. the same month a year earlier.
Meanwhile, the new round of chip company cost-cutting is bad news for producers of chip-making equipment, such as Applied Materials (AMAT). So is a temporary but huge excess of manufacturing capacity. Motorola's eight chip fabs (down from 20 five years ago) are operating at less than 75% of equipment capacity, says Motorola's Burgess. Plus, lots of used chipmaking equipment is available at fire sale prices. Thus, chipmaker capital expenditures should fall 10%, to $30 billion, next year, estimates Len Jelinek, an analyst with iSuppli. Still, the chip industry's excess capacity should fall to 16%, from 22% today, by mid-2003, he says, when semiconductor makers could start buying new gear.
Indeed, signs of an upturn are starting to appear. Already, says IBM's Concannon, his foundry is seeing strong orders in consumer-related markets such as global positioning systems and wireless local-area networks, plus chips for the newest cell phones designed for high-speed Internet access. Companies like TI and wireless technology powerhouse Qualcomm (QCOM) are sure to benefit. On Jan. 1, for instance, Nokia will debut the Nokia 3650, which can record and play video footage and take pictures. In fact, in a few years, the cost of the chips in phones will about equal what goes in today's PCs, says Mario Morales, a vice-president at IDC.
BACK TO LIFE. Another big pocket of growth is consumer electronics, such as DVD players, game-player consoles, and digital cameras. Overall, that category accounted for 18% of the total chip market in 2001, according to Wachovia Securities, vs. 14% as recently as in 2000. IBM has been shipping processors for Nintendo's GameCube, and the revenues of Big Blue's custom chip division are actually rising, notes Tom Reeves, vice-president of that operation.
Even the PC market is starting to stir, says Brian Matas, an analyst with chip consultancy IC Insights in Scottsdale, Ariz. Matas thinks it will come back to life as corporate technology spending recovers in the second half of 2003. And Intel should give demand a boost when it introduces several new products next year -- including its Banias chip, which allows ultrathin laptops to connect to high-speed local-area networks (LAN) wirelessly. Intel will also introduce a superfast processor for PCs, code-named Prescott, that could help speed a boom in broadband.
So, the chip industry may not cruise through its next upturn at the same speed it has the past several. But by late 2003, life should sure feel better than it does right now. By Olga Kharif in Portland, Ore.