Chip cycle forecasting
By Mary Lowengard, illustration by Tom White - December 1, 2000
Like the shadow of the moon passing over the sun in an eclipse, the specter of the trough of the semiconductor cycle is coming-the only question is, when?
Soon, according to Wall Street. Salomon Smith Barney semiconductor analyst Jonathan Joseph downgraded the sector from "outperform" to "neutral" in July, his action serving as the modern-day equivalent of throwing down a gauntlet. U.S. Bancorp Piper Jaffrey analyst Ashok Kumar's downgrade of Intel Corp.'s stock followed in mid-September. The subsequent sell-off brought the company's stock down seven points. A day later Donaldson, Lufkin and Jenrette's Boris Petersik turned his thumb down on the earnings of memory chip maker Micron Technology Inc.
(Reality check: The consensus of a number of securities analysts was that Micron would earn 97 cents a share for its fourth quarter. The company reported in early October that it earned $1.16. In September, Petersik forecast that Micron would earn $2.43 a share for fiscal year 2000. In fact, the company earned $2.56. So Petersik was off by 13 cents a share, a wide margin.)
Thomas Kurlak, the former Merrill Lynch & Co. semiconductor analyst, who spent more than two decades following the industry and a dozen years on the Institutional Investor All America Research Team (six of those years in the No. 1 seat), is somewhat sympathetic to his former Wall Street colleagues' positions.
"Wall Street analysts are very concerned about not repeating the mistakes they made in 1995. That left deep scars, so there's a lot of rear-view analysis going on," notes Kurlak. From his vantage point as managing director at Tiger Management LLC, where he manages hedge fund legend Julian Robertson's portfolio, the unrest in the financial community originated when one analyst noticed over the summer that the unit growth rate, year over year, had reached a level that had previously been a high water mark. "That was his tip-off, and he sounded the alarm," Kurlak says.
This rolling stone gained momentum when "a few oddball events" (Kurlak's characterization) happened: some glum pre-announcements by companies that they were going to miss their earnings numbers and then a succession of analysts lining up to say it's all over. "The problem with their analysis is that it doesn't take into consideration the United States' or the world's economy," Kurlak summarizes.
Nor any of a half a dozen other factors, it would seem. These include the broadening of the end-markets, the increasing importance of the communications and automotive segments; the impact of the fabless industry on production and, ultimately, ongoing globalization of the industry. All these in concert are conspiring to change the cycle-not causing the industry's cyclical nature to disappear but moderating it.
Industry analysts, a group more insulated from the vagaries and pressures of Wall Street analysts in their analysis, present a sunnier outlook. The Semiconductor Industry Association stands by its estimate that chip sales will end the year up 31%, at $195 billion. It projects 60% total growth in chip sales over the next three years. Industry executives line up behind these numbers and say, "Downturn? What downturn?"
South Portland, ME-based Fairchild Semiconductor International Inc. CEO Kirk Pond thinks it's important to remember three things when considering this cycle: "One, there are more people in the world with more disposable income buying more PCs, cars, pagers and telephones today. Two, the number of semiconductors within products is increasing. And three, new applications that expand the business are developing. Just think about the Internet!" (For "The gospel according to Pond," see box.)
Pond's points are supported by statistics generated this summer by Cahners In-Stat Group, as reported in its August publication Semiconductor Outlook 2000: Nothin' But Blue Skies. Here, it's made clear that "everything pertaining to communications will be hot in the coming decade. The primary growth drivers will be the wireless, cellular and information appliances markets, Internet networking equipment, and broadband applications markets."
|"There are more people in the world with more disposable income buying more PCs, cars, pagers and telephones today." -Kirk Pond, CEO, Fairchild Semiconductor|
Indeed, cellular handset shipments are anticipated to increase from 250 million in 1999 to 1.4 billion by 2004. In-Stat believes the worldwide wireless market will rocket from a $46.2-billion industry in 1999 to $131.5-billion by 2004, and the Internet-related networking equipment market will grow from $37.2 billion in 1999 to $104.3 billion five years later. In-Stat foresees calculated compound annual growth rates over the period 1995 to 2004 in double digits for two end-use segments: automotive and communications, with the latter leading the way with a 13.99% compound annual growth rate.
Perhaps the greatest sea change chip industry analysts point to when differentiating this cycle from those in the past is the development of more-and potentially more robust-end markets for semiconductors. "Past cycles have been based on the DRAM experience," declares J.J. Yamaguchi, vice president of marketing at NEC Electronics in Santa Clara, CA, a unit of NEC Corp. "Now that the industry is moving toward other markets, you're going to see the cycles get flatter."
He cites specifically two examples: The automotive markets, where the silicon content of the 18 million cars produced annually in the United States rises every year, and telecommunications. Chips for networking are driven by Internet usage, he explains, and the telecomm opportunities are virtually "limitless."
G. Dan Hutcheson, president of market researcher VLSI Research Inc., San Jose, CA, theorizes that the historic lack of stability of the chip industry can be traced directly to the lack of stability of its historic chief customer, the computer industry. But does this mean that as the end markets for products broaden the chip cycle will stabilize?
Not necessarily. For example, it appears that predicting annual handset production is just as tricky-and unreliable-as predicting sales of PCs. Points out Jack Quinn, president of Micrologic Research, a Phoenix-based firm focused on the wireless communications industry, there's no denying that "just about anything using semiconductors is a growing market." But he also characterizes forecasts issued at the beginning of 2000 for equipment demand for handsets as "wildly overoptimistic," starting up in the 600-million unit range and gradually being scaled back to between 300 million and 400 million.
To Kirk Pond at Fairchild, even errors of this scale are irrelevant. The point, as he sees it, is that the "telecomm market has been growing quicker than any of the other markets, driven by wired and wireless. You see all these people running around hysterically about `the numbers'; but whether it's 400 million [handsets] or 405 or 402, there's no denying that we've grown ourselves a tremendously large industry here."
And it's not just the wireless digital phones, other new Internet-ready devices also are driving growth. It's also the build-up of the necessary infrastructure to support both the devices and the bandwith on which they operate. Pond believes his own company mirrors the growing importance of this communications segment: "Our business has grown from 15% telecom to over 30% between 1997 and last quarter," he states. "Of course, the overall pie is much bigger," he adds.
|"Unchecked commentary from the rumor mill is this industry's worst enemy." -Thomas Kurlak, managing director at Tiger Management LLC|
Jack Quinn of Micrologic Research says that the broadening of the markets for semiconductors has been shifting gradually over a period of 10 years, with the communications segment becoming more and more important over time. But that doesn't mean that this area will be immune from an imbalance in the supply-demand equation.
Still, "for the next year it doesn't look like there's going to be any slowdown in demand," he says. A self-confessed technophobe (he bought his first mobile phone just last summer), Quinn doesn't put much stock in technical indicators. "They're always great in hindsight," he says. "[Yet] this cycle just doesn't seem like the ones we've gone through in the past. It just keeps going and going and going."
The second-fastest growing sector this year is consumer, projected to have increased by an astonishing 43% in 2000 by In-Stat, which is both good news and bad news. The good news is that it's another robust market to help diversify the end-product mix. The bad news is that consumer markets, accounting for about 20% of semiconductor end-use, according to VLSI Research's Hutcheson, exhibit their own cycles that, when layered over the semiconductor cycle, complicate and obscure. All of a sudden there's a back-to-school cycle and Christmas buying. It's Hutcheson's theory, for example, that the 1998 crisis was less precipitated by the collapse of the Asian financial markets than by a fall-off in consumer demand for PCs. Simply put, Hutcheson says, "Nobody needed one."
Mary Olsson, principal analyst in the semiconductor department of market researcher Dataquest Inc., San Jose, CA, integrates several different trends to come up with her pronouncements: "We follow the products, the applications markets, anything that goes into the electronics industry." Dataquest says North America is going to end up at 35% to 36% growth this year, and will play off that momentum over the next 12 months. The firm's forecast for 2001 is 38% growth. "You don't just turn off the spigot once the market gets going," Olsson observes. She too points to the proliferation of start-ups hitting the market with new products. "There are so many more end markets now. There's nothing on the horizon we can see that says this is going to stop."
These new markets have paved a path for smaller companies to jump into the action, like San Jose, CA-based Radiata Inc., which manufactures wireless LAN chipsets. And some of these start-ups have up-start ideas that shake up the ways their industries traditionally have been run.
"At the end of the day, people who are manufacturing chips are having to take on the engineering knowledge base of their customers to understand the systems and issues. This is a real stretch for the folks focused on the PC sector," observes Christopher Fisher, Radiata's vice president of sales and marketing. "This is one reason smaller guys like ourselves have been able to take on the bigger guys and succeed."
He also finds that "there are a lot of preconceived notions about the chip industry, some of it based on history and some of it based on conjecture." He emphasizes, "We're in the wireless communications business, where a lot has changed in the last 24 months, including a number of technological innovations that allow you to do things you couldn't do three years ago." So opinions tend to get outdated quickly, whether they are right or wrong.
The chips are down
Brian Matas, vice president of market research at the Scottsdale, AZ-based consulting firm IC Insights Inc., believes that supply will start to meet demand in the middle of 2001. Once that happens, growth will slow and the cycle will have turned, he predicts. His view is that although "there's a lot of activity in communications and the consumer segments, when you get right down to it, PCs still dominate the industry with easily 50% of the marketplace."
Matas sees red flags when capital spending as a percentage of semiconductor sales starts to exceed industry-wide averages of 22% to 23%. "During the uptimes, demand starts to increase, so manufacturers start putting in requests for more facilities. Then everyone starts doubling their spending. When you get two years of that in a row, that's when you start to get a surplus. It only takes one Compaq [Computer Corp.] to say `I don't like being charged $10 for these parts,' and boom, there's your downturn." So far, Matas reports, capital spending is up 75% over 1999, accounting for 27% to 28% of semiconductor sales.
|"There are so many more end markets now. There's nothing on the horizon we can see that says this is going to stop." -Mary Olsson, principal semiconductor analyst, Dataquest|
Tom Kurlak's personal measuring stick comes straight from the top. "Gordon Moore [Intel's chairman emeritus] once told me that the semiconductor businesses should grow between two to three times faster than the gross domestic product, so if it grows next year at 3% to 4% it's still healthy." Kurlak also cautions about making year-to-year comparisons against depressed numbers. "There's also a lot of misinformation," he believes. "Unchecked commentary from the rumor mill is this industry's worst enemy," he says.
Another metric Kurlak considers critical is inventory levels. "It's a given that consumption of chips goes up every year. Prices rise and fall based on supply and demand." The best way to measure whether supply and demand are out of balance is to look at customer inventory and supplier inventory. Right now, pricing appears to be pretty stable, with no dramatic rises or falls, and inventories, Kurlak maintains, "seem to be very much in control. So long as the Fed doesn't do anything crazy."
Dan Hutcheson says his money is on a slowdown in 2002, not 2001. "The economics of this industry are kind of obtuse," he maintains. "When you're growing so fast with so many new people coming in, there's a kind of cumulative psychology that holds it together. Nobody wants to see it drop."
The fabless manufacturing effect
Another critical element of the big picture is the role of the fab industry in helping mitigate the volatility in the chip cycle. Notes Cahners In-Stat Group, "The semiconductor industry is ramping up fab capacity at record rates, adding greenfield fabs and capacity to existing plants." In-Stat predicts that foundry "outsourcing" is one of the trends (along with continued consolidation, virtual supply chain management and better coordination between chipmakers and their equipment suppliers) that will ameliorate industry growth.
"The foundries are the `outsourcing future'," In-Stat maintains, estimating that by 2012 they will be supplying half of all chips produced. Both IC designers and integrated device manufacturers are shifting to the outsourcing model, which helps them avoid the high cost of new fab construction and exposure to capacity problems.
"It's healthy for our industry to have developed this fabless phenomenon," states Fairchild's Pond. "It creates opportunities for small design houses with a lot of creative talent." Although he acknowledges the sentiment that the foundries might serve to dampen the volatility of the cycle, Pond believes "the jury's still out on that."
He cites the example of [Austin, TX-based supplier of high-performance analog and DSP chip solutions for Internet entertainment electronics] Cirrus Logic Inc., "real innovators who managed to get themselves in a world of trouble agreeing to too many deals." Just because you're buying capacity doesn't mean you don't have to put up the money. At Fairchild, approximately 20% of the company's wafers are farmed out to foundries.
A world of opportunity
Yet another factor to consider is the increasingly global nature of the industry. Dataquest's Mary Olsson states that "the Asia-Pacific region is going gangbusters, at 40% growth compared to last year, Japan at 37% to 38% and Europe at 39% to 40%." Semiconductor Equipment and Materials International, a San Jose, CA-based trade association, predicted in mid-September that the market for chip manufacturing in China alone would grow nearly 400% to $4 billion by 2003.
Ron Leckie, the CEO of an Austin, TX-based market research firm called Infrastructure, observes that the market is actually global in two senses: There's global demand for the end products, such as cell phones; and "there's global spread of the manufacturing base." It's a sign of the times-good times-when "lots of people around the world want to build DRAM fabs. They think it will make them a first-world country," observes Fairchild's Kirk Pond, who adds dryly, "It's like owning your own airplane." Still, he thinks that this time around, people learned a lesson from last time around, and are holding their impulses in check.
Sunnyvale, CA-based Silicon Storage Technology, reports Derrick Best, senior vice president of marketing and sales worldwide, believes that China represents the new frontier for the silicon chip industry. His company supplies flash memory to the consumer, networking, wireless and Internet markets. "We see very rapid growth [in China.]" Indeed, Best says that demand in China may be what prevents the industry from sliding into a trough. "We see the softness in the PC segments, with a lot of companies overbooking. The market isn't growing at the same 15% to 20% rate. But we regard it as a breather until China comes on line."
Whom do you trust?
So with all this conflicting forecast information, whom do you believe? Thomas Kurlak's advice is straightforward. "Trust your customers," he urges. "Spend your time figuring out what your customers are delivering in their products. If they're doing well, you'll do well." In the semiconductor industry, he says, Wall Street analysts "spend their time worrying about the 10% of the time the industry is down, and generate all this fear. Your best bet is to talk to your customers, constantly monitor inventory levels and make sure lead times correlate with inventory levels."
"Between me and my customers, I think we know a lot more about the business than Wall Street," states Pond. While he acknowledges the analysts who follow his company do a good job, he feels their analysis about the industry is inevitably flawed because they are forced to synthesize information from many sources. "In the semiconductor industry, there's just one thing that drives the business and that's new products," he states firmly. "This is not new news. It's our obligation to create products that are better, faster, cheaper, more highly integrated, lower power and solve a problem." New products at Fairchild generate more than 30% of his company's revenue monthly now. "And they're made hand in hand with the customer," says Pond.
Also not new news
The bottom line is this industry is cyclical. "No matter what you hear from the analysts or the companies, what goes up inevitably comes down. As soon as you start to hear industry representatives say it's different now, there is no cycle, you know the end is near," IC Insights' Brian Matas muses. "When all they talk about is how their growth is straight up and to the right, pull out your umbrellas."
"History tells us it's going to get worse," states Fred Shlapak, president of the Motorola Semiconductor Products Sector, an Austin, TX-based unit of Motorola Inc. "But what does slow growth really mean? With an industry average compound annual growth rate of 17% growth over 40 years, 20% growth is still pretty good." He also maintains that, in the networking sector, software development is lagging the hardware being produced. Once it catches up, there may be a considerable sales surge.
Others, such as Hutcheson, urge executives to just go with the flow. "The issue of next year is not all that important in the long run. You can't run a business that way. You have to run it knowing that the future is uncertain, and concentrate on building market share." His advice: "Be prepared for anything, wherever the cycle is."
Retorts IC Insights' Brian Matas: "[Remember,] it's never a soft landing. It's just one big cliff."
Mary Lowengard is a contributing editor at Institutional Investor. Contact her at email@example.com.