Real Men Hire Fabs (from The Red Herring magazine)
Chip companies are cutting costs by outsourcing.
March 1999 issue
By Blaise Zerega
redherring.com
After a third consecutive year of poor financial performance, chip makers are determined to halt the trend. In many cases, they are looking to outsourced manufacturing for salvation. Given the rapid rise of contract manufacturing and the overwhelming costs of construction, equipment, and maintenance, it's possible that some chip companies will never build another fab.
In the short term, the new trend toward outsourcing is immediately and dramatically affecting chip companies' balance sheets. But the future impact will be more profound as semiconductor companies seek to define their core competencies and rebuild their companies around particular markets. As this happens, the ranks of pure-play foundries, test shops, and designers will swell. "You'll see the entire production cycle fragment as companies specialize," says Tom Starnes of the research firm Dataquest.
FAB CRUNCHES
Already, the major chip companies' new interest in outsourcing represents a striking about-face: the industry's attitude until recently has been that "real men have fabs." The outsourcing movement is being led by Motorola (MOT), whose Semiconductor Products Sector (SPS) plans to outsource the production of half its chips by 2002. In 1998 just 6 percent of Motorola's chips were manufactured by third-party foundries.
According to Hector Ruiz, president of Motorola SPS, the outsourcing decision reflects a choice to concentrate resources on development of the architecture, software tools, and design support needed for "systems on a chip" and other next-generation technologies rather than on fabs. It is these innovative technologies that will most likely lead Motorola -- and the industry -- out of its current slump.
Motorola SPS's outsourced products will include both mass-market commodity chips with relatively low profit margins and highly profitable programmable logic and application-specific integrated circuits (ASICs). To minimize the foundries' cut of the profits while still guaranteeing itself access to manufacturing capacity when demand picks up, Motorola plans to provide them with proprietary manufacturing, or "process," technology specific to Motorola products but possibly applicable to entire classes of chips. By all accounts, this strategy is smart; when chip demand rises and bidding wars for capacity erupt among chip developers, a preferred relationship with a foundry will really pay off.
Cypress Semiconductor (CY) has inked a similar agreement to have a foundry build its 0.18-micron programmable logic chips, which will account for 10 percent of its chip production. Chris Norris, vice president of Cypress's programmable logic division, cites reasoning similar to Motorola's. "It's a question of where to invest research-and-development dollars," he says. However, Cypress will continue to make its own logic and memory chips in order to leverage its process technology for those chips, which lets it pack high numbers of transistors and large amounts of storage on them cheaply and efficiently. Margins are another major factor in Cypress's outsourcing decisions. The company sees very profitable outsourcing opportunities for advanced processing chips whose gross profit margins are roughly 60 percent even after a foundry is paid, but it is reluctant to outsource memory chips, for which gross profit margins are in the neighborhood of 30 percent.
VALUE MATTERS
Cypress's priorities are typical. In general, where chip companies see little added value in controlling the manufacturing process, like for ASICs and other custom chips whose value depends chiefly on their design, they will begin to hire foundries. Where value is added in manufacturing know-how, particularly for low-margin, workhorse products like memory chips or where design and manufacturing are tightly integrated, companies will keep production in-house.
Intel (INTC), for example, owns all the steps involved in making its chips. "Intel feels that ownership of their fab is part of their strategic advantage. They need state-of-the art process technology to get their high volumes and high reliability," says Jim Turley, an analyst at the semiconductor research firm MicroDesign Resources.
Charles Bellfield, a marketing manager at NEC (NIPNY), which flirted with outsourcing some of its production last November only to pull back and continue its investment in a new fab, makes a similar argument. He says that in-house production pays off because of the close links between manufacturing and technological innovation: "Common sense dictates that new products come from and need new process technologies." Anecdotal evidence, like IBM's (IBM) copper interface and silicon-on-insulation innovations, whose discovery may have depended in part on the close relationships among design, testing, and production within a single company, supports NEC's position.
But aside from issues of innovation, there is another important argument for outsourcing manufacturing that has little to do with the chips themselves: the cyclical nature of the chip business. "Foundries will help with the big swings of overcapacity vs. undercapacity," explains Joanne Itow, an analyst at Semico Research. Until recently, chip companies have had few options short of opening or closing fabs when weathering a slump or sudden spike in demand. Mr. Bellfield points out that preserving a company's manufacturing capacity is a good way to manage short-term peaks and valleys because even though capacity may idle for periods, being able to activate production on short notice is sufficient compensation.
Either way, chip makers must also consider that because of foundries' economies of scale, the net expense of outsourcing is inevitably smaller than the net expense of in-house production. To illustrate the financial effects of keeping manufacturing in-house, Blake Johnson, a Stanford University assistant professor of engineering economics, adds to the host of metaphors concocted to describe the semiconductor industry. "A rabbit is a great thing for a snake to catch, but it will leave all but the largest snake immobilized," he says. "In my view, large fabs will eventually lead to the demise of all but the largest manufacturers."
Such reasoning is probably behind Motorola SPS's outsourcing announcements, which were precipitated by disappointing financial news. In the second quarter of 1998, sales were $1.8 billion, an 11 percent drop from the same quarter of the previous year; the following quarter saw a 14 percent drop from the previous year. For both quarters, Motorola SPS operated at a loss. (Fourth-quarter figures were not yet available at our press time.)
But even though the current slump is cyclical, many companies will probably continue to increase their outsourced manufacturing over the long haul. Going forward, Motorola plans to operate only one fab to support its broad product line, and although it is officially planning on outsourcing 50 percent, Mr. Ruiz says going higher is possible. Cypress also hints that its publicly stated outsourcing figure of 10 percent may increase. "Our trend internally is toward more fab arrangements for our specialty businesses," Mr. Norris says. A sign of how fundamental the shift toward outsourcing is: even Advanced Micro Devices (AMD) -- whose CEO, Jerry Sanders, is often credited with the "real men have fabs" declaration -- uses foundries when demand necessitates.
OUTSOURCERY
The willingness of chip companies to consider outsourcing is welcome news to the investment community. "In the semiconductor industry, it's very hard to be both a source of value-add in chip design and a source of value-add in manufacturing," says Drew Peck, a semiconductor analyst with SG Cowen Securities. Mr. Peck prefers to invest in those companies that take an intermediate approach, retaining some manufacturing capacity but using foundries when it is more profitable or necessary to meet a spike in demand.
The embrace of outsourcing suggests that in the future, chip companies will very rarely build a new fab, even for low-margin chips. "It's a business shift that's going to last quite a while for everyone except Intel," Mr. Turley says. Of course, the net profit margins in the chip business must be high enough to support both the foundries and the chip companies. Mr. Starnes of Dataquest warns that if the foundries' share of the profits grows too large, jealous chip companies could resume building fabs. But chip companies that do not stray from their outsourcing strategies could well reap the rewards for decades to come.
|
