by Debra Vogler, Senior Technical Editor
Dean Freeman, research director at Gartner-Dataquest, pulled no punches during a presentation at SEMI’s Silicon Valley Lunch Forum (Santa Clara, CA, 3/29). With silicon manufacturing fast becoming — if not already — a commodity, those who do not have IP and design as value adders will start dropping lower and lower on the food chain, and with that fall will come a drop in profitability, he told the audience.
In addition to the fab-lite strategy implied by recent announcements from Texas Instruments [“TI cuts capex and shifts R&D strategy“], STMicroelectronics [“ST: Crolles to stop at 45nm, 32nm partners sought“], and others [e.g. Freescale and Sony], Freeman pointed to the new reality he calls ‘process-lite.’ “Firms are saying, ‘we are no longer going to do process development because it costs us basically $500M/year, or about $1.5 billion/technology node, and we’re going to take that money and put it to the bottom line,'” he said. With that decision made, these fab lite/process lite companies can let TSMC or IMEC, for example, do the process development technology.
This new trend for IC manufacturers also means a tougher sell for equipment and materials suppliers, “driving fewer and fewer customers for equipment and materials suppliers to sell their wares to, and also to get opportunities to do R&D,” said Freeman. He urged that it will be “very important to get integrated with your clients and their consortia to ensure that your technology is adopted into the next technology node.”
Still, there are many more opportunities than existed in the 1990s to make money at the trailing edge. Freeman presented a slide (see figure above) showing how devices ranging from discretes/power, to analog/linear, and so on, all the way to digital mixed-signal ASICs with SOC, DRAM/NAND, and MPUs could be manufactured between 0.7µm to 90nm.
Updating Gartner-Dataquest’s revenue forecast for the semiconductor industry(see figure below), Freeman said that computing (getting a push from the multicore wars and servers), consumer electronics, wireless communications, automotive, and industrial applications are expected to drive CAGRs up over the next three years. So far early in 2007, however, the market is showing some weakness, and overall 2007 is projected to be a slower year with 6.4% growth to $276 billion in revenue. The industry is projected to exceed the $300 billion revenue mark in 2009 (at $306 billion), and settle into a 6.2% revenue CAGR between 2006 and 2011. “We’re no longer in the world of 17% CAGR,” he said. Pricing wars amid memories and microprocessors are driving down growth in those sectors, though ASSPs and nonoptical sensors, analog, and MEMS provide upside growth, “especially as analog and MEMS are being married more and more.” — D.V.