Let's start with the assumption that Doug Spreng, former president of Applied Microcircuits Corp., was dead-on in predicting at the Network Processors Conference that consolidation would get much worse in 2003. For the fabless semiconductor company, the best one can hope for in the coming year is sheer survival, in business and consumer applications as well as in communications. The number of chip companies will shrink to a fraction of their former breadth, with the biggest numbers lost in the telecom field.
Even operating under those assumptions, we're hearing some pretty wild ideas about who will survive. Andrea Cuomo of STMicroelectronics has a short list of chip companies making it in the last half of the decade, where the name of Texas Instruments is not to be found. And network processor analyst Linley Gwennap predicts that only Intel and AMCC were sure bets in NPUs, with IBM and Motorola on the fence and others left behind.
Hold on a second. There seems to be some wishful thinking involved here, in which the biggest survivors are so anxious to vote competitors off the island that they're forgetting the ability to evolve in particular environmental niches. If survival were relegated only to the richest, broadest, most technically savvy companies, then how could specialists like Atmel and AMI Semiconductor continue to find important markets?
Now for another prediction: The strongest players in general communication markets in three years will be Intel and Motorola in integer-based processing, TI in DSP and ST in mixed-signal. But that leaves a lot of room for specialized strategies. Agere, for example, can parlay general expertise in ATM and Sonet into a presence in aggregation processors. The strength of AMCC, Vitesse, Marvell, Broadcom and PMC-Sierra in physical-layer to data-link-layer processing will ensure all five some degree of viability in communication processors. IBM can look to its quiet but loyal customer base for continued success in ASICs and NPUs.
On the process front, Intel and IBM (and to a lesser extent, TI and Motorola) assume that only the integrated company able to support a billion-dollar fab will make it in comms-specific markets. Sure, all companies will need to leverage 0.13-micron CMOS designs. But does everyone need strained-lattice silicon, 90-nanometer feature sizes, low-k dielectrics and silicon germanium? Not if they're pursuing specialized, midrange apps.
If the recovering communication industry resembles a string of islands of different sizes, all positioned in different ecological niches, the giants should realize they can't vote the scrappier little guys off.
RSS
RSS| 
Digital| 
Mobile
