Are the present generation of managers in the semiconductor industry too lily-livered to make the big decisions?
Gilles Delfassy, head of wireless products at Texas Instruments, tells how TI was prepared to make a chip (the TMS320C5X DSP) for $900 and sell it for $42 for a year in order to hang onto a design win in an Ericsson TDMA basestation in the early 1990s.
Another TI-er, Dr Larry Hornbeck, developed digital light processing (DLP) technology for displays based on movable mirrors for 19 years before it was put into production.
Intel co-founder and former CEO, Dr Gordon Moore used to say “The only sensible strategy is to bet the company regularly.”
Nowadays, as company after company goes fab-lite, there are fewer and fewer CEOs willing to bet the company by building a fab. And, unless you’re Intel, or a major foundry, you do bet the company if you build a fab. But, there again, it was always like that.
Fabs cost the same now, in relation to industry revenues, as they did 20 years ago, according to a presentation by Malcolm Penn, CEO of Future Horizons, at last month’s IFS2007 meeting.
As for selling a chip for $42 which costs $900 to make, would a company be prepared to do that now? “We see less of that than in the past”, says Delfassy, “people are less adventurous and not willing to take such risky bets.”
It’s a bet which paid off. Wireless products account for the biggest portion of TI’s $14bn+ annual revenues.