For some time I have been talking about the semiconductor industry as the Fab 5, since there have been five process “clubs”. A few players hedge their bets and are in more than one club. The fab five are Intel (a club on its own), UMC (along with Xilinx and Texas Instruments), IBM (along with Samsung, ST, Infineon, AMD, Sony, Freescale and Chartered), Japan Inc (Renasas, Toshiba, Fujitsu, OKI, Sharp, Sanyo, Matsushita) and the big one TSMC (with AMD, TI, NXP, ST, LSI, Sony, Qualcomm). Japan Inc in particular is messy with Toshiba tied closely to NEC (in the TSMC club) but to Sony (in the IBM club too), Renasas and Fujitsu are still sort of going it alone. Japanese politics would indicate that they will all get together somehow.
Big changes are afoot. Here are some of the things going on, ST, NXP and Ericsson wireless are all merged together into a new company (called, yawn, ST-Ericsson). Nokia has also sold its wireless unit to ST so it is presumably in there somewhere. Toshiba looks like it is going to really join Japan Inc (as if there was any doubt). TI and Freescale are both trying to find a home for their wireless groups but nobody wants them at a price they want to sell. The IBM club have deepened their technology agreements and ARM (although fabless) seems to be sort of joining the IBM club to help create energy-efficient SoCs, with Samsung both building and consuming the volume (and so I hereby rename the IBM club the Samsung club).
What about everyone else? AMD, ATI (also in AMD for now), MIPS, nVidia, UMC, NXP, Infineon, Motorola, Texas Instruments, Freescale were all bleeding cash even before the downturn got really bad, and they are reducing their footprints. All of Japan Inc except maybe Toshiba were also bleeding money (and Toshiba would have been except for all that flash going into phones and iPods, and is now hurting more after losing Xilinx to Samsung over price).
So based simply on financial strength it looks like the 3 fabs are going to be TSMC, Intel and Samsung (taking over the name badge for the IBM club) long-term. Of course other people like ST won’t lose their fabs overnight but they won’t be able to afford to keep up. And it is unclear how many of the memory houses will make it through the current downturn. Qimonda is clearly comatose already and isn’t going to wake up.
So the Fab 5 will become the Fab 3. For EDA this just emphasizes that there are too many EDA companies, as I’ve said before. Or maybe that EDA will go internal again, which is a discussion for another day.
Who would have predicted 20 years ago when TSMC was a small foundry with a non-competitive Philips process that it would be the dominant player. Kind of like predicting that Ringo would be the last Beatle of the Fab 4…oh wait, maybe that’s going to happen too.