The way I see it to the customer:
ASSP ASIC FPGA
no chip design time design time + Si time short design time
faster prototyping
great for std stuff too expansive for std OK for 1st guy who
eg. DVD when LSI stuff wants fast to market
even provides ref
design
no changes in SI need time to change easy to change design
high performance high performance lower performance @
very high end?
minimum lead time 1 month+ lead time no lead time, program
for product when you need it.
hopefully cheapest most expansive, IPs good comprimise
for same performance cost + Si
As much as FPGA gets more densities and faster, ASIC also gets bigger and easier to design. I think both will co-exist and grow as semis in general will grow. I look at it more of a question of the growth rates that they have, not one will take from the other. My guess is FPGA is growing faster than ASSP/ASIC/SOC.
When I initially picked LSI, I wanted 'diversification' vs betting on one product line/type/company. FPGA was known to be the fastest growing sector in semi and was really between ALTR vs XLNX. If I were to diversify and buy stocks in both and one kills of the other, my average results will be average. To bet on one vs the other, really I did not have the expertise to pick. As it turns out, both won because the market grew enough for both.
I picked LSI because of the diversification of its products and the TAM that it played in. I was willing to give some gains because I knew that when the semi turns, LSI will participate because of the breadth of its products. Of course when LSI dived when it restructured .... it immediately turned into the buy of the year!
I have always looked back now at ALTR/XLNX vs my original $24 buy-in on LSI. My analysis back then was correct. Would have made a little more money on FPGA like the analysts recommended. But then, for the money that I gave up, I did get that little bit extra sleep.
And FPGA isn't without any enemies chasing it either. When the market grows big enough, you get general products replacing it at commodity prices.
In short, semi not a fixed pie but one that is growing and hopefully will grow rapidly when the upturn really takes hold. One should look at growth rates for different segments of that pie and not as a Pac man win-lose game between the two sectors.
Wilf was also correct about fablss becoming chip-less when there's a capacity crunch. Hence at the last round all the fabless guys were buying into new fabs to assure of their wafer supply. Is this round going to be different? TSMC was making 50% margins during the last round. They were re-couping their fab investment within the first year! Those fabless guys did not hand over those margins willingly, I can tell you that. During downturns, you don't want to have a fab. But during upturns, you'd better have a fab (remember Wilf comments about CEOs calling him directly to assure of their product supply?)
patrick
PS Come to think of it, it is better to farm out some of LSI's products to foundries now. That ties up more capacity under LSI's control for the upturn. Not a bad strategy. You've got to hand it to Wilf and his crew. They are actively managing the company and also keeping their eyes open while doing it. |
