How Powerful Are The Foundries Going To Get?
Just how much power is the foundry industry going to have? Intel says it will ship its first 45nm production microprocessors in the second half of this year. TSMC is saying it will run commercial 45nm wafers in September with volume production in the first half of next year.
There may be a few months difference in it, but only a few. Whereas Intel has only one client; TSMC has to make its technology applicable to many clients. That’s a bigger task. For a foundry to be a few months behind the leading IDM is a big achievement. If foundry is going to be this good, one can see why Freescale, Texas Instruments, STMicroelectronics and NXP have all recently said they are going to give up pursuing fundamental CMOS process R&D. They all say they will buy in basic CMOS processes and develop specialist derivative processes for specific applications from that basic CMOS process. But now w TSMC is saying it is accelerating the rate at which it introduces its derivative processes. Soon there won’t be much point for IDMs to replicate foundries’ derivative processes, and then their fab-lite strategies will slip into fabless strategies, and they will become design and marketing houses. How much power will the foundries have then? As Wally Rhines, CEO of Mentor Graphics, pointed out recently, the yield of chips increasingly depends on knowing, and designing for, foundry processing data and the decision of how much, or how little processing data is released, is entirely in the hands of foundries. To be succinct, the foundries will have the semiconductor industry by the balls.