IIRC, Intel used to due something like you are suggesting with their older fabs but for memory, prior to exiting that business. They also used to run the chipsets a node behind the CPUs.
The problem with being a node or two ahead is that periodically there is a fundamental tool change needed (EUV for example) and the cost of getting across that early means the early adopter foots the entire bill. Way back, BTW, Intel also developed a lot of tools in house, and given their resources, I somewhat wonder why they don't expand that approach and try to take the critical tools proprietary. I think for awhile they did have either EUV or Xray work in house, but I don't know what happened with that.
But as I've pointed out before, Intel does not have a birthright for a node ahead. The Asian fabs run a lot more area than Intel does, and there is no assurance that Intel will always be ahead. At some point when the $ spent by TSMC, UMC, Samsung & GF dwarfs Intel, then one would expect that the foundries might close the gap, assuming they decide to pool they R&D effectively. I've often wondered why China Inc doesn't decide that they might use $50B of their FX holdings to see if they can't take the lead in semiconductors. They could afford to fund the battle from tool supplier to fab if they really wanted to. They could replace AMAT, KLA, INTC, etc. So what if it takes them a decade or two?
But the real issue that I see currently is the Genie is out of the bottle on ARM SOCs being commodity products, and I don't think the market will ever revert to sole source x86 in the mobile sector. So times have changed for Intel, and until they show they understand that from a business perspective, rather than just thinking if they get the best part, everyone will flock to them, I don't think they will do will in the new reality. |
