Intel Corp (INTC)

[wbmw]

isn't it more typical in the semi- biz to leave the lower ASP/more commoditized products on the trailing edge processes. e.g. I believe there's still quite a bit of 65nm activity going on..

I know you meant this tongue in cheek, but let me give you an idea about older and more mature process nodes.

- Wafers are less expensive
- IP is mature and readily available
- Nodes are available for long >10 year spans
- Longer time in between product refreshes saves on R&D

In the case of mobile phone SOC's, all but the first are not really relevant.

IP is usually mature by the time it's on n-1 node (no need to target n-3), long lifespans are usually irrelevant, since there are few customers interested in SOC stability over those timespans, and yearly refreshes are almost required to gain entry into this market.

Having said that, the first bullet is pretty important, in that newer nodes are skyrocketing in terms of the cost of building the factory network, which all gets amortized into the wafer cost. Meaning, that nodes after 28nm will not get significant cost/transistor improvements, such that low-cost vendors will be willing or able to move, given the margin structure in these segments.

Intel has a huge advantage, in that they are once again in control of their own fabs, and once they transition to 14nm, do not have to pay TSMC as a middleman. That's going to result in state-of-the-art process benefits, as well as a much better margin structure. If they can make headway with SoFIA at 28nm, then by the time they move it to their own 14nm, will be a huge competitive advantage for them.