"INTC faces a big conundrum... focus on servers to grow their monopoly there and keep margins up, or spread the risks by trying to grow market share in the slow-to-no growth PC and now phone/tablet spaces, where margins will only decrease."
[This is what you call analysis??? Intel is going to continue all three lines. There is no conundrum. They will continue to dominant servers, they will continue to dominant PCs and they will build profitable mobility market share. As indicated in quite a few posts and articles, the margins do not in any way prevent Intel from a profitable share of smart phones and tablets. And Intel's fabrication will provide a distinct edge in all three markets....]
Well, you may not agree with the viewpoint I presented, but at least unlike most of your posts, it's not simply a regurgitation of articles I can find and read without your 'help'.
- They will continue to dominant servers -
They very likely will, but whether or not they will be able to maintain their high margins is not a given.
- They will build profitable mobility market share -
That remains to be proven.
I suspect there will be difficult challenges for INTC to grow profits. Maybe you can explain why the market appears to be pricing in weak prospects for INTC if their future is really as bright and certain as you portray. And please don't respond with the simplistic argument that the market is ignorant of INTC development and production capabilities. I don't buy that for a minute.
I agree that they very likely will remain on top. It is hard to estimate the margins. Each process shrink doubles the die per wafers and geometrically reduces the defects per die. Costs will go down. Power consumption drops to the resulting systems are cheaper to build.
One big, new feature on Haswell servers will be the transactional memory (TSX). This feature alone should provide performance improvement on the order of double that of Romley. Think of speculative execution of locks. When software sets a lock, Haswell monitors the data it is going to access and does not disrupt the bus access unless there is a REAL conflict. Lock conflict is a real scalability bottleneck. I don't know of any similar ARM capability.
tablets and phones -
Intel will chew down. Lexington has a $6 cost. IMO, PC will remain flat to up a couple % over time. tablets and phones are generally incremental.
Technology wise, everything is going low-power, all three of these market segments. Intel does not have a fabrication advantage in low power. It's trying to build up one by accelerating deployment of 14nm by exploiting its first mover advantage on finfets. All the 3 major foundries are very well funded and correspondingly trying to match Intel's acceleration with their own accelerated schedules. It's not going to be a cakewalk for Intel on fabrication and even having a one node fabrication advantage at low power does not guarantee success. For example, assuming Intel is able to bring 14nm to low power one year after its 22nm low power deployment (end of 2103) in the end of 2014, by that time, foundries will have 20 nm in volume and big.LITTLE configurations will be both power and performance competitive with Intel's 14nm SOC offering. That's ARMs side of the story. Whether Intel will successfully bring 14nm to low power in end of 2014 and whether big.LITTLE will actually make a difference are both speculative at this point, but the changing funding trends for the foundries don't look good for Intel.