John.Donovan - NO!. GG - Very true! Intel's understand that their strength is in the fabs. The investment in 14 was not a waste - Intel is poised to take over Mobile even with the current node and sticking with FAB development can produce more than any other investment they can make. Fables is vulnerable in the long run to automated design, FPGA and changing standards as the latest Intel moves will probably show. Instead of Smith I think that it is Jerry who will end up correct. By the way AMD and Global are watching the Intel's moves in Mobile and will probably make a move for it too.
Qualcomm is killing it lately, and has become the 3rd lagest chip maker already , with only upside from now on. The only real competitor to them is Samsung now, if they decide to become more serious about dominating the market with Exynos chips. So far they seem to be using them more like "nice to have" chips, only in some of their devices.
 - http://techdomino.com/qualcomm-3rd-biggest-chip-maker-by-revenue
Intel is too late for mobile just like MSFT. MSFT has however finally got it right and has a nice story going with Xbox, pc, mobile all providing similar UI. Also, msft has caught up to the competition and then some. Intel has a tougher problem to solve. They should spin off the fabs.
Also one reason most other companies have given up the Intel practice of short-lived foundries. Intel's business case is kind of an odd-one, anyway... they're after 14nm to keep x86 competitive in mobile against ARM, predicated on the idea of ARM not actually getting even more serious on low power, and of ARM trailing x86 in process. Those were good bets up until now, but then again, Intel is only now starting to nip at ARM's mobile heels.
My thoughts used to be like you
But I can tell you it is cheaper to buy wafers/chips at foundry then use Intel internal manufacturing.
How can be? Take TSMC 180nm today. TSMC 180nm 1st came on line 10 years ago making Qualcomm modem chips. 10 years later its now used for PMIC and other chips. Between modems and PMIC, many many other chips were made. 10 years from now TSMC will likely still be making chips (analog?, medical device chips? or something else). Compare and contrast to IDM Intel 's 180nm capacity is scrapped. All Intel's 90nm 65nm and 45nm is even scrapped. 32nm is being scrapped soon.
Now think about this. Intel is spending about $30B US dollars to outfit plants for 22nm and 14nm. Just look out 4 years (Intel's roadmap) when it moves to 450mm wafers at 10nm or 7nm . All that $30B will be scrapped in 4 years. What a expense. Foundry is much more efficient since it makes many many more chips over far longer time.
I am sorry, but the fabless model still relies on SOMEONE having a fab. The whole concept that you can somehow draw excess profit by passing your fab costs on someone else is rather ludicrous. Why would TSMC enable Qualcomm with leading edge process technology if they did not capture a fair portion of the profit? With the fab options of the Qualcomm's of the world shrinking expect to see their wafer costs go up. The fabless model only works for as long as the fab is more or less a commodity.
I can't resist a shout out to Rodney Smith for pioneering the fabless model over hoots from the "Real men have fabs!" camp. The model has long since been validated, but I'm sorry Rodney isn't still around to savor the latest vindication of his vision.
Here is the original Irish Times article: http://www.irishtimes.com/newspaper/finance/2012/1109/1224326364155.html
Possibly they are cutting training costs, which means Intel is facing tight money conditions now. Yet a quicker ramp to 14 nm wouldn't help? Maybe 14 nm is a small problem compared to something else..
Blog Doing Math in FPGAs Tom Burke 24 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...