The decision could have been made 18months ago. "A" parts could have been designed to run on gate last, made and qualified. How would we know any different? The problem has been a secure source of the right kind of memory.
There is a lot of capacity underutilized at Intel, it is reasonable that that capacity gets filled soon.
Apple/TSMC is a bad joke.
If Intel priced Apple chips at the Samsung price ($17.50 according to iSuppli [I think more like $25]) they would generate about $16,000 revenue per $5000 cost 22nm wafer. Moving to 14nm the cost would go to about $3.50 while the resale price would stay at $17.50...why would it go down? So now you have 80% gross margins.
By that time Intel will be making new architecture x86 based SoCs that run circles around the A chips. What will Apple do then?
Oh, Intel will do the apple business alright, but not until memory sourcing is firmly under control.
You can bet that Intel has already seen the new Atom cores in an SoC in engineering volume. I would bet that it just dances around anything any foundry can do now or in the future. So, how long do you think Apple would have Intel fab and inferior product for them while Intel is supplying x86 SoCs to the rest of the world?
In the unlikely event that the Apple "A" chips outperform the new Atom based stuff, Intel BETTER get in bed with Apple...or get ready to sell those excess fabs for 50 cents on the dollar.
I would be really surprised to hear of an Apple/Intel deal. I am about 90% certain that Apple already has a deal with TSMC and has given Samsung its walking papers. Not to say an Intel/Apple deal won't happen down the road but it just isn't going to happen in the next year. JMO
I pretty much agree with that assessment but at the same time I think Apple is totally nuts putting their future in the hands of TSMCs ability to quickly ramp 20nm to the levels that Apple requires while dealing with the rest of their customers. Just one of those things where one would usually conclude "this is going to end badly". If TSMC drops the ball, Apple will likely never recover. Those competitors buying from Intel will make sure of it...
This is why the press given to the idea that Intel will get 10 percent of the business only sounds prudent to me. With all of Apple's money, why not hedge the bet with making sure a rapid conversion to Intel is possible...
Apple's Swift and 32nm Atom are in the same integer performance ballpark with the chips swapping leads in benchmarks. Atom is also taking the big step of going out-of-order at 22nm which should put it clear of Swift which already has used this technique. Atom has more performance potential.