... Apple will need to use a different solution for its 2013 iOS device models, with options including a die shrink of the current A6 chip from 32-nm to 28-nm, a more substantially revamped A6 chip, or an early A7 chip based on a 28-nm process.
Lowest risk: die shrink of the current A6 chip from 32-nm to 28-nm.
Bigger risk + bigger benefit: an early A7 chip based on a 28-nm process.
Shrinking chips is pretty routine. Safe and sane.
But I think it would be best for TSMC to start working with the A7 ASAP.
Maybe just with 2 cores like the A6 currently has. This would let TSMC get used
to manufacturing the A7, bring up yields, and ship the 28nm version in 2013.
Then they could shrink it for 2014, crank up the clock speed, and add 2 more cores.
Apple needs to ship a quad-core processor next year, especially if iOS 7 / 8 have
some new processor-intensive features (augmented reality or whatever.)
Alternatively, if TSMC isn't ready to build 28nm A7s in massive numbers, TSMC could
build a die-shrunk A6 for the mobile iOS devices while experimenting with A7 production.
Apple could put the early A7 in Apple TV, like they (apparently) did with early production
samples of the A5 in 2012. They could just disable the weaker of the two cores.