They'd also have to license all the manufacturing IP, and for cash because they have none of their own to share. The people licensing them that IP are the people who would much prefer to be getting Apple's business as a customer, so there's not much motivation to make it cheap. And they'd either build a plant with spare capacity that they'd have to sell, in which case they'd need to sell their services to a competitor in the CE business - or they'd not have enough, in which case they'd need to buy capacity from a competitor in the semi business.
However it isn't impossible for a valid case to be made with regards to building your own foundry. It depends upon just how much control Apple wants over their future. Two years ago people would gave questioned the wisdom of Apple designing their own SoC. Today people cant wait for the next "A" chip to come out.
Really? I don't think I would have freaked out if people suggested Apple build their own SoC, it's a licensed CPU core with a licensed GPU module and some licensed RAM components with a bit of utility here and there. Now if they were talking about designing their own cores I'd be a bit freaked. The last integrated computer maker to do that successfully that I can think of was Acorn.
Designing a SoC probably only required a relatively small team, some nice computers and some fancy software. No multi billion dollar investments were needed. The people with the skills to do it existed in lots of firms, whereas the people with experience of operating a Fab are pretty thin on the ground.
Anyway if Apple were seriously going to invest that kind of money to build a key component themselves then the component they should be going for is the display. Users don't perceive the CPU, but they have a strong relationship with the screen, especially on a device like the iPad or iPhone. Not that I think they'll do that either.