Note: I produced a proper piece of analysis of the iPhone 4S (and lots of other things) for Enders Analysis - contact me for details
A little uninformed observation: Apple isn’t running one release cycle around the iPhone - they’re having to manage five.
First, there is the semiconductor side of things. I don’t play close attention to this anymore: Dean Bubley has some interesting thoughts on the process of migration from third party chipsets to the A4 and A5, the incorporation of the Qualcomm combined CDMA/GSM platform and future plans for LTE. But I think the interesting thing is that a lot of this is driven by timetables and roadmaps that don’t necessarily map well to annual phone launches. Most obviously, LTE chipsets are clearly not mature enough to go into a phone with decent battery life (setting aside the fact that there are almost no networks to connect to outside the USA and 70% of iPhone sales are outside the USA). Additional funnies like Soft SIMs (regardless of how much credence you put on that rumour) only add to the complexity of this.
Second, there is the physical design of the phone. This is always tough, as you get smaller and more worried about battery life, but with the iPhone 4 Apple set a new bar for complex and technically difficult-to-manufacture packaging that stands in obvious contrast to the black plastic that comes from China and Korea. Only Nokia has a similar interest in materials. And both the iPhone 4 and N9 make decisions about ‘the pretty box around the chips’ that affect the underlying platform - by putting the antenna around the outside of the phone Apple gained more space for, amongst other things, battery, and made the whole thing physically stronger (and then made the rest of it in glass to add back breakability in time for the new handset ;). Doing this takes much more time than designing a new piece of injection-moulded plastic - the white version of the iPhone 4 was nine months late for heaven’s sake.
Third, of course, there’s iOS. Apple isn’t going to ship a new phone two months before the release of a major new version of the OS - apart from anything else the device replacement cycle is an important tool for keeping the base up to date. iOS 5 wasn’t ready until now (I’ve been using the betas). That ALONE might have been reason to delay the new phone, whatever it looked like.
The thing is, each of these has its own rhythm and each reaches points where there’s something to ship at different times. Apple has to try to synchronise these so that two or preferably three of them hit the ‘we can ship something’ point at roughly the same time.
Meanwhile, you have to think about upgrade cycles. Almost all iPhone users are on 12m contracts and a high proportion are on 18m or 24m contracts. In other words, people can upgrade their phone every two years, but Apple releases a new phone every year. That means that the people who bought a 4 had a 3 but not a 3GS, the people who will buy a 4S will have a 3GS but not a 4 - and down the road, the people who have a 4 now will only be in the market for a new phone when the 5 is released. (Horace Dediu puts this very elegantly)
(Of course, all of these issues apply to everyone, but when you’re Samsung and you’re spraying two dozen phones into the market each year they manifest themselves in different ways. If you’re only doing one phone they get… amplified.)
In other words, Apple is actually running two overlapping phone release schedules, both of which have to synchronise with the chipsets, the hardware design and the OS, for a total of FIVE development cycles. Nice problem to have to manage.