Ceramics

The simple fact of the matter is, Apple has scaled themselves into a position where any major material change to the iPhone must follow one of three paths:

Machine a New Material
Similar to the switch from 6061 to "7000 series" aluminum, Apple could find a new material that could be swapped into the current manufacturing footprint in a straightforward fashion. 7000 series aluminums machines about 20% slower than 6000 series, so they could absorb a somewhat substantial cycle time hit, but continue to leverage the decade of process knowledge, capital equipment and partnerships they have cultivated. In a way, we see this with the Watch being made from 4 different materials, with 3 of them using very similar processes. Perhaps a new magnesium or titanium alloy? Machined carbon fiber (a trend in the high-end knife world... and I know lots of Apple engineers and ID folks are knife nerds).

A New Material with a Short Cycle Time
This is basically the promise of Liquid Metal - essentially a very crazy strong metallic that can be injection molded, but comes out of the mold with all the fine details and beautiful surfaces in one, fast shot. Apple has had this partnership public for many years now with an exclusive license agreement with the Liquid Metal patent holders. They could be working with them, while also working with injection molding equipment makers to build modified/tuned machinery (::cough::). With this process, Apple would go from 20, 000 CNC mills making iPhones, to a few hundred, or perhaps a thousand, highly modified/specialized/proprietary injection mold machines. Doing this would be a major undertaking and huge technical achievement, but it is entirely within the capabilities and resources of Apple.

A New, High Cycle Time Process and Material
The final option, is for Apple to bring a new process online with cycle times similar to the few minutes of machine time required to make the current iPhone. To do this, Apple would need to essentially flip the switch on an investment of roughly the same size of the current, machining footprint. Not only would this be a logistical nightmare, but finding a company to produce that much machinery without 3-4 years of lead time would be almost impossible (as a reference Haas and DMG Mori are neck and neck as the world's largest CNC machine manufacturers, and they each only make about 15, 000 a year).

While the sheer scale of that last option might only be fully comprehensible to someone like Horace Dediu, the best argument I can make against it is from The Hunt for Red October. In one of the movie's best scenes, the US National Security Council is made aware of a new crisis in Russia when the satellite flyover of their major Atlantic port reveals heat blooms in the engineering compartments of every ship in the fleet. For Apple to bring a whole new long-cycle time process online for the next iPhone (now 10 months from launch), they would need warehouses with thousands of machines already in situ, with thousands more in production. Teams of analysts would have been reporting on such a move for months already.