There is a fourth option to consider that would also be interested in the funding: UMC.
While I agree that it makes sense for Apple to use multiple foundies, this can still be a challenge, not only because of differences in processes, but also cost. Using multiple fabs will require new mask sets for each fab, even if they are on the same or similar processes. Although Samsung and GF's processes are likley to be somewhat similar, the fab resources used by Samsung are not part of the Common Platform alliance. So, I doubt that it is the same process. Also, Apple has already invested a significant amount into having dedicated resources at Samsung.
A few other items to consider are services, IP, and IP protection. Samsung and TSMC both offer more extensive services, IP, and IP protection than the other foundries, which are all considerations. In other words, switching from one fab to another is not like going to a grocery store across the street. It is a major engineering effort with a significant cost. As a result, any change takes time. In the end, I would expect to see two of the foundries, not more, manufacturing for Apple, and even then, they may be manufacturing different products.
I think flashing $1 billion in front of GloFo at this juncture would be serious enough for it to give Apple a serious consideration BUT GloFo can do more outside of Apple's business. GloFo should certainly leverage its USA fab presence to its advantage while there is a trend toward more on-shoring.
While majority of the MEMS market is still dominated by players with the IDM-model, GloFo can leverage its experience with its MEMS fabs to advance the fabless model for MEMS.
Where is evidence of US fab plan for TSMC? As I know it was media speculation based on Chang comments only that he wouldn't rule it out as a possibility. This is not like any plan. Then later he was more firm and said basically TSMC would stay concentrated in Taiwan to make best utilization benefit.
Blog Doing Math in FPGAs Tom Burke 24 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...