And because IDC had mentioned it before, its not that the additional capacity for D1C will be making only xeon parts (as either plant can make either chip), but the additional capacity will be used to add retail sales of xeon chips and expand clarkdale/arrandale production (we'll probably see an expanded line of value clarkdale/arrandale chips and 32nm atom at this point. The number of atom codenames has confused me, AFAIK the atom SoC chips are TSMC or GF and the not SoC chips are still from Intel's fabs).
Atom won't matter as they'll want that chip to be on the 32nm SoC process, which is seperate, and it will be a year later from the regular 32nm, which makes supply issues moot.
As for TSMC, its not as clear, but this is as far as I can find.
Moorestown:
Lincroft: 45nm Intel SoC process
Langwell: 65nm TSMC process
Medfield:
32nm SoC CPU/IMC/GPU with Intel process fused with 32nm TSMC I/O chip?(like how they do with Lynnfield and the PCI Express controller) Looks pretty likely as they have a roadmap for SoC process well into 22nm, and I'm pretty sure Intel in hell won't give TSMC their 32nm process technology.
Arrandale/Clarkdale is ENOUGH alone to choke the 32nm supply. These processors will sell probably magnitude more than Bloomfield/Lynnfield did. That plus delay/cancellation of their 45nm dual core Nehalem part plus a bit of the economy tangled up in a rope might explain it.