Well it seems this is a serious consideration now. But to succeed as TSMC-type foundry, you have to have the ability to have many different focus areas. Besides logic, also mixed signal, analog, embedded memory, high voltage, CMOS imaging, RF, BiCMOS, etc., across several generations (at least as old as 0.35 um). But Intel probably already sold off all or most of its pre-90 nm 200 mm equipment. Serious foundry planning probably would have prevented some fab closures. So it would mean Intel has at least several years catchup to TSMC. It would be interesting to watch the competition at high-end (sub-32 nm) though. Would Intel beat out Samsung and TSMC and GF in this key area? Would commmoditization of foundry services eventually happen?
If I can put myself in Intelís position, I will go with one company that is not that big but has a unique technology that may be prevailed in the future. That is much lighter weight on their tiny business area where they want to growth. ASIC and Processor business area are not profitable enough, but FPGA or PLD will be. Altera is already working with TSMC, if I remember right.
For readers interest: About 2 years ago during an earnings Conference Call, Otellini was asked about running a foundry operation. He didn't reject the idea and mentioned various issues such as competitive, legal and whether they could offer services on their best node-1 basis. Also, in an SEC filing about a year ago, Cray mentions a "contract" with Intel for a certain unspecified "silicon" device.
Having opened up a huge lead in the foundry nanometer nodal shrinkage, Intel can now exert enormous leverage throughout semiconductor manufacturing. Will they offer their 22nm technology also to Altera, for instance, to transform their competitiviness against Xilinx who will be using ARM technology at 40 nm, say?
Intel can pick where they provide key advantages of speed, price, and even power consumption to companies collaborating with them to the detriment of competitors.