I've thought about this, and this is what I think :

1) One of the chief benefits of POET in the long term is the ability to integrate the function of several separate chips into one:

Here's the math:

1 + 1 = 1

or in the case of a cell phone:

1 + 1 + 1 + 1 + 1 + 1 = 1

My point is that in the end we should require fewer wafers for the same product.

2) POET should be able to fit more chips on a wafer as they continue to scale past 40nm.

3) As has already been mentioned, the demand for POET-based chips should necessitate the mainstream use of 8" GaAs wafers.

4) Many older underused fabs will start churning out wafers at full capacity. They will need to add the MBE. I actually looked at Riber's financials to see what they get for one of their MBEs and it either costs only $1M each, or that's Riber's profit.

Peanuts.

5) POET will enter the market gradually so that the infrastructure for a POET ecosystem can evolve

I don't think this will be a smooth road, however. Fabs best set up to do POET will be able to profit from a lack of readiness in the whole market.