Well, I dare to predict that POET will, once introduced at 40nm, scale down much faster than Silicon.

First reason:
The work is already done for Si, so to adopt all this for GaAs will take much less time.

But - possibly more important:
The leading Si-Industry (GF, TSCM, Intel) is just building up the infrastructure for 17 and 14nm. A multi Billion investment! As soon as POET is out (not only PET), this Si-technology will be outdated. It will possibly be similar fast and energy efficiant like PET, but missing the optical and the analog part. So these production lines will be very expansive - "dead" capital. My prediction is that they will do whateveer possible to adopt POET to thee smaller node sizes as fast as possible.
And - well, POET Technologies will ask for their share...