Hi Roy--thanks for the post. It has been an interesting few years for Silicon Photonics. I have the opinion that it is still very early; the M&A and Intel and IBM status notwithstanding. My motivation comes from analyzing a huge amount of the R&D literature both from all the commercial folks but also the much larger group of universities doing improtant work. There are so many possible building blocks for a deeply integrated photnics IC that I think it will be years, possibly 10, before the "known, good, scaleable" choices are ironed out. From a business perspective, it feels to me that the early ins (like Intel, and of course Luxtera) risk having later fabless designs come in with more performance, lower power, etc. Of course that depends on ongoing advancement of open fabs, which is not certain either.
Daryl and Roy, do you feel strongly that Moore's Law is coming to an end (a theme that repeats in your excellent book), or merely slowing? While I tend to agree that this has limited impact on your overall trends forecast for silicon photonics, it is also hard to imagine that CMOS process technology advances will come to a grinding halt; the semiconductor industry has demonstrated remarkable inventiveness. Your prediction that silicon photonics integration will likely occur in switch chips from Broadcom and others by 2020 is incisive; it will be interesting to see how Broadcom acquires the requisite expertise to make this leap, and whether integration of III-V compound semiconductors will be necessary for the light sources.