Scientific Computing has an article from Eric Bender, MIT, that is worth reading:

Silicon Photonics Meets the Foundry

What I like most is the following:

• The article highlights silicon photonics' main problem: The light source has to be manufactured separately from different materials and it has to be mounted on the chip in an expensive production step. "In the long run, everything should be monolithic," says Lionel Kimerling. So true – and so far away! What Kimerling doesn't know: Geoff Taylor has done that already, though not with silicon. As POET shareholders (should) know, silicon cannot emit light, but gallium arsenide can.

• Rajeev Ram wants to use standard foundries to manufacture the photonics his team is developing without costly modifications of those foundries. He is not the first with this idea. Geoff Taylor had this in mind when he developed the POET process.

The article is showing how far POET Technologies is ahead of the curve. While monolithical manufacturing of silicon photonics including a light source is wishful thinking, POET Technologies has developed just that using GaAs and is close to production. I am keen to see the next steps …