This is definitively very interesting and very impressive! Good find, dcmd99!

However, the article is lacking technical details and I am unwilling to pay $$ for the original Nature article. At least we know that the solution is based on silicon photonics. Though the text doesn't include the word "monolithical", I believe it is monolithical, based on the original paper's literature list.

With one very important exception: the laser comes as an external component – typical for silicon photonics. The image Fairchij posted is showing how the laser light enters a chip. Here's another one that visualizes where the light comes from:

Sorry for the bad quality! The image is from the free image preview section of the Nature article.

However, in spite of the bad quality you should be able to see the processor chip on the left hand side and the memory chip on the right. In between and external to the chips there's the optical link between the two. Also external to the chips is the laser that is fueling the optical capabilities of the whole system and feeding the laser light into the two transmitters. This laser is not monolithically integrated.

The POET solution, however, integrates the lasers monolithically and thus has it internal to the chips. This makes the POET solution superior, at least regarding space and costs, and quite likely also in terms of power consumption.

Have a Merry Christmas everyone and a blessed New Year!

Greetings from Germany
Rainer