I always found it a bit of a mystery why POET would build the full transceiver (rather than just the optical engines) which I felt should be done by others and avoid the costs associated with the equipment, floor space, manpower and distribution channels to the end users.  The Intel model of focusing on just the supply the optical engines to transceiver manufactures seemed to be the most logical path.

But I think POET wanted to maintain the option of selling the full blown transceiver as a means to ensure that they would have a marketable end product to sell directly to end use customers. Like a contingency plan to ensure they get good value.  It may provide some incentive for transceiver manufactures to give them fair value for the optical engine because if they don’t POET could build the entire transceiver themselves and undercut the pricing of the transceiver manufactures.

I don’t know if that is part of the equation but in the description of POETs offering the following statement is provided.

Proprietary embedded dielectric waveguides combined with component-level integration and packaging give POET’s line of ROSA, TOSA and TXRX optical engines an extraordinary cost and performance advantage over Silicon Photonics or conventional free-space optics. Initially serving the 100Gbps Ethernet and LAN WDM markets, the form-factor compliant designs are scalable to 200G and 400G. POET’s optical engines offer a compelling and sustainable competitive advantage to transceiver manufacturers and end-users.

No mention of POET building the transceiver but for the first time (to my knowledge) POET makes direct reference to the TXRX optical engine. And at the same time they are talking about end users?

So two interesting facts emerge from POET’s statement.

First, they plan on selling the TXRX optical engine which is the combined ROSA and TOSA with (mux and de-mux on the same substrate).

And secondly they apparently plan to sell directly to end users as well as transceiver manufactures. If we take this literally this suggest that they could be supplying optical engines directly to data center companies. I think there has been a trend of some of the very big companies that are interested in building their own infrastructure to the point of building their own servers. And the target price of $1/G may inspire some to remove the middle man? Speculation but interesting to contemplate.