This comment by Suresh that they are early with the 1.6T modules, as differentiated from the 100G, 400G offerings that never really led to significant income is very key.

POET was coming with a "me too" product that was incrementally better in several ways, including price, but the customers already had "known good parts." That's a hard way to get into the business.

Now POET has 800G modules with very little competition and a route to 1.6T and 3.2T assured; that is a very different conversation. They've also proven that the OI allows very rapid customization for the burgeoning need for copackaged optics. CAI is representative. There will be others.

Much as we grumble, POET has been around a long time, showing up at OFC and the others, doing NREs with significant players, and having membership with the standards body and SHINE in Singapore. 

All that work is paying off, with Luxshare and Foxconn joining the party. So, all that earlier effort did not yield fruit, but in a way, it has, by laying the foundation for the next few crucial product releases. The AI server business cannot wait for some major R&D effort. They need this stuff now or on their 15-month horizon. The risk is high for those hyperscalers who cannot provision the AI needs. POET is in a very enviable position. It has been painful but necessary.