All of the major partners that come up regularly here (CAI, Mitsubishi, Foxconn, Luxshare) don't do small stuff. These are very large companies that want to be relevant in their segments. They did not announce their POET tie-ups to do a few thousand units. Think millions of modules with POET chips for packaging and networking in these vast data centers.

You can expect that their ambitions for these millions of units will match the outsized deployments of the AMD and NVIDIAs of the world.

They all claim to be working closely with the largest players in the hyperscale data centers, which, according to Dave Lazovsky represent 70% of the data center business. The smaller players will follow.

These large companies are in a desparate race to take their share of the AI market. They have always been competitive, but this is different. They see it as existential, a fight to the death. That may be dramatic, but it is the driver of the need to modernize the data centers beyond the GPU, CPU advances. 

There is talk about building nukes to power these things. That is indicative of the pressure on these companies to scale up and make it work, whatever the cost. 

POET seems to offer a quickly deployable, elegant and more and more proven way to address a large share of the power problems, as well as increasing the effectiveness of the expensive xPUs and HB Memory that are driving the semiconductor sector. 

It's all falling POET's way so far.