I am reviewing Dave Lazovsky’s recent interview for a second time because it provides some very relevant insights into the industry and POET’s market. 

https://www.youtube.com/watch?v=GMMGtw7FRM8 

The magnitude of the scale of a single socket win …. I have never seen anything like it.

He talks  about the big 4 that he is working with. US hyperscalers which he obviously can’t name and dodges naming them specifically.

The top five hyperscalers spent $158 billion on data centers in 2022, led by Amazon's $64 billion, Google's $32 billion, and Meta's $31 billion. Microsoft chipped in $25 billion, while Oracle added $6 billion. Just five years earlier, these same five firms spent a combined $42 billion. 

BreadAndCircuses provide a very good picture of CAPEX savings: https://agoracom.com/ir/POETTechnologies/forums/discussion/topics/803981-gpus-and-photonic-fabric/messages/2419665#message 

But to me the biggest long term driver is sustainable energy requirements and costs. Industry is under great pressure to utilize CAI’s approach to move data using POET’s light source. The electricity requirements are not sustainable at current trends. Solar and wind are not going to cut it. Hydroelectric energy is maxed out and energy constrained based on seasonal inflow. There are also issues of location and transmission constraints for bulk production regardless of energy type.

Coal fired generation has to continue being displaced for the health of the planet but eventually so will natural gas. Natural gas has been the go to for controlling variable generation requirements that used to be provided by coal. 

Nuclear generation growth is the only real solution but that generally takes more than a decade to go from drawing board to production and costs are at the highest of the resource mix. 

I had planned to review the NERC 2023 electricity reliability report to observe the planning to deal with AI hyperscale energy demands but the site is unavailable this weekend for mtce.  The following link may provide some good insights.

file:///C:/Users/poet/Downloads/2023-U.S.-Data-Center-Markets.pdf  

It is very clear to me that the big 4 hyperscalers see the CAI solution as a means to allow the current AI growth trends to continue. But the reality is POET actually has the ability to reduce the power demands beyond what CAI has achieved by avoiding the use of the gratings that CAI relies on for surface mounting injection points but that is another complex topic and behind a veil of secrecy.