Let's play with some numbers, shall we? Please fill in the blanks.
A ridiculously long post, and it's filled with assumptions, inaccuracies and unanswerable speculative questions. How long is a piece of string?
-100PH by when? Who determined that? Do you have that in writing and are you considering the possibility of a plateau in difficulty rises? So that is a guess with a large margin of error which you fail to note. Total hash could be 75PH by the time the farm opens so reducing the required capacity of our farm by 25% - future hash rate it is an unknown figure.
-The 4U expansion units are not 1TH each, there are 6 expansion cases and one master case for our full-sized 55nm-based 10.488TH/s rig giving a figure of 1.498TH/s per unit. You are such an expert (proved substantively wrong twice already today) but you couldn't work that out with
available figures. So your guestimate on this figure is 50% out - because you didn't do the research.
-So we knock 33% off your 16k boards figure to give 10,666 boards.
Suddenly outlay and workload and investment in resources is reduced by a third - and that is fact not a guestimate.
-Wafer numbers - the UMC foundry we are using has a capacity of 45,000 12inch wafers per month. I can't find details on how many chips per wafer UMC will be fabbing.
-The rest of your post is speculation. What are my time estimates on 10k boards? I don't have any, I'm not a PCB fabricator you would need to ask Ken's engineering team that one. Certainly your knowledge of the industry seem a bit sub-par so with respect I'll trust Ken's expert team on their estimates for this project over yours.