What China semi is smart enough to do, and US semi stupid enough to fall for, is to provide the world with the highest value engineer in terms of number and value/cost. The US semi company then employs China engineering for the obvious reason. The effect is the long term scientific/technical/engineering value of that US semi company being exported to China.
My point was about what it will take for the US to remain competitive in the semi business - it must have fabs. Your point is that Chinese labor is cheaper than US labor, including engineering. With it's large and educated work force, China has a strong card to play and it is no doubt unwise to try to beat them by using sweatshop labor. A better approach is to manufacture with fewer lowskilled workers and more automated/mechanized manufacturing.
OK so not undertilization as a conscious decision one makes when selling the output would become difficult or detrimental. What you seem to be suggesting is that they would not reach more than 50% theoretical/planned capacity and got to ask, how. As I see it, they would begin with some 5k wafer starts per month. get to yield, then move it into high volume and ramp. Where do things go so wrong and why shouldn't we call it "inability to ramp". Ofc 50% would mean that they ramp the entire thing and depreciation kicks in when they aren't ready to ramp but what would pressure them into that irrational rush? Maybe huge delays but they seem ahead of schedule so far, according to these rumors. Anyway, the claim that they will have 64L this year is interesting, would suggest that they might start high volume production in second half 2019 and have substantial output in 2020- any idea how many bits per cell, from a cost prespective it would be interesting if they had viable QLC from the start.
No one is suggesting companies should run a memory fab at 50% utilization. They are suggesting utilization may turn out to be poor on the first phase of a project, discouraging owners from building subsequent phases.
Second time you print that quote about running a memory fab at 50% utilization and ,got to ask, why would anyone do that? Under what circumstances does it make any kind of financial sense to run at 50%? Selling the output is not an issue with mainstream memory, underutilization would only increase losses with zero benefits.