What still worries me is that ramp up time for fabs like these are on the scale of years - if we assume a fab started right now, we'll be making semiconductors in ~2 years at the earliest (very optimistic)? It seems like the root of the issue seems to be the inventory tax and accounting rules - perhaps an exception for the semiconductor industry would help rather than even a single domestic fab (which could be impacted by natural disasters as well).
A 250M$ - 1G$ investment / year is big, but if it's self-sustaining in a couple years, that's a huge win.
TSMC announced the Arizona fab in May of 2020. They estimate it will be at volume production by 2024. The challenge is that it's not just building the fab, but also the lead times for all of the equipment involved, getting the specialists in to install+calibrate, then doing some initial runs to work out issues before going to volume production.
Who will be billed for this negative equity?
TSMC's economists certainly know what they do when they put a $12B investment: this can only happen with US government guarantee, thus taxpayers paying.