This statement about computer chips has nothing to do with foundries, it's something that is immediately obvious if you read the white paper.

The question is not whether one can go out and produce 1 trillion exohash by tomorrow morning.... more of a question of, what happens if in late 2046 there is a fallow period in the AI boom, and someone takes an opportunity and spends $100 million to produce enough chips to wreck the system. You don't need to be expert on VLSI to understand that what is expensive to make today will probably be cheap to make in the future, and also that we can't predict the market swings and geopolitical economics in the future. What if China invades Taiwan?