he already produced the maskset and they probably had a few test wafers done.
IIRC he stated, in this thread, that his first chips came from a shared wafer at a university.
Some quantity from a portion of a wafer, not 'a few wafers' as a test run at a commercial fab in preparation to just cranking out commercial amounts from the same fab.
I have no idea what would be involved in moving from a proven test run from a 'shared university wafer' to x number of wafers at commercial fab. But there is bound to be some delay. And risk.
Okay, I missed that, interesting.
Well, universities dont have fabs, so where ever they fab those chips, you could go there and have your asic produced. There are a few possible ways; either you could re-use the shared mask set that only contains a few of your own chips, and the fab will be able to fill an entire wafer with this using a stepper. Not all fabs do this, and the ones that do charge a significant surcharge for it, because its a lot of work per wafer. But for a bitcoin chip this is probably still a good way to go, as volume will be small for an asic project, and wafer costs insignificant.
Another way is to have a new maskset built based on your now-proven design. You would still be tied to the same fab though and it will cost a pretty penny and take even longer, but your per chip costs would be a lot lower.
Im guessing they are going with the first option.