>> How stupid is that? Let's bank something that continuously decrease in value as it sits on a shelf and increases inventories to boot. <<
Remember the original plan was that K10 server parts would be in volume production in q3 '07. So a shutdown of Fab 30 at year-end '07 would leave plenty of overlap. problem is that K10 servers won't be available until q2 or q3 '08. fab 30 still has to be shut down, otherwise no 45 nm conversion. So what to do ?
The original sin here was decision to skip 65 nm opteron shrink, and go directly to K10.