The fallacy in your argument is your premise that the practice itself needs to be sustainable.
When in fact, the practice may lead to a sustainable state that is self-funding and requires no further spinoffs.
Kind of like what you hope is going to happen to SGI in its own market... So you seem to be aware of the concept.
The R&D costs to participate in the hardware market that
SGI does is relatively modest and increases very little
over time. SGI's main problem, up until today at least, was
a bloated corporate structure left over from its Hollywood
darling days. The current reorg is leaving engineering/R&D
essentially unaffected. As they say in the army, a higher
teeth to tail ratio.
OTOH, for a fabbed semico the cost of staying at the poker
table goes up exponentially year after year. With flash
now gone AMD is now just a Futurama style head in a jar
with nothing left to pawn. What's it going to do the next
time it has to ante up? At least a company like IBM makes
loads'a'dough on services, software, and proprietary
systems. Oh yeah, and selling process tech to semico
have-nots like AMD.
