This is the biggest problem I see, even with Fermi coming in, there are really only two options, price match AMD and make smeg all on every card, or charge what the card needs to make a profit and have very low volume overall. The problem is their past 3 gens it would be normal to be selling Fermi and the 260-285gtx as their new midrange, at a profit with high volume. That segment is just not there now and likely not for a long time. They've already tried and failed to shrink it to a crap 40nm process.
You can't really see a return to a decent income until the midrange Fermi is ready, but Nvidia's game for the past 3 years has to not make a midrange part, so who knows if theres a Fermi midrange, or when it would be ready.
A midrange fermi would be able to cut size, mem bus, pin out, traces, pcb layers, power circuitry which would make it much more competitive, I just can't see it being ready before Q3 now(as Nvidia for quite some time have been releasing mid/low end 6 months after the high end).
Well A3 silicon is supposed to be back very early Jan, which should mean info about if they have done risk production alongside(put in a limited amount of waifers at the same type and pray A3 silicon is good enough, otherwise you throw away millions of dollars of waifers) they might have a few ready to go in Feb but not many. It would still be 2 months after A3 silicon comes back that they can get real supply.
So the thing is, if Nvidia partners make almost nothing on Fermi and desparately need a midrange, when is it coming. The high end part will change fairly little for their partners as I just can't see huge money in it.
Obviously if its 4 times as fast as a 5870 and only costs Nvidia £50 more to make, they can sell loads at £400-500 with ease and make a killing, but I don't think even the hardiest of fanboys thinks thats possible. Would be interesting though, I might have to buy an Nvidia card
