So you think the professional market are penny pinchers? In general, that isn't the case IMO. If the card helps them do their jobs, then the price will be higher than a normal consumer based GPU such as a GTX 780. I'll take a wild stab at the cost. This was actually mentioned in the video I posted above by Linus, it's about making more dense GK110 clusters, of course you can add more GK110 based CUDA development oriented cards to a cluster or chassis when there are two GPUs on board. If you're making a GK110 cluster based system for CUDA development, you can use a Titan black. Sure. What if you have 10 pcie slots and want 20 GK110 based cards. What then? Well, the Titan Z allows more densely packed GK110 clusters for professional use. That's what linus mentioned and it makes complete sense.
I'm going to take a wild guess that the professional market and CUDA developers aren't penny pinchers. If nvidia priced it at 3000$, then that's what the professional market will bear.
Now the card obviously isn't for me either. I'm a gamer. I'm not a CUDA developer. I would not buy this card for 2000$. Or 1500$. I would buy a GTX 780, though. That's my card, since it's oriented solely towards gaming. It isn't a hybrid professional / gaming card as the Titan cards are. Anyway, I think nvidia should be the ones worrying about the cost. If they think the scientific market or CUDA developers won't pay 3000$, but they'll pay 5000$ for a K6000 (which has less DP performance than the Titan Z). Well I dunno. Sounds fair to me for the professional market even if i'm not the target for the product.
You can't have both. If they are not penny pinchers, fine. They will choose more expensive professional solution.
If they go for titan blach/Z, then they are penny pinchers.
Also, the 3slot cooler effectively diminishes 2 dies per PCB. You will have to skip some of PCI-e slots due to lack of space between.