16nm FinFET - finally! This will be exciting. Considering how far Nvidia and AMD have pushed performance on the same 28nm node for the past many years, jumping up a full node (instead of the usual half-node) should allow for something truly special next gen.
That said, I'm not so sure we will all be dancing in this wonderland of cheap, plentiful 17-billion transistored 1080ti cards in Spring 2015, as that story implies.
1. The "full fat" 17 billion transistor model is probably going to be in limited supply 1H 2016, which means either Quadro/compute only, or a small supply of very expensive flagship consumer cards (i.e. Titan Y)
2. This limited supply will be compounded by availability of HBM2. If Hynix continues to be the only supplier, this could be trouble. It is a JEDEC standard so others should be able to make it, but Hynix is a co-developer and the only one with actual manufacturing experience, so I think they will be the major supplier or more like only supplier in 1H 2016. This means that unless Hynix are planning a major ramp up, and starting that ramp soon, Nvidia will be competing with AMD for supply in 1H 2016.
Just speculation at this point, but I hope there's another GPU in the wings, a "GP204" - i.e. the "1070 ti" equivalent of the 970 ti, rather than what will probably be the "1080 ti" model to replace the current 980 ti. That's where we will most likely see the new enthusiast sweet spot from Nvidia.