People are getting ahead of themselves.

We haven't seen if the mid-gen upgrades have had any effect on next-gen uptake yet.

If MS rebrands the X as XB2 with backwards and forwards compatibility, does it still count as next gen?

In the Xbox One X's case it was also a product of ditching the ESRAM which took up a massive percentage of the chips transistor budget.
Xbox One @ 28nm was 363mm2.
Xbox One S dropped that to 240mm2 @ 16nm.
Xbox One X brought it back up to 359mm2 @ 16nm.

Keep in mind that 16nm is derived from a 20nm process, so it's more like a half node shrink from 28nm rather than a full one.
Next Gen will start off with 7nm (Which isn't a true 7nm process, more like 10nm Intel equivalent) and then probably get a half step to 5nm, which should be market ready in just a few years with some luck.

"nm" is more or less just advertising fluff.
5nm from one company could be roughly equivalent to another company's 7nm, you need to compare the individual geometries to see where things truly stack up to get an accurate representation.

You only upgrade whenever the hell you feel like it.
There are Radeon 7970 users who still game just fine.

Or they might start off with "12nm" which is just a refined 14/16nm process.

It's upper mid-range.
The 1060 is squarely mid-range.
The 1080/1080Ti are high-end.
Titan is Enthusiast.

There is actually another way we can go.

It's working for NAND, basically they started producing chips on a larger fabrication process and just went the stacked method.

It's pretty much the standard now.

Oh I didn't know that. Always suspected though with the 16nm thing as I would have expected a full shrink of 28nm to be 14nm. 

But now the 7nm thing is totally new to me. Isn't that flat out lying to market a 10nm "half shrink" as a full on 7nm process? 

Yes, if they do it like in this generation I will have no problem with it.

The next Xbox should be branded in the following format Xbox ꝏ

For years companies have been fluffing up their fab shrinks, they do it for various reasons.
Intels 14nm process for instance is vastly superior to the 14nm from Samsung/Global Foundries... Yet many people think they are somehow directly comparable.
To be fair, even Intel's "nm" naming scheme has deviated somewhat from the normal naming conventions, it's just less extreme than other foundries.

5nm though from TSMC/Samsung/Global Foundries (If they decide to go for it again) should actually be closer to 7nm.

So we still have a long time before electromigration/quantum tunneling becomes a real issue... And I think before we hit that wall we might go backwards for awhile... And start building chips on larger nodes and just stack chips before shrinking again, worked for NAND.

AMD at the moment though is taking the option of building multiple smaller chips (which assists with yields) and throwing everything that doesn't scale down in nm very well onto an older, cheaper process, not everything needs to be on 7nm, but they can only take that so far as well.