I don't think Nvidia cares much about the mid or lower tier of GPUs, especially consumer GPUs. Sure they manufacture them to stay competitive in that area, but I don't think they care too much about it.

Their core revenue comes from - in decreasing order - really expensive data center GPUs, somewhat expensive workstation / enterprise GPUs, and finally, expensive top-of-the-line consumer GPUs like the 3090 / 4090 (and upcoming rumored 5090) which they can mark up higher than the lower tiers.

I'm sure that Nvidia knows that even though consumers / gamers will complain, the companies and hobbyists buying them will have no choice and be willing to pay for that power and extra performance and VRAM, whether it be for creatives (3D or game creators) or AI / ML / LLM hobbyists trying to GenAI at home.

It does matter a bit re: the affordability of the used market.

To be fair I think they're delaying it primarily because any new memory increase would make it less likely that Chinese consumers buy the memory cut version of it which would inevitably be similar in performance to the generation they already have available, at least right now they have a good enough product over there that Chinese consumers can keep buying.

I personally (not in China) bought the 4080 Super and honestly regret the investment already, right now it's a pretty bad time to be spending money with the upcoming recession, especially for products not good enough to run your own competitively performing local LLMs on