So many excuses in ther it beggars belief-so many games bring made replaying a game is pointless - couldn't giv a cows arse re companies not being able to make more gaming computer hardware end of the day the gamer wants access to play games at the best technology available and if thts a subscription with streaming then so be it - but I'm sure streaming and gaming hardware can live together for the time being - noone can predict wher gaming wil be in 20 years
If you don't want to revisit things then so be it. Others seem to do OK in that though.
Best technology has been a fairly decent path for many but there are also serious perks to alternative routes (see many of Nintendo's efforts since a long time really), and cheapness also has its perks if you don't need to have someone on your team handling nighttime skybox animation. Or if you prefer films have somewhat topped out on cash (or at least students and small countries can afford to compete with Hollywood of the 90s, and it is not like they have gone from strength to strength since then), music long ago did and books probably longer still. Do you not imagine some kind of plateau for games? Indeed I am not seeing the perks of the PSBone over the PS360 at this point in terms of AI, level design/size, story, music and input.
Subscriptions do also have serious downsides (some already noted and others skimmed at best -- constant payment or at least payment that runs out if you buy a time/data block, you can be fairly assured that no second hand options will get given here (or at least not without them kicking and screaming the entire way), far less scope for mods, far less scope for cheats, needs internet connection, needs bandwidth on said connection) and might well be things people avoid.
While I am not sure hardware makers will abandon ship (there are other fields that need hardware) if the massive earner that is gaming dries up then it will change some things. While I would consider game streaming's failure slowing this process a happy perk* it is not enough for me to take active measures just yet.
20 years prediction... we do have roadmaps for the likes of Intel and possibly Nvidia at those kinds of ranges (graphics might be harder but what is done on render farms for pixar today is likely to make it into a card that many years from now, along with the lighting effects), and the tech pathways are also fairly known (what is reserved for uber expensive "this needs to happen, make it so" today will probably give a good idea of what is to come here for consumer tech), I doubt we are going to see infrastructure refreshment do anything radically good. There could of course be some kind of black swan event/discovery that changes things radically (someone might crack memristors, maybe some quantum).
It is probably too soon for a big boy neural interface worth noting here (or at least for it to go widespread enough for a market... wonder if brain interfaces for those that do have them will be considered like using a mouse and keyboard on a console game today). Might be in time for some kind of electroactive polymer gloves though.
Procedural generation might be better, will almost certainly allow for fully voiced games (think every game is bastion if it wants to be) from said generation and quite possibly some odd takes on acting (the face swapping stuff already seen can be expanded a bit) but whether it will suffice for level generation and outright script generation (might be enough to give sidequests a bit more flavour or have inventories populated with many more items doing similar things). What will go for AI might be a bit different.
*the thought of the following future somewhat scares me and it has already been demonstrated today that companies will go there given half the chance and governments and people are spineless enough to allow it to happen. The shift to streaming for games meaning all entertainment is handled there serving to eliminate one of the needs for local computing power.