Its understandable- everyone wants their cellphone to complete the journey of the desktop- but the battery is just to bad. Even with a dedicated chip for drawing - open pubg and the batterys power-supply falls like a rock.
What is a game other- then a lightshow- No-Sleep-Bug?
In my opinion, as long as there is demand and physics unaligned, "building it and they will come" will not archive anything.
So why not instead build a miniserver with gfx cards into WiFi and do the computations for AR, VR and games (mostly downstream data) and catch that with a passive WiFi-Reciever? It is still video displaying - but it does not come with the usual computational burden.
I know, that is a bad architectural choice - away from the expected model, that should batterys get good, would triumph any day now. But if you plan in for that eventuality - you can still program architecturally near neutral and you could walk ahead into the future, not just glimpse it while tied to a charger.
https://passivewifi.cs.washington.edu/files/passive_wifi.pdf
https://bib.irb.hr/datoteka/575153.Energy_Consumption_in_And...
https://pdfs.semanticscholar.org/9934/9551d3ef0e0efc3c54aaab...