I'm by no means an expert on the subject, but I want to explore/experiment with an idea I have for the Metaverse: right now, all the computation and load is handled client-side. The client needs powerful CPUs and GPUs to render everything, and it's really difficult to fit that powerful technology onto a small area like AR glasses. In addition, the Metaverse will inevitably expand to thousands to hundreds of thousands of concurrent users all in one place (e.g. an extremely large concert or something), so these tiny CPUs and GPUs will inevitably reach a limit at some point.
What if we moved all this necessary computational power over to the server-side? The client would just send the server certain actions/movements, and then the server would apply these actions/movements to some scene and then send an image back to the client of what they're viewing. Send images fast enough and it becomes a video, so all the client-side would need to optimize is the rendering and displaying of videos, and this scales very well in a scenario where there's much more activity, such as a large concert, as it's much easier to scale server infrastructure since it's not limited by the same size constraint as the client-side technology.
Of course, there are many issues with this approach, such as the need for high data-transfer and extremely-low latency to make the client feel like they are moving in real-time. As a non-expert in these subjects, I don't know how difficult it is to achieve extremely-low latency, but I do know that we've already scaled high data-transfer when it comes to videos: just look at YouTube, which needs to send massive amounts of data to millions of concurrent users! I'm hoping to do some more research and experiments into this area (when I get the time), but for now I thought I'd just share this idea with people who know more about the subject than I do to let me know of the problems with moving all the computation server-side instead of client-side.
Giving everybody access to a bay of extremely powerful GPUs is probably not sustainable—it's possible for a group of people to overload it and ruin access for others. To solve this problem, we could adopt an approach similar to renting machines on AWS, where you "rent" out a GPU on some cloud that would be in charge of rendering your scene. There could be some "free tier" GPU for users that need to access the metaverse for free, and this way, one user or group of users overloading their GPU wouldn't affect the performance of others. In addition, in an event like a huge concert or sports match, the host could rent out a powerful GPU such that all attendees, regardless of how powerful their client-side machine is, would be able to experience the event without lag issues related to rendering (that is, assuming their internet connection speed meets a baseline criteria, which hopefully will be possible with 6G and whatnot in the coming future).