Gen5 since late March.....

Infrastructure is part of it, but I think it's really a stopgap to alleviate what exists now: people attempting to stream HD movies, subscribed to cable-cutting technologies that are too high a rate, incompatible cameras, latency-dependent VPN, or running any other kind of system-intense technologies that are good enough for terrestrial but will bring limited satellite resources to its knees. [Edit: Not to mention all the IoT stuff like smart scales, refrigerators, lighting systems, door locks, microwaves, washer/dryers... all of these things that load down a particular user's bandwidth, let alone combined on a limited beam]

We did the math on this several weeks ago. It only takes a relatively few people (in the thousands, iirc) on any one beam to use any of these at any one time to drag everyone on it into the dirt.

So beam capacity is only one part of the problem that might level the load across a beam better rather than compressing it amongst fewer servers and outroutes. You still need to level those resources across the beams routed to a particular gateway (or even intra-gateway, which would certainly increase latency). This way you can pull in resources from underutilized beams (and gateways), and dynamically make full use of the satellite's capacity across the userbase, focusing it where it's needed at any given time.

This all assumes the bottleneck isn't the beam itself (which I really doubt), and lies mostly in the processing required at the ground stations.