Hyperscaler operated MECs built at Teleco stations were supposed to be a revolution for low latency compute - gaming, metaverse etc. But this didn't materialize and a lot of telecos grandfathered the service. However the same system could be used for edge AI inference. Are there any major barriers for this? Anyone with first hand experience?