Voice chat isn't inherently costly to host, usually the server acts as a broker and then people connect P2P. Do you remember jitsi? It was used as an alternative to zoom during covid, it was a webpage like Google meet where you opened a room, and people joined with the link. Actually under the good the first one that joined the room acted as a host so all voice communications never reached jitsi's servers.

Jitsi still exists btw, although I understand that a more integrated solution like matrix is preferable. Matrix IS the federated discord-like solution, although last time I read about it it had some issues with scalability and the typical foss Dev drama.