It’s probably more like replacing one big server with smaller servers where it’s now cheaper to overprovision some for standby/failover, so you don’t have to fix them as they fail.