What you say about interconnection fees or rather termination rates is nonsense.
First of all mobile termination rates have come down dramatically in the past decade and have been
sitting at $0.017/min since 2015, so that's not a relevant expenditure for an MNO and actually has never been. And that's because, as you might have witnessed yourself, people do not only place outgoing calls, but sometimes they do also receive calls and when they receive calls their MNO is being paid a termination rate by the originating network operator. So the same way MNOs are charged for terminating calls into third-party networks, they are being paid by third parties for terminating calls into their own network.
Unless the ratio of incoming and outgoing calls is asymmetric the expenditure and income from termination fees in principle neutralizes, regardless of the number of subscribers.
However, there's indeed one small exepmation and that's because the termination fee for calls into fixed line networks is lower than into a mobile network. That means that an MNO pays less for its own mobile subscriber calling a fixed line number than the other way around. Hence, based on symmetric call volumes an MNO earns more from receiving calls from a fixed line telephone than the cost of terminating a call from a mobile subscriber to a fixed phone. As a result fewer customers would actually mean less income from terminating call from fixed networks, but again that cost is negligible today.
With regards to data usage the cost of carrying data, fact is that technology dictates the increments in which you can size your network capacity and those increments are very large. When you upgrade a tower's capacity you can't do a 10% upgrade or downgrade but adding an extra carrier frequency (which you do when you run out of capacity) adds at least some 100% capacity. The backhaul to the towers runs either on fibre or microwave links - whether you lay or rent a fibre pair to the next aggregation point or buy a pair of microwave antennas that stuff is sold or leased out by unit and can't be downsized. On the aggregation network you install switches and routers that are always overdimensioned as you anticipate traffic growth for years ahead. Again you buy this stuff by units, not actual load. You connect your aggregation points by leasing capacity on longhaul fibre routes where the DWDM technology dictates the capacity you can buy, typically 10Gbps or 100Gbps. Again these are coarse increments where you can't adapt for 10% more or less load, particularly as you always need reserve capacity. In the core network you need various servers doing all kind of jobs like orchestrating all the the towers and making sure you can seamlessly switch from one tower to the other while moving, keeping a database of subscribers and their locations, billing, etc. All that stuff is again sold by units and has quite some reserve capacity.
IP transit to connect to the Internet is ridiculously cheap in Australia where you can get a 100Gbps upstream for $6000/month and I doubt that Optus see more than 200Gbps peak traffic generated by their 10.5m mobile subscribers. Again this is stuff is leased in huge increments of 10 or 100Gbps and it's cheap.
Apart from the fact that the entire network is built in expectation of traffic growth and usually always has reserve capacity all that suff has already been bought and installed. You can't downsize it (and it wouldn't make sense as data usage is growing steadily) and in fact it's all sunken cost.
The only variable cost where Optus might be able to save a bit if they lost AYS' subscribers is the software licenses for the various elements on the core network but typically you commit to a certain subscriber number, so you can't easily downgrade.
And because of these economics of mobile network where you face massive fixed costs and negligible incremental costs to serve additional subscribers, economies of scale are key. And that's why TPG will be keen to get a deal with AYS which again will be the motivation for Optus to buy AYS in order to avoid losing their economies of scale.
