Quantifying cost saving when migrating a monolithic app to microservices

I don't think migrating a monolith to microservices generally results in direct cost savings, although it's certainly possible in some situations.

If the monolith is currently scaled for peak load, then there's obviously some potential costs savings moving it to elastic infrastructure, where the cost saving is going to be related to the difference between peak and base load. Additional direct savings from moving to microservices could come in if there's any event-based or v.peaky capabilities that can be hived off into microservices, and therefore removing the need to scale the whole monolith infrastructure elastically.

To try and calculate the potential savings in such cases you'd need to work out the difference between base and peak load for the monolith's various capabilities, and then create some indicative figures if there were actual savings in elastically-scaling individual capabilities.

However, it's also more expensive to operate microservices, since there's more moving parts, so my suggestion would be to do those calculations, but be prepared for the move to microservices potentially costing more.