Are you aware of the requirements to run a Splunk indexer on VMWare?
Splunk has pretty significant resource requirements for indexers to begin with and adhering to the best practices for virtualized indexers outlined in the document linked above is pretty much the only path to happiness.
There are a lot of things an indexer needs memory for. The indexing pipeline is one of them, but indexers are also handling search requests from the search heads. Things like large lookup files can increase your memory footprint.
As long as you don't experience symptoms of a memory leak, where RAM usage is increasing until the VM fails, anything between 1 and16GB is pretty normal for a busy indexer.
