Yes, you should split it evenly between both, assuming you're using Splunk forwarders to load-balance between the two. You should allocated more storage for unevenness, but this is hard to say. For example, a server going down, loading of historical or archived data, or writes of a large file all at once (vs appending of a log file over time) will tend to cause unevenness. As the number of servers goes up, unevenness in volume across indexers tends to go down.
