Most of our databases are multi-tenant, so a single cluster will handle all customer data. The exception is Redshift, which has a separate cluster for each customer since we allow them to have direct access to it (https://amplitude.com/blog/2015/06/05/optimizing-redshift-pe...).