When you do any initial clone of a table object [no matter the size – it could even be hundreds a table of TBs], there is no storage cost at all. This is mainly because the cloning of an object is a very small sized metadata change. The cloned object references the base objects it was cloned from on the backend by pointing to the original partitions. BUT if you insert, update, and make any changes to the table that was cloned then Snowflake makes corresponding write ahead immuable partitions for EVERY micro-partition that was changed. This is where you can start to incur storage costs. These costs can also be very significant if many of the partitions change frequently AND especially if there is a long time-travel period on the table clone.