In a PPI (Partitioned Primary Index) table, data distribution across AMPs is determined by the primary index hash value, not by partition definitions. Partitions control how data is physically stored within each AMP and affect query elimination, but distribution uses the standard primary index mechanism.