I shall try to play devil's advocate.

awk variables have no sigils, so the code appears cleaner - there are downsides to that, for example it cannot support interpolation (same problem with Python).
It's internal loop when reading text files is faster than using <ARGV>, however there is no significant difference when perl is used with -p (at least, not with the benchmarks I have run).
Awk's system footprint is smaller, so on overloaded systems it might perform better on simple tasks.

Having done quite a lot with awk in the past I can only suggest that we thank it for giving Larry the inspiration to write Perl, then move on.