We came up with nearly the exact same solution. I believe you can eliminate $val, however, by operating directly on the hash value.

use strict;
use warnings;

my %hash;
my $id;

while( my $line = <DATA> )
{
    chomp $line;

    # if( $line =~ m/^>(.+?) / )  # changed to \S
    if( $line =~ m/^>(\S+)/ )
    {
        $id = $1;
    }
    else
    {
        $hash{$id} .= $line;
    }
}
[download]

ewijaya, if you are trying to read sequence files in fasta format, you might want to look at bioperl's Bio::SeqIO class.

Update: ihb made a good point about the regex. I made the assumption (based on the example data) that a space will always follow the sequence ID, but that may not always be true. Therefore, ihb's regex is a bit safer for that reason (although you should keep in mind that \w will not match '.' or '-', so \S is probably better).