just cheat and make sure the input doesn't have more than one per line :)

tr ' ' '\n' < /tmp/hh | grep -c hello
[download]

alternatives:

plain, explicit count (leverage s/// to provide # per line):

perl -lne '$ct += s/hello//g; END {print $ct}' /tmp/hh
[download]

Same technique as the tr/grep above, just in perl flavor:

perl -0040 -ne 'END {print "$x\n"} $x++ if /hello/' /tmp/hh
[download]

Fun w/perlrun:

perl -Fhello -0777 -lane 'print $#F' /tmp/hh

# shorter:
perl -apFhello -0777e '$_=$#F'  < /tmp/hh

# with check for word boundry (e.g. to exclude "Othello" or mispelled 
+"cello")
perl -apF'\bhello\b' -0777e '$_=$#F'  < /tmp/hh
[download]