The domain is the admissible values that you can feed into
gf, but when you feed a number into
gf, you feed it into
f first and the domain of
f is
x⩾0, this will always produce an output
f(x)⩾5 so you can safely feed it into
g and you have the domain of
gf to be
x⩾0.
This answer would change if the range of
f didn't match up with the domain of
g, say you had a value
f(x)=1, then you couldn't feed
f(x)=1 into
g because the domain of
g is
x⩾2, so it cannot accept
1 as an input. you would then have to restrict the domain of
f (and consequently
gf) so that the restriction of
f had a range which is a subset of the domain of
g, fortunately you need to do no such thing here because of the first paragraph.