About areg command for fixed effects

Hi.
I have a question about areg command for fixed effects.

areg y x1, absorb(x2)
predict x2_fixed, d

Using above command, I can get coefficients (fixed effects) of x2.

If I run this command:

reg y x1 i.x2

the first one is omitted. That is, if x2 is a five categorical variable, the number of estimated coefficients is four.
However, if I use areg and predict command, I can get all five coefficients.

How are they calculated?

Thanks