numeric precision when raising one numeric to another.

It appears from checking the output of exponentiation of one numeric to
another, the output is actually in floating point. Is this normal and /
or expected?

Now, given that

create table test2 (i1 numeric(20,0), i2 numeric(20,0));
insert into test values (123456789012345,123456789012345);
select i1*i2 from test2;
gives:
?column?
-------------------------------
15241578753238669120562399025

it seems odd that

create table test (i1 numeric(20,0), i2 numeric(20,0));
insert into test values (2,55);
select i1^i2 from test;
gives:
?column?
---------------------
3.6028797018964e+16

Now, I can get an exact answer if I'm willing to twiddle with breaking
the exponent down:

select (2^60)::numeric;

Gives:
numeric
---------------------
1152921504606850000

While, select (2^30)::numeric*(2^30)::numeric;

Gives:
?column?
---------------------
1152921504606846976

So, numeric can hold the value, but it looks like the exponent math is
converting it to float.

I'm not bothered too much by it, as I don't really work with numbers
that big. I was mainly wondering if this is kosher is all.