This fails because MATLAB cannot represent an integer larger than 2^53-1 as an exact integer, when working in double precision.
14^27 is far larger than that of course. So it fails. This leaves you with two basic options, to use symbolic arithmetic, (or perhaps my VPI toolbox) or to use powermod.
The powermod solution is better of course, in the sense that the huge number 14^27 never had to be fully computed.