The important arithmetic operations over finite fields include exponentiation, division, and inversion. An exponentiation operation can be implemented using a series of squaring and multiplication operations using a binary method, while division and inversion can be performed by the iterative application of an AB$^2$ operation. Hence, it is important to develop a fast algorithm and efficient hardware for this operations. In this paper presents new bit-serial architectures for the simultaneous computation of multiplication and squaring operations, and the computation of an $AB^2$ operation over $GF(2^m)$ generated by an irreducible AOP of degree m. The proposed architectures offer a significant improvement in reducing the hardware complexity compared with previous architectures, and can also be used as a kernel circuit for exponentiation, division, and inversion architectures. Furthermore, since the Proposed architectures include regularity and modularity, they can be easily designed on VLSI hardware and used in IC cards.