First, it is useful to realize that quantum computation can be simulated on a classical computer, and any classical computation can be represented as a circuit.
Interestingly, if a quantum circuit does not have entanglement, it can be simulated efficiently on a classical computer. Due to the absence of entanglement, each qubit in a quantum circuit is independent of any other qubit. Hence each qubit's state can be represented by a 2d complex vector. Mathematically, for a quantum circuit $U$ with no entanglement and an $n$-qubit product state $|\psi\rangle = |\psi_1\rangle \otimes \cdots \otimes |\psi_n\rangle$, we can write:
$$U|\psi\rangle = U_1 |\psi_1\rangle\otimes \cdots \otimes U_n |\psi_n\rangle,$$
where $U_i$ are $2\times2$ complex unitary matrices such that $U_1\otimes \cdots \otimes U_n = U$. Such computation can be done in polynomial time on a classical computer because we just need to do $n$ matrix multiplications. Moreover, it can be done in parallel, and it will take constant time in $n$.
Now, even more peculiarly, certain quantum computation can be efficiently simulated classically even in the presence of entanglement. This is due to the Gottesman–Knill theorem. The theorem states that some highly entangled circuits can be simulated efficiently on a classical computer.
In general, we can represent an $n$-qubit quantum state by a $2^n$ complex unit vector. Of course, generally, we have an exponential overhead when trying to simulate an $n$-qubit quantum system. So while simulating quantum circuits is possible, the classical circuit will take exponentially many resources.