The advantage of scalable digital quantum computers when/if they get built is that they can do certain very specific mathematical operations in polynomial time (instead of exponential for classical computers). This is a fairly imprecise statement (but gives good intuition). It is conjectured without yet being proven (still, there are many clues to it being true). It applies only to very specific problems, so there are "classically easy" problems which we will just continue solving with classical computers, and there are many problems that are infeasibly difficult for both classical and quantum computers.

An example of where quantum computers will excel is simulating large organic molecules (which is impossible today) ushering a new era of drug discovery and material science.

While there are toy applications of small computers, you probably need hundreds or thousands logical qubits. Given the low quality of the hardware and the environmental noise it is expected that you will need to use error correction codes where you encode a single logical qubit on top of hundreds or thousands of physical qubits.

In other words, once we have a machine with 10000 physical qubits that can all be well controlled and entangled with each other, with gate fidelities of 0.99 nobody (almost nobody) will be able to argue that it is not a quantum computer.

This device here has 20ish qubits with gate fidelities of 0.8 to 0.98.