That is true -- all we can show is sustained exponential improvement in figures of merit like qubit lifetime and coherence, scalability, gate fidelity over the

That is true -- all we can show is sustained exponential improvement in figures of merit like qubit lifetime and coherence, scalability, gate fidelity over the span of 25 years. As long as this exponential pace of improvement does not stop (we are far from fundamental limits) we should be able to build these machines. Then there is the question of what they can be used for. Their applicability probably will remain niche, as only very special problems are solvable better on a quantum computer than on a classical one.