The most tempting, most oversold application
Pick the best combination among zillions: delivery routes, factory schedules, power-grid dispatch, crew rosters. These problems are everywhere, worth billions, and often NP-hard, so "quantum solves optimization" became the industry's favorite pitch. The honest version is narrower.
Quantum approaches (annealing machines built specifically for optimization, and algorithms like QAOA on gate-based hardware) genuinely explore solution spaces differently than classical methods. But decades of brilliant classical work (integer programming, simulated annealing, tabu search) set a brutally high bar, and for realistic industrial problems the best classical solvers still generally win. Grover-style quadratic speedups exist on paper but are modest, and no proven exponential advantage is known for general optimization.
What the pilots actually showed
The documented pilots are real and instructive. Volkswagen ran quantum-assisted traffic-flow routing for buses in Lisbon during the 2019 Web Summit — a genuine live deployment, and also a problem classical computers handle fine; the point was learning the workflow. Manufacturers and logistics firms have run similar scheduling experiments on annealing hardware. Read every one with the question: did it beat the best classical solver on cost or quality? So far the public answer is no: the value has been readiness, not advantage.
That is not a dismissal. Optimization is where hybrid quantum-classical workflows are being engineered and where any future hardware improvement plugs straight into an existing pipeline. It is simply the area with the largest gap between marketing and measured results: keep the evaluating-claims checklist close.
需要说明的是: No general guarantee of practical advantage. The benchmark is the best classical heuristic, not brute force.