Results from IonQ Forte: structure matters more than qubit count
We ran two different combinatorial optimisation problems on real IonQ Forte hardware. The results came out in opposite directions, and that is the interesting part.
The first problem was a 24-qubit graph cutting benchmark, the standard MaxCut formulation. QAOA at p=1 with 4,096 shots returned a cut value 46.2% above the classical baseline on this instance. The run is recorded in the IonQ logs as Job 019e94b0.
The second problem was a 25-customer vehicle routing puzzle. We tested four QAOA variants on Forte. None of them beat classical simulated annealing. The classical solver found better tours every time.
Two results, same hardware, same team, same week. So what is going on?
The answer: problem structure matters more than qubit count. Graph cutting has a structure that QAOA exploits well. The cost function counts cut edges, and that aligns with what a parameterised quantum circuit can express in a few layers. Vehicle routing does not have that property at the scales we can run today. Classical routing solvers are mature, the heuristics are decades old, and they handle constraint structures like time windows and capacity limits better than a shallow quantum circuit.
This is not a popular story to tell. Most quantum announcements lead with the wins and quietly drop the losses. We are publishing both because operators making real decisions deserve to know where quantum methods help them and where they do not.
The practical implication for grid planners, fleet operators, and logistics teams: if a vendor tells you "more qubits will solve your problem", ask them which problem and why. The structure of your decision shapes whether quantum buys you anything. For combinatorial siting and selection problems where the search space grows exponentially in candidate count, the structure looks promising. For point-to-point routing at the scales most fleets need today, classical methods remain the right answer.
We built the Pizza Race game to make this visible without needing a quantum background. You can play it in three minutes. Stage 3 is a routing puzzle where classical wins. Stage 4 is a graph cut where quantum wins. Same player, same hardware family, opposite outcomes. That is what the hardware shows.