Optimization is widely expected to be one of the first areas where quantum computing earns its keep in business. Finance, logistics and AI all depend on choosing the best option from an astronomically large set of possibilities, and those problems strain even the largest classical computers. Quantum machines approach them differently, which is why banks, carmakers and airlines have been running experiments for years.
It is also the area where expectations most often run ahead of evidence. Quantum computers do not simply try every answer at once, and as of 2025 there is no clear, independently confirmed case of a quantum computer beating the best classical methods on a real business optimization problem. This article explains why optimization is so hard, what quantum approaches offer, where companies are experimenting, and how to engage sensibly today.
Why some optimization problems are “unsolvable”
Many business decisions are combinatorial. A delivery route with 60 stops has more possible orderings than there are atoms in the observable universe. A portfolio manager choosing among thousands of assets under dozens of constraints, or a factory scheduling hundreds of jobs across shared machines, faces the same explosion. Checking every option is impossible, and for many of these problems no efficient exact algorithm is known.
In practice, organizations do not solve these problems perfectly. They use heuristics, approximation algorithms and commercial solvers that find very good answers in reasonable time. The opportunity for any new technology, quantum included, is to find better answers, find them faster, or handle larger and more constrained versions of the problem. Even a one or two percent improvement in fuel use, capital efficiency or machine utilization can be worth a great deal at scale.
What quantum computers bring to optimization
A qubit can exist in a superposition of 0 and 1, and qubits can be entangled so that their states are correlated. That lets a quantum computer represent a vast space of possible solutions at once. The popular claim that it therefore “tests every answer in parallel” is misleading, though: measuring a quantum system returns a single result, so algorithms must use interference to make good answers more likely than bad ones. That only works for problems with the right structure.
Three approaches dominate optimization work:
- Quantum annealing. Specialized machines, most notably from D-Wave, encode a problem as an energy landscape and let the system settle toward low-energy states that correspond to good solutions.
- Variational algorithms such as QAOA. The Quantum Approximate Optimization Algorithm runs on gate-based quantum computers, with a classical computer tuning the quantum circuit’s parameters in a loop.
- Quantum-inspired classical algorithms. Techniques developed by studying quantum methods, run on ordinary hardware. They are often the most practical near-term result of quantum research.
The theoretical ceiling matters too. For general search, Grover’s algorithm offers a quadratic speedup, not an exponential one, and most researchers do not expect quantum computers to solve hard optimization problems exactly in polynomial time. The realistic hope is better approximate answers for specific problem types on future, error-corrected machines.
Finance, logistics and AI: where companies are experimenting
Finance. Portfolio optimization, risk modeling and derivative pricing are popular targets. Monte Carlo simulation, used heavily in risk, has a known quadratic quantum speedup through amplitude estimation, though it requires more capable hardware than exists today. JPMorgan Chase has one of the most active quantum research teams in banking and, in 2025, published work with Quantinuum on certified randomness generated by a quantum computer.
Logistics. Vehicle routing, scheduling and supply chain design map naturally onto optimization formulations. Volkswagen ran an early pilot with D-Wave in 2019 to route buses in Lisbon, and airlines, ports and manufacturers have tested quantum and hybrid solvers for scheduling. Most results so far show feasibility rather than a consistent advantage over top classical solvers.
AI and data. Quantum machine learning is an active research field, covering feature selection, sampling and some training subroutines. The honest status is early: claims of faster training remain largely theoretical, and loading large classical datasets into a quantum computer is itself a major bottleneck.
Hybrid quantum-classical workflows are the realistic path
Quantum computers will not replace classical ones. The practical model is hybrid: classical systems handle data preparation, problem decomposition and most of the computation, and a quantum processor is called for the specific subproblem where it might help. Cloud platforms from IBM, Amazon, Microsoft and D-Wave already offer this kind of access, so experimentation does not require buying hardware.
Because hardware is still limited and error-prone, the value of experimenting now is mostly in learning which of your problems have the right structure, building formulations and benchmarks, and developing skills. Those assets carry forward as machines improve. For the broader state of the hardware, see Quantum Computing: The Next Frontier of Technology.
How to explore quantum optimization sensibly
- Start with the classical baseline. Make sure you are already using modern solvers and good data. Many organizations find large gains here before quantum is relevant.
- Pick one well-defined problem. Choose a recurring optimization with clear cost metrics, such as a routing, scheduling or allocation task.
- Benchmark honestly. Compare quantum or hybrid results against the best classical method on the same problem, including time and cost, not against a weak baseline.
- Use cloud access and partners. Rent time on quantum hardware and simulators rather than investing in infrastructure.
- Treat it as R&D. Budget it as capability building with a defined review point, not as a production system with a promised return.
The trade-off is opportunity cost. Money spent on quantum pilots could go into classical optimization or AI that pays back now. A small, well-scoped exploration usually strikes the right balance. Our AI consulting services help organizations improve classical optimization and assess where emerging approaches might fit. Quantum’s security implications are a separate and more urgent matter, covered in Quantum Computing vs. Cryptography.
Frequently asked questions
Has quantum computing beaten classical computers on real optimization problems?
Not convincingly as of 2025. There are promising experiments and some claims of advantage on specially constructed problems, but no widely accepted demonstration of beating the best classical solvers on a practical business optimization task.
When might quantum optimization become useful?
Estimates vary widely. Many experts expect clearer advantages to depend on fault-tolerant, error-corrected machines, which major vendors are targeting toward the end of the decade. Timelines should be treated as uncertain.
Do we need quantum experts on staff to experiment?
Not initially. Cloud platforms and software development kits lower the barrier, and partners can help formulate problems. What you do need is someone who understands your optimization problem and can judge results against a strong classical baseline.
Solving hard problems today and tomorrow
Delana Technologies helps organizations get more from classical optimization and AI now, and evaluate quantum approaches with honest benchmarks when the time is right. To discuss an optimization challenge, call 239.414.5126 or contact us.
Sources: Farhi, Goldstone and Gutmann, “A Quantum Approximate Optimization Algorithm” (2014); Grover, “A fast quantum mechanical algorithm for database search” (1996); Volkswagen and D-Wave announcements on the Lisbon bus routing pilot (2019); JPMorgan Chase and Quantinuum, certified randomness research published in Nature (2025); IBM, Amazon Braket, Azure Quantum and D-Wave Leap documentation.