Quantum hardware used for portfolio optimization

A recent social-media post circulated examples of quantum portfolio-optimization runs on real hardware, reviving a familiar claim in quantum computing: finance is one of the sectors where small experimental systems can be tested today. The underlying material does support that narrower point. Published papers, vendor documentation and open tutorials show researchers have run simplified portfolio-allocation and risk-style problems on current quantum systems, though at small scale and usually in hybrid workflows that still rely heavily on classical computation. (quantum.cloud.ibm.com) The experiments do not show a production-ready replacement for conventional portfolio construction. (jpmorganchase.com) They show that researchers can encode reduced versions of allocation problems into forms such as QUBO or Ising models, send parts of the computation to quantum hardware, and compare the outputs with classical baselines. ### What exactly was run on quantum hardware? JPMorganChase researchers Romina Yalovetzky, Dylan Herman and Marco Pistoia said in a September 10, 2024 post that they applied a hybrid version of the HHL linear-systems algorithm, called Hybrid HHL++, to small-scale portfolio-optimization problems on Quantinuum System Model H-series trapped-ion quantum computers. (jpmorganchase.com) They said the work aimed to bridge the gap between theoretical algorithms and circuits that can run on today’s devices. IBM Quantum documentation also describes a “Quantum Portfolio Optimizer” function that maps a dynamic portfolio problem into a QUBO, transforms it into an Ising Hamiltonian, and runs a Variational Quantum Eigensolver on quantum hardware with noise-aware post-processing. (qiskit-community.github.io) IBM says a sample case with seven assets, four time steps and four resolution qubits requires 112 qubits. ### What kind of finance problem are these demos solving? Qiskit Finance’s portfolio-optimization tutorial frames the task as selecting assets while balancing expected return against risk, then solving the resulting discrete optimization problem with quantum methods. (jpmorganchase.com) The tutorial also points readers to prior real-hardware experiments, including work on improving variational optimization with CVaR. A 2025 paper on a “real-world test” of portfolio optimization with quantum annealing said the dataset came from a real problem already used in production by Raiffeisen Bank International, but the experiment itself was still an optimization testbed built around a QUBO formulation. (quantum.cloud.ibm.com) That paper described the work as a collaboration between RBI and Reply. ### Are these full portfolio managers, or just reduced lab benchmarks? Nature Scientific Reports authors in a 2023 paper on best practices said results on different real quantum devices were obtained only for a small-sized example, and linked solution quality to processor size. That is consistent with the broader literature, which treats current hardware demonstrations as limited benchmarks rather than large investable universes. (qiskit-community.github.io) An arXiv benchmark published in 2025 said the literature had lacked an extensive comparison of quantum approaches against state-of-the-art classical methods on meaningful real-world instances. That framing reflects the field’s current posture: benchmarking first, claims later. (link.springer.com) ### Why do portfolio problems keep showing up in quantum finance? Portfolio selection is attractive because it can be rewritten as a constrained combinatorial problem. That lets researchers encode choices such as which assets to hold, how many to include, or how to trade off return and covariance risk into binary variables that fit quantum-optimization routines. (nature.com) Hybrid methods dominate because current machines are noisy and small. JPMorganChase said current hardware constrains the scale of demonstrations, while IBM’s documentation describes explicit preprocessing and post-processing around the quantum step. (arxiv.org) ### Where can readers check the code and benchmarks themselves? IBM’s Qiskit Finance tutorial remains a public starting point for portfolio-optimization code examples, and IBM Quantum’s documentation now includes a separate portfolio-optimizer function for dynamic cases. JPMorganChase’s 2024 write-up links its hardware demonstration to a published paper in Nature Scientific Reports. (qiskit-community.github.io) Recent benchmark papers are still adding comparisons across QAOA, annealing and other methods, with newer studies in 2025 examining noisy simulations and real-world instance sets. Those papers, along with vendor tutorials, are the next places to watch for updated hardware results and larger test cases. (jpmorganchase.com) (arxiv.org) (qiskit-community.github.io)