Introduction
Asset management converts mandates, beliefs and market data into portfolios subject to return objectives, risk budgets and operating constraints. The familiar mean-variance framework begins with expected returns, a covariance matrix and a trade-off between return and variance [1]. Institutional implementation adds cardinality limits, minimum and maximum holdings, transaction costs, turnover, liquidity, tax, regulatory, concentration and multi-period constraints. These additions can produce mixed-integer, nonlinear or otherwise combinatorial problems whose difficulty grows rapidly with the size and detail of the investable universe.
Quantum computing has attracted attention as a possible route to new optimisation and risk-analysis methods. The relevant evidence spans different technical categories. Some papers propose algorithms and asymptotic results. Others simulate quantum circuits on classical computers. A smaller group executes simplified portfolio problems on physical quantum processors. These categories answer different questions and should be reported separately.
The evidence in this paper uses five labels. Classical production denotes established portfolio mathematics or current operational practice. Quantum hardware demonstration denotes execution on a physical quantum processor. Quantum simulator evidence denotes results produced through classical simulation of quantum circuits. Theoretical result denotes a mathematical or asymptotic result under stated assumptions. Matchpoint scenario analysis denotes an illustrative operating case; it is not observed investment performance, a forecast or an allocation recommendation.
Physical-hardware experiments have become larger and more relevant to finance. A 2025 IBM-Vanguard preprint reports a simplified bond exchange-traded fund construction experiment using 109 qubits and circuits of up to 4,200 gates [9]. A 2026 trapped-ion preprint reports an end-to-end hybrid portfolio-selection pipeline using real market data, with quantum kernels reaching 78 qubits and 1,016 two-qubit gates [10]. Both studies depend materially on classical computation, problem reduction or post-processing. The latter reports that standalone QAOA failed to find the optimum for two larger indices in the tested results [10]. These findings are valuable demonstrations; they do not prove a general production advantage.
The portfolio problem
Production portfolios rarely stop at a continuous mean-variance formulation. A mandate can restrict the number of holdings, set minimum lot sizes, impose sector and issuer caps, limit factor exposures, require minimum liquidity, account for trading costs and taxes, and constrain turnover from the current portfolio. Some constraints preserve convexity. Others introduce binary variables, non-convex objectives or discontinuities.
A useful quantum programme therefore begins with a problem inventory. It records variables, objective terms, constraint types, data dependencies, required precision and the accepted classical solution method. A large convex problem may remain tractable using a classical optimiser, while a smaller discrete problem with tight interacting constraints may require extensive branch-and-bound, decomposition or heuristic search. The precise formulation determines the computational question.
What the Evidence Establishes
Algorithms and present constraints
QAOA is a hybrid quantum-classical method proposed for approximate combinatorial optimisation [2]. A quantum circuit produces a distribution of candidate bit strings while a classical optimiser adjusts circuit parameters. For portfolio selection, the bit strings can represent selected assets or discrete positions. Constraint handling can enter through penalties, a constraint-preserving circuit, encoding choices or classical repair.
Conditional Value-at-Risk objectives have improved the search for favourable measured outcomes in tested variational settings, including portfolio instances [3]. Warm-start QAOA uses a classical relaxation to initialise a low-depth quantum workflow [5]. These methods can improve tested instances. Their results depend on the formulation, circuit, device and optimiser, and they do not establish an automatic speed-up.
Current hardware performance is sensitive to topology, gate fidelity, circuit depth, shot count, optimiser behaviour and calibration. Weidenfeller et al. examine QAOA scaling with 7 and 27 decision variables on superconducting hardware [7]. Buonaiuto et al. study a four-asset formulation encoded with 12 qubits across simulators and real devices [8]. The small examples provide engineering evidence about design choices. They remain far below a representative institutional universe.
The IBM-Vanguard experiment
Agliardi et al. report a sampling-based variational scheme for a simplified bond ETF portfolio-construction problem [9]. The preprint describes execution on 109 qubits of an IBM Heron processor with circuits reaching 4,200 gates. It reports a relative solution error of 0.49 percent for the tested experiment and uses classical local-search post-processing.
The experiment executes a finance-labelled problem at a larger physical-qubit scale than earlier small examples. Its boundary is important: the portfolio problem is simplified, the reported metric applies to the tested configuration, and classical local search contributes to the final result. The study remains a preprint at the source-review date. It supports a large hardware demonstration for a specific problem; it does not establish general superiority over leading classical portfolio solvers.
The trapped-ion hybrid pipeline
Yalovetzky et al. report a 2026 pipeline using real market data from four equity indices containing up to 225 assets [10]. The hybrid qReduMIS method uses classical graph reductions to simplify or decompose the selection problem, then applies quantum routines to remaining kernels. Reported trapped-ion executions reach 78 qubits and 1,016 two-qubit gates.
The paper reports that standalone QAOA failed to find the optimum for the S&P 100 and Nikkei 225 cases in the reported experiments [10]. The stronger hybrid results rely on material classical reductions. This evidence supports classical problem reduction as part of the present operating model and shows why the contribution of each component should be reported.
Risk algorithms and simulator scale
Quantum amplitude-estimation methods can offer improved asymptotic error scaling under stated assumptions [12]. Woerner and Egger develop methods for expected value, value at risk and conditional value at risk, with a toy model on hardware and a simulated two-asset portfolio [12]. Stamatopoulos et al. study market-risk gradients and provide resource estimates under fault-tolerant assumptions [13]. State preparation, circuit depth, error correction and precision remain decisive resource boundaries.
Soloviev and Krompiec study an encoding that represents multiple portfolio variables per qubit and simulate a problem with more than 250 variables [11]. The reported implementation is a classical simulation of quantum circuits. Real-device execution and noise analysis remain future work. The Matchpoint evidence synthesis is that no broad production quantum advantage in asset management is established by sources [3-14].
A Benchmark-First Hybrid Operating Model
The current practical architecture is hybrid. Classical systems validate data, formulate the problem, create baselines and often reduce the instance. A quantum processor generates or evaluates candidates for a defined subproblem. Classical routines decode, repair and improve those candidates. Independent risk systems recalculate the portfolio. An authorised investment process makes the final decision.
Every stage should produce a controlled artefact. The input stage records data provenance, valuation time and transformations. The formulation stage versions the objective, constraints, penalties and encoding. The baseline stage stores solver configuration and results. The quantum stage stores circuit, compilation, device and sampling evidence. The repair stage records changes to measured candidates. The validation stage recalculates exposures and risk without relying on the experimental code path.
Use-case readiness
Discrete portfolio selection has the richest portfolio-specific experimental base in the reviewed sources [3-11]. It is a reasonable research use case when an institution already faces a hard combinatorial problem and can define a strong classical comparator. Quantum risk algorithms warrant theoretical research and resource estimation. Scenario-rich and multi-period methods warrant targeted exploration. Post-quantum cryptography warrants an active resilience programme.
A falsifiable hypothesis
An experiment should begin with a claim that can fail. A suitable hypothesis states that a named hybrid pipeline will produce feasible candidates for a specified problem family within an agreed objective-gap threshold, decision window and total resource cost. The protocol should pre-register instance selection, holdouts, baselines, metrics, hyperparameter-search limits and stop/go rules.
The classical benchmark should include an exact or bound-producing solver on tractable instances, a strong heuristic on larger instances and the current production process. Warm starts, reductions, repair and local search should be attributed. If a hybrid pipeline uses a classical routine, the benchmark should test the same routine with classical candidate generators where possible.
A minimum scorecard covers objective gap, hard-constraint violations, success probability, dispersion across runs, total wall-clock time, quantum-processing time, classical-processing time, shots, queue time, failure rate and applicable vendor cost. Investment validation adds expected-return assumptions, ex ante risk, turnover, transaction costs, liquidity and exposure stability. Timing begins when an approved input is available and ends when a validated candidate is ready for the authorised decision process.
Governance and diligence
The complete hybrid pipeline should enter the institution's model or decision-support inventory. The record identifies purpose, owner, users, data, theory, limitations, validation, monitoring, change thresholds and contingency arrangements. Independent validation reproduces core results and challenges both the quantum and classical components.
Vendor diligence should establish which stages are quantum and which are classical; which physical processor and calibration window produced a result; which exact solvers and heuristics were used; where data and logs are stored; what access and retention apply; and what queue, cost, portability and exit arrangements exist. Claims should be labelled peer reviewed, preprint, simulator-based or hardware-demonstrated.
Quantum-Safe Readiness
Public-key cryptography supports authentication, secure communications, certificates, software signing and key establishment across the asset-management ecosystem. A sufficiently capable future quantum computer could threaten widely used public-key schemes. The date of such a machine is uncertain. Long-lived confidential information creates a present concern because encrypted data can be retained for later attack.
BIS Papers No 149 identifies quantum-related cyber preparation as a current financial-system issue [14]. NIST finalised its first three post-quantum standards in 2024 [15]. FIPS 203 specifies ML-KEM for key establishment. FIPS 204 specifies ML-DSA for digital signatures. FIPS 205 specifies SLH-DSA, a stateless hash-based signature scheme. These are classical algorithms designed for resistance to known quantum attacks.
Cryptographic inventory
The inventory should identify cryptography in applications, infrastructure, network protocols, certificates, code-signing, hardware-security modules, backups, archives and third-party integrations. Each item records algorithm, key size, protocol, library, owner, data classification, confidentiality horizon, replacement path and external dependency.
Priority can be based on data longevity, business criticality, internet exposure and migration lead time. Long-lived client records, investment research, legal agreements and authentication roots deserve early attention. Asset managers depend on custodians, administrators, banks, market-data providers, cloud platforms, identity systems and software suppliers. Procurement and architecture processes should request each critical supplier's inventory, supported algorithms, milestones and fallback arrangements.
Matchpoint readiness roadmap
This roadmap is Matchpoint scenario analysis. It is an illustrative programme structure, not observed industry performance, a budget or evidence that a quantum workflow will outperform a classical system. The programme begins with a portfolio-problem inventory, accepted classical baselines and a cryptographic dependency map. A reproducible laboratory phase follows, with holdout results and independent quantitative review. A qualifying method can then enter a non-production shadow workflow before any governed decision-support role is considered.
Production consideration requires approval under the institution's model, technology, operational-risk, security and investment-governance processes. The initial role should remain bounded, reversible and monitored. The classical workflow remains an approved fallback. A programme can pause at any phase, retaining better formulations, stronger baselines, improved data lineage and clearer cryptographic dependencies as independently useful outputs.
Conclusion
Quantum computing has produced credible experiments relevant to portfolio optimisation. The evidence includes small real-device studies, larger simulator formulations, a 109-qubit IBM-Vanguard hardware experiment and a trapped-ion hybrid pipeline using real market data [8-11]. These results show technical progress and clarify the role of circuit design, device properties, classical reduction and post-processing.
The reviewed evidence does not establish broad production quantum advantage in asset management. The appropriate present programme is benchmark-first, hybrid and governed. It begins with a decision-relevant formulation and a strong classical comparator. It labels theory, simulation and hardware evidence separately. It measures feasible solution quality, reliability, total resources and control burden. It retains deterministic risk validation and accountable human approval.
Post-quantum cryptography warrants parallel action. NIST's final standards and BIS analysis support cryptographic inventory, supplier engagement and crypto-agility work now [14-15]. This resilience programme can advance independently of the timetable for useful quantum portfolio optimisation.
For asset managers, readiness is an evidence capability. A disciplined organisation can identify where quantum methods deserve testing, stop weak experiments, preserve valuable benchmarks and respond coherently as the technology changes.
