Quantum computing use cases in finance

Where quantum computing could apply in finance

A 2022 survey by researchers at JPMorgan Chase and academic partners groups the candidates into derivative pricing, risk modeling, portfolio optimization, natural language processing and fraud detection. Each group has a different level of public evidence. Natural language processing appears in the survey, but no bank pilot or bank-authored paper on it has been published, so it is left unrated.

Pricing and risk: the resource estimate and the pilots

Derivative pricing often uses Monte Carlo methods, which the Goldman Sachs and IBM authors say consume significant computational resources for financial institutions. A key public resource estimate is a paper by researchers at Goldman Sachs and IBM Quantum, which describes itself as the first complete resource estimate for useful quantum derivative pricing: A Threshold for Quantum Advantage in Derivative Pricing, first posted in December 2020 and published in Quantum in 2021. For two benchmark products, autocallable and target accrual redemption forward derivatives, it estimates 8,000 logical qubits and a T-depth of 54 million, and says advantage would require running the program in about a second. The authors state that these requirements are out of reach of current systems and offer them as a roadmap.

Crédit Agricole CIB ran two proofs of concept with Pasqal and Multiverse Computing, starting in June 2021, on the valuation of financial products and on credit risk. Its January 2023 announcement calls the results conclusive but gives no performance figures, and it notes that Multiverse’s algorithms can run on conventional systems. The announcement does not show a hardware advantage.

Portfolio optimization: small problems, mixed evidence

Choosing assets under constraints is an optimization problem, and banks and asset managers have run hardware experiments on it. Vanguard and IBM tested a variational quantum algorithm on a simplified bond ETF problem, using 109 of the 133 qubits on an IBM Heron processor and circuits of up to 4,200 gates, with a classical local search to refine the samples. The classical solver CPLEX can solve the problem to optimality at that size, and IBM reports that the hybrid workflow beat a purely classical local search, with the gap growing as the problem grew. It does not report beating CPLEX.

In May 2024 JPMorgan Chase, Argonne National Laboratory and Quantinuum reported in Science Advances evidence of an algorithmic speedup for the quantum approximate optimization algorithm (QAOA) on a benchmark known as the LABS problem. The scaling result came from a noiseless simulator on Argonne’s Polaris supercomputer, and the hardware work was small-scale, so the speedup has not been shown on a machine. The release lists financial modeling among possible applications without naming a finance use case.

Trading and prediction: the HSBC and IBM trial

On September 25, 2025, HSBC announced a trial with IBM on over-the-counter bond trading. Using real production-scale data from the European corporate bond market and IBM’s Heron processor, the hybrid approach improved the prediction of whether a customer’s request for a quote would be won at the quoted price by up to 34 percent compared with common classical techniques used in the industry. The accompanying preprint reports up to about 34 percent in out-of-sample test scores, measured by backtesting against models trained on the original data and on noiseless simulations of the quantum transform. The authors suggest hardware noise may contribute to the gain and say this motivates further study.

Fraud detection: simulator results and a warning about false positives

The public evidence on fraud is simulation. The National Quantum Computing Centre announced that Rigetti would lead an Innovate UK project with HSBC, the University of Edinburgh’s Quantum Software Lab and the NQCC, extending quantum anomaly detection models to money laundering. A Physics World feature sponsored by the NQCC (April 16, 2025) reports that the team ran quantum machine learning on a simulator inside HSBC’s own infrastructure, so that no customer data was shared. Rigetti’s project lead says the simulations show the potential of low-depth programs for fraud detection in the near term. He also says the hybrid approach produced more false positives than classical methods. The feature adds that even a fault-tolerant machine would struggle to process a bank’s full transaction data, because each data point takes time to run.

An earlier preprint by Oleksandr Kyriienko of the University of Exeter and Einar Magnusson of HSBC (August 2022) tested quantum kernels for anomaly detection on a public credit card fraud dataset from Kaggle, subsampled to 525 transactions that included 25 fraud cases. In simulations of up to 20 qubits the authors report a quantum-classical separation in average precision of 15 percent at 20 qubits. The paper states that it is not an HSBC product. Neither source reports a run on quantum hardware, and neither describes a deployed fraud system.

Certified randomness: a hardware result with a security use

In March 2025 JPMorgan Chase, Quantinuum, Argonne National Laboratory, Oak Ridge National Laboratory and the University of Texas at Austin published a protocol in Nature for generating randomness that a remote party can verify. A 56-qubit Quantinuum H2 trapped-ion computer acted as an untrusted server, and four supercomputers, including Frontier, checked the results. The authors report at least 71,313 bits of certified entropy. JPMorgan’s post lists lotteries, jury selection, encryption keys and differentially private machine learning as uses, and it does not describe a deployment in banking.

Quantum-safe security: where banks are already acting

The BIS paper Quantum computing and the financial system (October 2024) names two risks: quantum computers breaking widely used cryptography, and sensitive data stored now being decrypted later. NIST released three standards on August 13, 2024: FIPS 203 (ML-KEM), FIPS 204 (ML-DSA) and FIPS 205 (SLH-DSA).

The BIS Innovation Hub’s Project Leap tested the move in two phases. Phase 1 (report of June 5, 2023) sent payment messages between the Bank of France and the Deutsche Bundesbank using a hybrid scheme and found it feasible, with an effect on performance. Phase 2 (report of December 11, 2025) replaced digital signatures in an operational payment system, ran every test scenario successfully, and found that many system components had to change and that post-quantum cryptography takes considerably more processing time. The BIS’s July 2025 roadmap says to start the transition now, with broad awareness and a cryptographic inventory as critical foundations, and says migration may involve cryptographic agility, defense in depth, hybrid models and phased migration.

What the value estimates say, and what they do not

The McKinsey Quantum Technology Monitor, April 2026, estimates that finance use cases could be worth $400 billion to $600 billion by 2035, an average impact of 3.0 to 4.5 percent, excluding “black swan” breakthroughs. The report says the incremental impact of quantum computing overlaps with the impact of generative AI, so the total value at stake is not entirely additive, and it says value depends on multiple factors unique to each company. TQI reads it as an order of magnitude and not a forecast for any one bank.

A different kind of evidence comes from the industry itself. In QED-C’s 2025 market survey, financial services ranked second, after chemistry and materials, among the end-user sectors seen as most attractive for quantum computing suppliers in the near future.

How to read a quantum finance claim

1. What is the baseline? HSBC’s 34 percent compares with common classical techniques used in the industry, and the release does not say they were the best available methods. Vanguard’s result compares with a classical local search, not the exact solver.
2. Is it a best case? “Up to” means the top of a range.
3. Was it run on hardware or a simulator? JPMorgan Chase’s QAOA scaling result came from a noiseless simulator.
4. How big was the problem? Vanguard’s was simplified, and the Goldman Sachs estimate needs about 8,000 logical qubits.
5. Backtest, pilot or production? Every example here is the first or second.
6. Has it been peer reviewed? The HSBC paper is a preprint; the JPMorgan randomness paper and the Science Advances paper are published.

What a financial firm can do now

• Build a cryptographic inventory and plan a phased migration. The BIS says to start the transition now and names the inventory as a critical foundation, and NIST advises starting to integrate its standards immediately.
• List the workloads that are optimization or simulation problems and expensive today, then rank them by size and by whether a classical improvement is cheaper.
• Ask any vendor or adviser for the baseline, the problem size and the hardware behind every claim.

Questions

Sources

1. HSBC demonstrates world’s first-known quantum-enabled algorithmic trading with IBM, HSBC media release, September 25, 2025.
2. Enhanced fill probability estimates in institutional algorithmic bond trading using statistical learning algorithms with quantum computers, arXiv:2509.17715, submitted September 22, 2025 (preprint).
3. On the path towards a quantum economy, Physics World, April 16, 2025 (sponsored by the National Quantum Computing Centre).
4. The NQCC is partnering on the project awarded to Rigetti Computing by Innovate UK, National Quantum Computing Centre (announcement of November 2023).
5. Unsupervised quantum machine learning for fraud detection, Kyriienko and Magnusson, arXiv:2208.01203, submitted August 2, 2022 (preprint).
6. JPMorganChase technology blog post on certified randomness, March 26, 2025.
7. A Threshold for Quantum Advantage in Derivative Pricing, Chakrabarti, Krishnakumar, Mazzola, Stamatopoulos, Woerner and Zeng, arXiv:2012.03819 (v3, May 25, 2021), published in Quantum 5, 463.
8. Portfolio optimization with Vanguard, IBM Quantum blog, September 29, 2025.
9. JPMorgan Chase, Argonne National Laboratory and Quantinuum show theoretical quantum speedup with QAOA, Quantinuum press release, May 29, 2024.
10. Two real-world experiments conducted by Crédit Agricole CIB with quantum computing, Crédit Agricole CIB, January 26, 2023.
11. Quantum computing and the financial system: opportunities and risks, BIS Paper 149, October 9, 2024.
12. Quantum-readiness for the financial system: a roadmap, BIS Paper 158, July 7, 2025.
13. Project Leap, BIS Innovation Hub with the Banque de France and the Deutsche Bundesbank (phase 1 report June 5, 2023; phase 2 report December 11, 2025).
14. NIST releases first 3 finalized post-quantum encryption standards, NIST, August 13, 2024.
15. A Survey of Quantum Computing for Finance, Herman et al., arXiv:2201.02773 (v4, June 27, 2022).
16. State of the Quantum Industry 2025 report, QED-C, 2025 (quantum computing market survey results).
17. McKinsey Quantum Technology Monitor 2026, April 2026 (finance value estimate and overlap note).
18. Top Global Banks Exploring Quantum Technologies in 2026, The Quantum Insider, March 27, 2026.