Quantum computing in finance: portfolios, risk and security

Quantum computing in finance is the marriage that has been announced for a decade, because a sector that lives on optimisation, stochastic simulation and security meets a technology that promises something on precisely these three fronts.

The large international banks have had active quantum teams for years, and this is a signal to be read carefully, because it indicates the seriousness of the trajectory, not yet the maturity of the results. This guide separates the different levels, explaining what portfolio optimisation, quantum Monte Carlo and financial machine learning really promise, on what timescales, and why cryptographic migration, alone among them, is already a regulatory obligation with a deadline. It closes with a concrete path for an institution that wants to proceed methodically. In quantum computing applied to finance, more than anywhere else, what separates strategy from fashion is the baselines.

What are the application areas in finance?

There are four application areas, with quite different levels of maturity from one another.

• Security. Transactions, digital signatures and encrypted channels rest on algorithms vulnerable to Shor’s algorithm, and the migration to post-quantum cryptography is already a matter for supervisors.
• Optimisation. Portfolios, rebalancing, arbitrage, collateral management, combinatorial problems native to the sector.
• Stochastic simulation. Derivatives pricing and risk measures, where quantum Monte Carlo promises speed-ups proven in theory.
• Machine learning for fraud, credit and markets, the most exploratory area, to be read with the caveats of the page on quantum machine learning.

When it comes to quantum computing in finance, the areas that receive the most media attention, portfolio optimisation and machine learning, are not the ones an institution should act on first. If the four areas are ranked by how much they are talked about at conferences and in articles, cryptography comes last, because it seems the least spectacular. If instead they are ranked by real urgency, cryptography comes first, because it is the only one with a regulatory deadline already set, while everything else is still research and prototyping. These two orderings, the one of attention and the one of necessity, are therefore the reverse of each other, and this is the first thing to clarify in any committee discussing quantum computing, because the area that is talked about least is precisely the one that is already mandatory by law.

How does quantum portfolio optimisation work?

The portfolio selection problem with real-world constraints, cardinality, minimum lots, transaction costs, is combinatorial and classically complex. The quantum route formulates it as a QUBO (Quadratic Unconstrained Binary Optimization) and entrusts it to variational algorithms such as QAOA or to quantum annealing, with the aim of exploring the space of allocations better than classical heuristics. The real state of play consists of prototypes on tens of assets, not of demonstrated advantages on institutional-size portfolios, where classical solvers remain extremely strong and run in production on instances of thousands of securities that today’s quantum computing does not come close to. The value today, consistently reported by those who experiment, is above all methodological, because the rigorous formulation that quantum requires often improves the classical solution as well, and the pipeline built in emulation is already ready for the hardware to come.

What is quantum Monte Carlo for pricing and risk?

Much of quantitative finance rests on Monte Carlo simulations to estimate prices and risks, a method that works well but has a considerable cost, because to halve the error the number of calculations must be quadrupled. Quantum amplitude estimation would therefore make it possible to loosen precisely this constraint, offering the same level of precision with far fewer calculations, a quadratic advantage proven in theory that would translate into substantial speed-ups. The constraint is just as clear-cut, because this technique requires deep, coherent circuits, realistically hardware with error correction. It is the financial area with the soundest theory and the longest timelines, perfect to prototype today but quite premature to promise in the short term. Those who build competitiveness on amplitude estimation now will in any case have an advantage that is hard to close when quantum computers become practicable, because the hardest part is not so much the algorithm as integrating it into development processes tied to real risks.

Is quantum machine learning useful for fraud and credit?

The most plausible answer is ‘not yet’, and perhaps not in the form many imagine. Fraud, credit scoring and market forecasting work on massive classical data, the terrain where quantum machine learning meets its structural limits, from data loading to dequantisation. No demonstrated advantage exists today on these tasks over classical computing methods, which in this sector are moreover already state of the art. The sensible directions remain kernel components inside hybrid pipelines and low-cost methodological research that brings the two worlds of quantum and classical computing together. A bank therefore does well to maintain a knowledge watch, while it is premature to put quantum machine learning into industrial plans as a short-term lever. This distinction, here more than elsewhere, helps to plan the budget better.

Why is cryptography the real urgency for the sector?

Finance is the perfect target of the strategy known as ‘harvest now, decrypt later’, because it generates encrypted flows of very high value, data with multi-year confidentiality and digital signatures whose legal validity must last for decades. A contract signed today with vulnerable cryptography could be repudiated or forged once the attack becomes practicable, and interbank communications archived today will be readable tomorrow. Regulators have already put this on record, because international financial authorities and the European operational resilience framework call for cryptographic inventories and migration plans with defined horizons. For a bank, post-quantum cryptography is not so much innovation for its own sake but rather real and urgent compliance with a deadline, and the quantum line in the budget should start precisely from here, before any computational exploration. The paradox of quantum computing in banking is all here: the first mandatory expense serves to defend against the technology, not to use it.

What are banks really doing?

Banks, in deeds more than in press releases, are essentially doing three things.

• Research programmes with quantum vendors on optimisation and pricing, active for years in the largest international banks, a serious exploration that nevertheless still produces only prototype results.
• Quantum key distribution trials on metropolitan routes, more demonstrative than operational, and not to be confused with quantum computing, because key distribution is quantum communication, a different technology.
• Cryptographic migration programmes, with growing concreteness, the only ones with structural budgets and defined deadlines.

The correct reading for an average institution is that there is no need to replicate the laboratories of the giants, but to replicate their discipline, with a small, competent team that can read other people’s results, plus serious execution of the one mandatory building site. It is a strategy within the reach of any IT department, and it keeps the institution within the conversation on quantum computing without exposing it to its speculative costs.

How do you assess a financial use case?

A filter of three questions applies, valid for every sector, plus one question specific to finance.

• Is the problem combinatorial or stochastic, with a complexity that explodes?
• Is there a quantified classical baseline, that is, the current cost of the best solvers or of Monte Carlo in production?
• Does the value of improving it justify a dedicated programme?
• Specific to finance. Does the expected advantage survive the latency and regulatory constraints of the real process?

A theoretical speed-up on pricing is worth little if the bottleneck lies elsewhere in the chain. Once the filter has been passed, the path is the proof of concept in emulation with a baseline alongside, with your own numbers, on your own data, before any commitment to hardware.

What role does emulation play in the financial journey?

Emulation plays the role of the laboratory in which everything that precedes the hardware happens at infinitely lower cost. On the EQM platform used by HumanQ, a bank could prototype the QUBO formulation of its own allocation problem and measure its scalability, implement amplitude estimation on reduced payoffs and verify its depth requirements, and have the security team run Shor’s algorithm on small instances, demystifying the cryptographic threat. Every run is reproducible and comparable with the classical baseline, the necessary condition for promoting investments based on numbers rather than mere proclamations based on slides.

What are the realistic timelines, in brief?

Here is a roadmap for the finance sector.

• Immediately. Cryptographic inventory and start of the post-quantum migration, because the regulatory deadline already exists and the technical timescales are long.
• One to three years. Proofs of concept in emulation on portfolio optimisation and amplitude estimation, with the building of an in-house team.
• Three to ten years. First expected advantages on pricing when error correction makes deep circuits practicable, with optimisation as a possible forerunner on selected instances.

Quantum computing in finance will reward, as always in this sector, those who have done their homework before the deadline, and it is worth remembering that today there are two deadlines, one mandatory and one preparatory, in the right order.