Quantum computing is starting to signal it could provide value for financial services problems, with recent results from banks and asset managers showing early signs of the benefits quantum computers could hold for useful applications.
In two separate announcements earlier this year, HSBC and Vanguard revealed the results of collaborations with IBM that hint at how quantum computers could begin to tackle some of the most computationally demanding tasks in finance. HSBC presented data that quantum-enhanced models helped improve predictions in corporate bond trading. Vanguard explored new algorithms that optimized portfolio construction under real-world constraints. Both experiments leave more to be explored as both quantum computers and new algorithms continue to scale. Yet, taken together, the findings may mark an early turning point, with finance emerging as one of the first industries to show the potential of measurable benefits from a technology that has long existed in the realm of experiment.
“Despite its emerging nature, quantum computing is slowly becoming a complementary explorative tool to tackle real-world problems in financial markets, and our work with HSBC and Vanguard provides some concrete exemplary quantum-enhanced approaches with learnings to inform future research directions,” said IBM Quantum Industry Applications Lead Dr. Manuel Proissl, in an interview with IBM Think. “It is all about finding practically relevant intersections of quantum and classical algorithms that can address financial modeling challenges.” This has been at the heart of these collaborations using IBM’s latest quantum computing systems.
For HSBC, the tests were conducted on real production-scale bond trading data, not on artificial examples. The experiment demonstrated an up to 34% improvement in predicting whether a bond trade would be filled compared to baseline methods. However, further research is required to determine how to reproduce them for different market regimes.
Empirical evidence like this advances a potential understanding of how quantum computers could be applied to financial problems and where they could provide value. Just as in the early days of classical computing, we hope this course of experimentation with our partners will yield significant progress.
The team also developed a framework that enables the approach to be scalable. Instead of inserting quantum computers directly into the high-speed trading loop, which is impractical today, they proposed a hybrid model. Quantum circuits, which are the programmable building blocks of quantum algorithms, generate features offline. These are then stored in a database and reused whenever similar market conditions arise.
“There is also a technique introduced, called classical-to-quantum event matching, which allows you to reuse quantum-generated features, regardless of the quantum algorithm,” Proissl explained. “If a similar market event observed in the past occurs again, then you can reuse these quantum features and respective models in real-time since the market event’s outcome is independent in this solution approach.”
That design, which blends cloud-based quantum computations with the speed of existing systems, would offer firms both flexibility and cost control when scaled. They could decide when to invest in generating new quantum features and when to rely on cached ones, making integration more realistic with today’s limited hardware.
Portfolio optimization offers another path
At Vanguard, the focus was different. The challenge was constructing an exchange-traded fund portfolio while navigating the messy constraints of the real world. Portfolio construction is a canonical problem in finance, dating back to Harry Markowitz’s 1950s theory of balancing risk and return. In textbooks, the task involves tracing the efficient frontier and selecting portfolios that maximize expected return for a given level of risk.
In practice, however, portfolio managers face thousands of assets, nonlinear constraints and regulatory rules that make the problem exponentially harder. Classical solvers can handle small versions of the problem, but as the universe of assets grows, the search space balloons beyond what classical methods can manage in a reasonable time.
Vanguard and IBM turned to a variational quantum algorithm, or VQA, a hybrid method designed for today’s noisy quantum devices. A VQA uses a quantum processor to generate trial solutions, while a classical computer evaluates and refines them in a feedback loop. This design makes the algorithm resilient to the errors that plague current hardware. The team utilized 109 active qubits out of 133 available on an IBM Quantum Heron processor, executing circuits with up to 4,200 gates, and then refined the quantum samples using a classical local search algorithm.
The results were encouraging. On a simplified Exchange-Traded Fund portfolio construction problem, the hybrid workflow performed on par with a state-of-the-art classical solver and in some scenarios surpassed it, particularly as the problem size increased. The quantum classical method consistently outperformed a purely classical heuristic baseline.
The lesson was not that quantum had suddenly leapt past classical computing, but that it had joined the family of heuristic approaches finance has long relied on. Optimization in practice is rarely about exact solutions. It is about good enough answers found quickly enough to be useful. Variational quantum algorithms, IBM researchers argue, represent a new kind of heuristic, one that explores complex landscapes differently than classical techniques.
“The idea was to demonstrate that we can do the same with real problems, and we did,” Proissl said of the Vanguard study. “But it also gave us clearer directions where we have to go next. Eventually, we aim to handle many more assets, other asset classes and more complex constraints and market scenarios.”
IBM’s quantum roadmap
These experiments are direct results of technological progress being made on IBM’s broader quantum roadmap, most recently updated this year with plans to deploy a large-scale, fault-tolerant quantum computer by 2029—a quantum computer with hundreds of logical qubits capable of running one hundred million operations, which could accelerate time and cost efficiencies in fields such as drug development, materials discovery, chemistry and optimization. But hardware and software progress between now and the end of the decade will, IBM believes, deliver applications with a “quantum advantage” by the end of 2026—when a quantum computer can run a computation more accurately, cheaply or efficiently than a classical computer.
For IBM, both the HSBC and Vanguard results feed directly into its development strategy. “The goal of our mission is to build useful quantum computers, and the word useful here means solving real-world problems,” Proissl said. “To build devices and systems that are practical and useful to the industry, we need these types of collaborations with HSBC, with Vanguard.”
