Sumitomo Mitsui Trust Group, through its digital subsidiary Trust Base, is actively partnering with OQC to explore quantum workflows, not just research, but integration into existing financial processes, the company says. The project focused on three specific financial areas: Value-at-Risk, Credit Valuation Adjustment, and derivative pricing, areas with “mathematically rich and commercially material” problems.
OQC and Trust Base combined classical Monte Carlo baselines, quantum-enhanced methods, and Physics-Informed Neural Networks to assess the practical route toward quantum finance, finding that “near-term value comes from disciplined benchmarking, hybrid architectures and hardware-aware research.” Classical Physics-Informed Neural Networks are a practical approach today, balancing pricing accuracy, runtime and workflow integration while longer-term quantum solutions are developed.
OQC and Trust Base Explore Quantum Finance Workflows
This work underscores that practical quantum advantage will be won not just through algorithmic innovation, but through a comprehensive understanding of the entire computational process. The collaborative project prioritises establishing robust classical benchmarks against which quantum-enhanced methods can be rigorously tested; without reproducible baselines for established techniques like Monte Carlo simulations and neural PDE solvers, accurately assessing the added value of quantum approaches proves impossible.
This methodical approach extends beyond simply demonstrating quantum advantage in isolation, instead concentrating on how these methods integrate into existing financial processes. Evaluating the complete workflow, encompassing pricing, sensitivity analysis, exposure calculations, and risk assessment, is central to determining practical utility, as runtime, calibration costs, stability, and seamless integration are all critical considerations.
The work with Trust Base underscores the importance of hybrid architectures, where quantum components augment classical systems, selectively applied to problem areas where they may offer improved representation, sampling, or optimisation, according to the company. “Near-term quantum systems are most likely to contribute as part of a broader classical stack,” the researchers note, highlighting a phased implementation strategy that acknowledges the current limitations of quantum hardware.
The company views this as an important step in preparing for advancements in qubit quality, qubit count, latency, and error correction, anticipating a future shift in the balance between classical and hybrid methods. The research being conducted, they believe, will define which workloads will be prepared when that transition occurs.
For Trust Base, the collaboration represents an opportunity to explore the “substantial” potential of quantum computing within the financial sector, but only through a rigorous and collaborative approach grounded in both market realities and machine capabilities. “The route to it starts with work like this: rigorous, collaborative and grounded in the realities of both markets and machines,” a company spokesperson stated, emphasising the importance of practical application over theoretical promise. This partnership signals a concrete investment in quantum workflows, moving beyond research to explore integration into existing financial processes.
Classical and Quantum-Informed Neural Networks for Derivative Pricing
Classical Physics-Informed Neural Networks offer immediate benefits to financial modeling, according to recent collaborative work between OQC and Trust Base, a digital subsidiary of Sumitomo Mitsui Trust Group. These networks balance pricing accuracy with runtime efficiency and ease of integration into existing financial systems, providing value even as more complex quantum solutions mature. This approach explores practical routes to research focusing on distant, fault-tolerant quantum computing.
The project deliberately examined three complementary methods, Physics-Informed Neural Networks, Quantum Physics-Informed Neural Networks, and Quantum Monte Carlo, to assess their viability within established financial workflows. Physics-Informed Neural Networks, or PINNs, directly learn solutions to pricing equations from governing parameters, boundary conditions and payoff conditions, creating a continuous pricing surface without reliance on fixed grids. Quantum Physics-Informed Neural Networks, specifically Quantum-compressed PINNs, then extend this architecture by incorporating a parameterised quantum circuit as a trainable quantum feature map.
Evaluating operational usefulness required more than simply demonstrating theoretical quantum advantages; the team focused on how these methods compared against robust classical baselines, the company says. A key consideration was model stability, as a pricing model producing unstable sensitivities is unsuitable for real-world risk management, even if it achieves accurate pricing. The staged implementation strategy acknowledges that a transition from classical to fully quantum systems will not be instantaneous.
Instead, financial institutions must build capability incrementally, identifying suitable workloads and integrating quantum methods into existing risk and pricing pipelines. Quantum Monte Carlo, exploring amplitude-estimation techniques for payoff expectations, represents another avenue for investigation, though its ultimate impact hinges on advancements in hardware quality and tighter integration with classical computing, Trust Base reports. The project’s findings suggest that a practical path to quantum value involves a phased approach, leveraging near-term solutions like classical PINNs while simultaneously preparing for the potential of more advanced quantum algorithms.
Quantum Monte Carlo Faces Hardware Realities in Finance
OQC and Sumitomo Mitsui Trust Group’s digital subsidiary, Trust Base, investigated these methods alongside established classical techniques to determine a practical path toward quantum advantage in finance. The investigation revealed that current quantum systems are most effective as components within a broader classical computing stack, rather than wholesale replacements. Quantum Physics-Informed Neural Networks, or QPINNs are currently best positioned as a means of studying hybrid architectures, not as immediate production replacements for existing classical solvers.
Researchers found that these networks can be competitive in parameter count, but practical deployment remains contingent on hardware advancements. While theoretically appealing, its near-term performance is heavily influenced by practical considerations such as reflector choice, compilation strategy, qubit placement, and the inherent noise present in quantum hardware.
The project’s findings indicated that simpler circuit designs could sometimes yield smoother, more stable results than more complex, theoretically precise constructions. “Lower-cost circuit choices could sometimes produce smoother empirical behaviour than more exact but deeper constructions,” the team reports, highlighting the importance of pragmatic design decisions.
Practical Quantum Readiness: Benchmarking and Hybrid Architectures
This approach allows financial institutions to begin realizing value from advanced computational techniques even before fully fault-tolerant quantum hardware is available. The project’s findings emphasize that achieving practical quantum advantage requires more than just innovative algorithms; it demands a careful assessment across the entire computational stack. Banks and financial institutions are not expected to transition immediately from classical systems to fully quantum solutions, but rather to adopt a phased approach focused on building capability and identifying suitable workloads. A key takeaway from the collaboration is the importance of disciplined benchmarking and hybrid architectures.
This rigorous evaluation revealed that near-term value is most likely to emerge from combining the strengths of classical and quantum approaches, rather than relying on quantum solutions in isolation. The focus on derivative pricing as a testbed proved particularly insightful, as it highlights the need for models that accurately value financial instruments dependent on future market uncertainties.
While established methods like Black-Scholes, finite-difference solvers and Monte Carlo simulation are well understood for simpler products, more complex derivatives require computationally intensive techniques like Monte Carlo simulation. The project’s results suggest that Physics-Informed Neural Networks can provide a viable alternative, offering a balance between accuracy and runtime.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
