TechCapsules

AI-summarised brief · reviewed before publication

Oxford Quantum Circuits (OQC) partnered with Trust Base, the digital‑innovation arm of Sumitomo Mitsui Trust Group, to benchmark classical, hybrid quantum‑classical and fault‑tolerant quantum algorithms on core financial risk‑modeling tasks. The study compared four architectures—classical Monte Carlo, classical physics‑informed neural networks (PINNs), quantum‑compressed PINNs (QPINNs) and quantum Monte Carlo (QMC) via amplitude estimation—across Black‑Scholes/Garman‑Kohlhagen, Dupire local‑volatility and Hull‑White models. QPINNs showed higher parameter efficiency but suffered from latency and noise‑sensitive Gamma calculations, while classical PINNs delivered faster runtimes and greater numerical stability. Optimizations such as quantum signal processing cut QMC resource needs by up to 16‑fold in T‑gates and four‑fold in logical qubits. Fault‑tolerant analysis indicated that raising the error‑correction threshold from 1 % to 5 % could shrink a 32‑qubit encoding’s physical qubit count from roughly 404 k to 130 k, outlining a path for hybrid quantum solvers in enterprise risk pipelines.