IonQ runs quantum portfolio test
IonQ reported portfolio‑optimization results using trapped‑ion quantum hardware on 250 S&P‑500 assets with 2020–2025 data, highlighting QUBO advantages for dense problems and cross‑generation scalability. It’s a concrete data point for quantum approaches to combinatorial asset‑allocation research. (x.com) (x.com)
IonQ and collaborators published a preprint titled "Large-scale portfolio optimization on a trapped-ion quantum computer" on arXiv (arXiv:2602.23976) with a March 2, 2026 date and author affiliations that include IonQ and Kipu Quantum. (arxiv.org) The methodology applies RMT-based correlation-matrix denoising and community-detection clustering to partition a 250-asset universe into hardware-sized QUBO subproblems using a correlation-guided greedy splitting scheme. (arxiv.org) Subproblems were solved with a bias-field digitized counter-diabatic quantum optimization algorithm (BF-DCQO), a non‑variational approach that the authors say avoids classical parameter-training loops. (arxiv.org) Hardware experiments ran on IonQ systems including an IonQ Forte and a 64‑qubit Barium development device (executions reported on 36‑qubit and 64‑qubit configurations, with subproblems up to ~60 qubits), followed by recombination of low‑energy candidates and a two‑stage post‑processing (fast repair and cardinality‑preserving swap local search). (ionq.com) The study reports systematic improvement in final objective values and risk–return tradeoffs when increasing executable subproblem size from 36 to 64 qubits, and compares results against the classical optimizer Gurobi and randomized baselines. (arxiv.org) Authors benchmark the pipeline on daily log‑returns over a multi‑year window and frame the task as a cardinality‑constrained Markowitz QUBO, positioning the work as a hardware‑tested route that quantifies the tradeoff between executable qubit budget and reconstructed portfolio quality. (arxiv.org)
