Quantum reordering optimizes liquidity in TARGET2 payments

The study extends prior work by McMahon et al. (2024) to the Italian segment of TARGET2, a high-value payment system (HVPS) in the Eurozone.

Using a Constrained Quadratic Model (CQM) solver, researchers optimized payment batches, yielding average daily liquidity savings between EUR 23.16 million and EUR 38.35 million over a 35-day sample.

These savings represent 0.4 percent to 0.7 percent of daily liquidity reserves for batch sizes of 70 and 140 payments, respectively, with maximum savings reaching 4.4 percent.

The quantum annealing process for each batch takes approximately 5 seconds, demonstrating its efficiency in real-time gross settlement (RTGS) environments where liquidity efficiency depends on transaction settlement order.

This novel approach focuses on liquidity recycling through optimal reordering, a computationally intensive problem effectively addressed by hybrid quantum solvers.

A Machine Learning (ML) framework was developed to identify payment batch characteristics predicting liquidity savings.

Batches with fewer than five participants acting as both senders and receivers are less optimizable, while optimizability increases up to 15 such participants; an increasing number of unique receivers also enhances optimizability.

This ML framework can help allocate limited optimization resources more effectively.

The study also benchmarks the quantum approach against a classical Simulated Annealing Algorithm (SAA), which produced comparable savings for smaller batches.

The SAA was extended to handle larger batch sizes up to 700 payments, estimating daily liquidity savings of up to EUR 3.9 billion, demonstrating its potential as a scalable alternative despite increased processing delays.