Quantum reordering frees up to €167 million daily in TARGET2
Reordering payment batches with quantum and classical algorithms reduces daily liquidity requirements in TARGET2 by up to €167 million, according to research by the Bank of Italy and the Bank of Canada. The study tested 35 days of Italian settlement data across multiple batch sizes.
Two solvers deliver matching gains
Applying D-Wave’s Constrained Quadratic Model hybrid quantum solver to 35 days of Italian TARGET2 transactions generated average daily liquidity savings of €23.16 million for batches of 70 payments and €38.35 million for batches of 140 payments.
These reductions represent 0.4 percent and 0.7 percent of daily liquidity reserves, respectively, with single-day peaks reaching €167.7 million.
A classical simulated annealing benchmark achieved virtually identical savings within the same execution constraints.
When scaled to batches of 700 payments on classical hardware, simulated annealing yielded average daily savings of €269.3 million, though accumulating larger batches required an average wait time of 25 minutes.
Network topology dictates efficiency
The authors trained a LightGBM machine learning classifier using Shapley additive explanations to determine why certain payment batches yield greater optimization gains.
Batches showed minimal savings when fewer than 5 institutions acted as both sender and receiver, with optimizability peaking at 15 dual-role participants before tapering off.
A higher count of unique receivers also increased savings potential, whereas general network clustering metrics had negligible predictive power.
Classical algorithms win the present
Quantum hardware provides no practical advantage over classical simulated annealing at current operational batch sizes.
Major liquidity gains in larger queues require settlement delays that real-time payment operators cannot accept.
Near-term implementation should focus on selective filtering rather than quantum upgrades.