Mastercard and D-Wave: Quantum Computing for Financial Services

The Problem

Credit card fraud costs the global payments industry tens of billions of dollars per year. Mastercard processes billions of transactions and flags suspicious ones in real time, typically within milliseconds. Alongside fraud, the company manages other computationally heavy problems: anti-money laundering screening, cross-border settlement, and the design of consumer loyalty and rewards programs, all of which involve searching enormous spaces of possible combinations for good solutions.

Current systems handle these with classical optimization and machine learning. They work well, but several of these problems are combinatorial at their core, exactly the problem shape that quantum annealing was designed to attack. That is the context for Mastercard’s quantum partnership.

The Mastercard and D-Wave Partnership

In July 2022, D-Wave and Mastercard announced a multi-year strategic alliance to accelerate the adoption of quantum computing in financial services. The documented facts:

• The partnership covers research and development of quantum and quantum-hybrid applications
• Named application areas are consumer loyalty and rewards, cross-border settlement, fraud management, and anti-money laundering
• The technology base is D-Wave’s annealing quantum computers and quantum hybrid solvers, delivered through the Leap quantum cloud service with real-time secure access via Mastercard’s network
• Mastercard CIO Ken Moore framed the rationale around quantum computing’s “unique ability to analyze huge numbers of potential combinations” to deliver optimal solutions

D-Wave’s subsequent investor communications have continued to list Mastercard as a partner and customer building quantum-hybrid applications in these areas. Neither company has published benchmark results, accuracy figures, or details of any production deployment from the collaboration.

Why Annealing Fits These Problems

D-Wave’s machines are quantum annealers rather than gate-model computers. An annealer is built to do one thing: find low-energy configurations of a QUBO (quadratic unconstrained binary optimization) problem. Many financial services problems can be cast in this form. Choosing which subset of reward offers to assign to which customer segments under budget constraints is a combinatorial assignment problem. Netting payment flows in cross-border settlement is a constrained matching problem. Even parts of the fraud and AML pipeline, such as selecting an optimal subset of features or rules from a large candidate pool, reduce naturally to binary optimization.

The example below is a simplified educational illustration of how a feature-selection problem from a fraud-modeling pipeline could be written as a QUBO using D-Wave’s open-source Ocean tools. It is not Mastercard or D-Wave production code, and it does not represent any result from their partnership.

import numpy as np
import dimod

# Toy problem: pick a small subset of candidate fraud-model features
# that are individually predictive but not redundant with each other.
n_features = 6
np.random.seed(7)

# Mock "usefulness" score per feature (e.g., mutual information with label)
usefulness = np.random.uniform(0.2, 1.0, n_features)

# Mock pairwise redundancy between features (e.g., absolute correlation)
redundancy = np.abs(np.random.uniform(0, 0.6, (n_features, n_features)))
redundancy = (redundancy + redundancy.T) / 2
np.fill_diagonal(redundancy, 0)

# QUBO: maximize usefulness, penalize redundant pairs chosen together
bqm = dimod.BinaryQuadraticModel(vartype="BINARY")
for i in range(n_features):
    bqm.add_variable(f"f{i}", -usefulness[i])          # reward selection
for i in range(n_features):
    for j in range(i + 1, n_features):
        bqm.add_interaction(f"f{i}", f"f{j}", redundancy[i, j])

# Solve with an exact classical solver (a D-Wave annealer or Leap
# hybrid solver would accept the same BQM at much larger scale)
solution = dimod.ExactSolver().sample(bqm).first
chosen = [v for v, val in solution.sample.items() if val == 1]
print(f"Selected features: {chosen}")
print(f"Objective value:   {solution.energy:.3f}")

The same BinaryQuadraticModel object can be submitted to a physical annealer or to Leap’s hybrid solvers, which combine classical heuristics with quantum hardware to handle problems with up to millions of variables. That hybrid path is what D-Wave promotes for near-term enterprise workloads, and it is the technology stack named in the Mastercard announcement.

Honest Status

It is worth being precise about what is and is not known. The partnership is real, multi-year, and publicly committed to specific application areas. What has not appeared in the public record is any measured outcome: no published accuracy comparison for fraud detection, no settlement-efficiency figures, and no announced production system. Quantum annealing also has genuine open questions, most notably whether it outperforms the best classical optimization heuristics on real enterprise problems at current hardware scales.

For learners, the value of this case is strategic rather than numerical: one of the world’s largest payment networks judged quantum optimization promising enough to sign a multi-year alliance and route secure access to quantum solvers through its own network. Watching whether published results eventually follow is part of understanding how enterprise quantum adoption actually unfolds.

Learn more: Quantum Annealing

Sources

• D-Wave and Mastercard Take Quantum Leap into Future of Financial Services (D-Wave press release)
• D-Wave and Mastercard Take Quantum Leap into Future of Financial Services (Business Wire)