A European freight forwarding company (name anonymized in the published report) released its internal benchmark data from a six-month D-Wave Advantage2 pilot program this week. The company's logistics team ran real container shipment routing problems alongside their existing classical Gurobi solver, and the data is worth examining in detail.

The Problem Formulation

Container shipment routing at this scale involves 800–1,400 containers per weekly planning cycle, each with origin, destination, time window, and 3–5 carrier options. The optimization objective minimizes total freight cost plus late delivery penalty. This maps naturally to a QUBO formulation after encoding carrier selection as binary variables, but the resulting problem graph has 3,200–5,600 binary variables — too large to embed directly on Advantage2's Pegasus graph without minor embedding overhead.

The team used D-Wave's hybrid BQM (Binary Quadratic Model) solver, which decomposes large problems into sub-problems that alternate between quantum annealer and classical CPU components.

The Actual Results

The headline "15% cost reduction" refers to the improvement over the company's baseline greedy heuristic solver — not over Gurobi. When compared to Gurobi with a 120-second time limit:

  • D-Wave Hybrid: 8.3% better than greedy baseline, 1.2% better than Gurobi (120s)
  • D-Wave Hybrid: 4.1% worse than Gurobi with a 10-minute time limit
  • D-Wave Hybrid: Significantly faster than Gurobi at solution times under 30 seconds

This is an honest mixed result. For time-critical re-planning scenarios (port delays, carrier failures requiring rapid re-routing), the D-Wave hybrid solver provides solution quality competitive with Gurobi at a fraction of the classical computation time. For the weekly planning cycle where compute time is unconstrained, Gurobi remains superior.

What This Tells the Market

The freight forwarder framed this as a deployment success, and they're not wrong — but the value proposition is speed, not optimality. D-Wave's hybrid solver delivers near-Gurobi solution quality in under 30 seconds on problems where Gurobi takes minutes. For operational contexts where 30-second replanning windows exist, this has real value.

The lesson for evaluating quantum annealing pilots: ask whether the comparison baseline is the customer's existing heuristic or a well-tuned classical optimizer with adequate compute time. The honest answer often looks like this report — competitive in specific time-constrained scenarios, not yet dominant across the board.