Classiq and Israel Natural Gas Lines used hybrid quantum-classical optimization to search gas pipeline pressure settings. IonQ Forte-1 produced physically valid candidates close to the classical optimum, but the feasibility test did not demonstrate quantum advantage.
IonQ hardware has now been used to test a quantum optimization workflow for a real engineering model: the pressure settings that determine how much natural gas a transmission network can deliver. The reduced experiment returned physically valid candidates close to the continuous classical optimum, but the result is a feasibility demonstration rather than evidence that a quantum processor outperformed classical optimization.
Classiq Technologies and state-owned Israel Natural Gas Lines Ltd. formulated gas delivery as a constrained optimization problem. The objective is to maximize throughput while preserving mass at junctions, maintaining consistent flow direction and keeping endpoint pressures above required minimums. INGL is Israel's national operator for the transmission network, so the case is tied to an operationally relevant engineering domain even though the reported instance was deliberately reduced.
The underlying hydraulic relationship is nonlinear. The study uses the Panhandle-B equation and approximates its flow exponent with a second-degree polynomial, taking αPB as 0.51. Once pressure values are discretized into binary choices, the number of possible assignments grows exponentially with the number of decision nodes. That structure allows the engineering problem to be expressed as a quadratic unconstrained binary optimization model, or QUBO, suitable for a QAOA workflow.
The technical result appears in an arXiv preprint. It therefore provides accessible technical documentation of the work, but the supplied material does not establish peer review or independent replication. The available description is centered on the preprint and announcement material rather than on a journal publication with external validation.
The full test case represented a six-node network with five directed edges. Five decision nodes were assigned three qubits each, giving 15 logical qubits and 32,768 possible states. Classiq's synthesis platform and simulator ran QAOA with 30 layers, while the classical optimizer COBYLA converged in approximately 50-60 iterations.
In that simulation, the workflow recovered the maximum-throughput valid operating point identified by exhaustive classical evaluation and hydraulic simulation. This is useful validation of the encoding and constraint handling. It is not a quantum speedup: the classical reference was used to check the answer, and the supplied material does not report a comparison showing lower runtime, energy use or total workflow cost.
The experiment also differs from a conventional statistical study. It concerns a reduced optimization instance rather than a population sample, and the available description does not report repeated-trial counts, p-values, confidence intervals or an uncertainty analysis. Those omissions do not invalidate the engineering demonstration, but they limit what can be inferred about robustness across network topologies, noise conditions and alternative discretizations.
The hybrid design matters because QAOA does not remove the need for classical computation. A classical optimizer adjusts circuit parameters, the quantum processor samples candidate bit strings and engineering software checks whether those candidates obey the physical model. The practical question is therefore the performance of the complete loop rather than the qubit count alone. This distinction is consistent with the methodological caution expected in research communities associated with MIT, CERN and the journal Nature, where algorithmic claims are normally separated from demonstrations of system integration.
For hardware execution the researchers reduced the instance to 10 logical qubits. The setup used six decision qubits and four auxiliary qubits, with two decision nodes and fixed customer pressure. To limit errors from circuit depth, the IonQ Forte-1 trapped-ion quantum processing unit ran only two QAOA layers.
The two leading valid outcomes were reported as (qBVS1, qBVS2) = (5, 4) and (4, 3). Together they bracketed the continuous classical optimum within one pressure-discretization step and were found compatible with the SIMONE hydraulic software. That is a concrete hardware result: the QPU generated candidates that survived the physical validity check. It does not establish that the processor found the optimum more efficiently than a classical method.
The hardware test also illustrates the central compromise in near-term quantum optimization. Increasing QAOA depth can provide a richer variational circuit, but it also exposes the computation to more imperfect gates and readout errors. The reported p = 2 circuit favored a shallow, lower-noise experiment, while the more complete p = 30 result was obtained in simulation rather than on the QPU. The comparison is therefore between different execution regimes, not a like-for-like benchmark of equal-depth quantum and classical algorithms.
The work sits alongside other efforts to connect quantum processors with conventional computing infrastructure, including this earlier quantum deployment, but the technical challenge here is optimization under hydraulic constraints rather than secure communications.
Classiq and INGL position the quantum workflow as a global search engine for promising operating scenarios. That division of labor is sensible. SIMONE remains the engineering-grade tool for hydraulic verification, while the QAOA layer proposes discrete candidates that can be checked against the network model. IonQ's public framing likewise emphasizes the practical value of testing quantum computing within complex energy-planning workflows, not a demonstrated replacement for established optimization software.
The evidence supports a carefully limited claim. A hybrid algorithm was formulated for gas transmission, solved successfully in a full simulator and executed on IonQ Forte-1 at reduced scale. The reported hardware samples were physically valid and close to the classical reference, yet the study supplies no measured quantum advantage, no fault-tolerant operation and no demonstration that the approach improves live network management.
For infrastructure operators, the important threshold is not whether a QPU can produce a plausible pressure assignment. It is whether the complete process remains accurate, fast and robust as networks grow, pressure discretization becomes finer and engineering validation becomes more expensive. The supplied results do not answer those scaling questions, so the credible significance is methodological: quantum optimization can be inserted into an existing verification pipeline without replacing the classical hydraulic model.
QAOA uses a parameterized quantum circuit to sample candidate solutions rather than evaluating every classical possibility in parallel. Its output quality depends on the encoding, circuit depth, parameter search and hardware noise, while the physical and auxiliary qubits in this experiment are not logical error-corrected qubits. That distinction keeps the result in its proper category: a near-term hardware feasibility test whose value will be decided by end-to-end classical comparisons and larger validated networks. In that respect, the experiment is best read as an engineering benchmark for integration and reproducibility, not as evidence that quantum hardware has already surpassed classical optimization for gas transmission.
