Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
The potential power of quantum computing has been demonstrated, but commercial viability is widely assumed to hinge on unrealized thresholds. More qubits. Lower error rates. Fault-tolerant machines.
But what if useful applications exist today?
New research from Synopsys and IonQ suggests they do. Even without error-corrected hardware, a quantum processor can take on one costly step inside an engineering simulation while classical systems handle the rest. The gains are confined to that operation, not the entire workflow, but they deliver a measurable improvement in simulations engineers run every day.
Three kinds of computing, one workflow
High-performance computing (HPC) remains the foundation of engineering simulation. GPU acceleration, reduced-order models, and surrogate methods have dramatically cut runtimes for large fluid and structural simulations, and the biggest supercomputers continue to push the ceiling on problem size. AI is also contributing, with machine-learned force fields making atomistic simulation orders of magnitude faster and helping researchers study larger systems over longer timescales.
Quantum computing joins that mix rather than displacing any part of it. HPC carries the bulk of the computation. AI finds patterns, builds faster approximations, and reframes problems into more tractable forms. A quantum processor takes on specific calculations that remain difficult for classical heuristics.
These calculations have a relatively narrow profile: small inputs, heavy computation, compact results. Entire engineering applications almost never fit that description. Individual steps inside them sometimes do.
Graph partitioning is one of them. Large finite element analysis (FEA) models must be divided across compute resources before a solver can run, and the ordering of the resulting sparse matrices determines how much work the solver must perform. Finding the best possible split is computationally out of reach, so solvers rely on approximation methods, and no single method is best for every model.
This represents a small, well-defined decision with outsized downstream cost, making it a good candidate for quantum acceleration.
What the research shows
The joint Synopsys-IonQ research covers two domains of simulation: FEA, which predicts how products behave under impact, vibration, and stress, and computational fluid dynamics (CFD), which predicts how liquids, gases, and heat move through and around them. The FEA results offer the clearest evidence of near-term commercial potential, while the CFD work shows progress toward a more technically demanding application.
In 2025, Synopsys and IonQ researchers reported wall-clock improvements of up to 12% using a variational quantum method in a variety of FEA simulations, including roof-crush, vibration, and blood pump behavior. The 2026 follow-up, which received the 1st Place Best Paper Award at IEEE Quantum Week, replaced that approach with a scalable non-variational method and reported end-to-end FEA gains ranging from 5.9% to 14.6% on far larger industrial models, including vehicles, jet engines, drills, and impellers with meshes reaching 35 million elements. The simulations in both studies were performed with Ansys LS-DYNA.
Two details are worth highlighting. The gains are end-to-end, measured across the full run rather than an isolated kernel. And they were achieved without error-corrected hardware, with physical execution validated on IonQ’s trapped-ion system.
The quantum computer is doing something narrow here, essentially cutting a graph, while everything else in the solver stays classical. This means the entire simulation code does not need to be ported to a quantum machine.
A wall-clock reduction of 5.9% to 14.6% may sound modest until it is applied to an automotive crash program that runs thousands of simulations per vehicle or a complex stress test that requires a week of continuous computing. Compute cost and schedule pressure are among the biggest constraints engineering teams face. For a seven-day simulation, a wall-clock reduction nearing 15% could shave a full day off the runtime.
Quantum-accelerated CFD also shows promise, though commercial viability is further off. Researchers have now run quantum fluid simulations on trapped-ion hardware under more realistic conditions, including flows that vary across three-dimensional space and interact with walls. But the simulations become difficult to scale because extracting results from the quantum processor and feeding information back into it takes too much time as the problem grows.
The gaps between research and production
Hardware access is the obvious constraint. Quantum systems are scarce and heavily oversubscribed, and production workflows need dependable access and predictable performance.
Consistency across models is another hurdle. Commercial simulation software supports thousands of use cases built around confidential customer designs that software developers never see. Demonstrating gains on a set of open research models is different from delivering them reliably across an installed base. The next phase of the research will expand beyond the current test set and explore whether the same methods apply to other optimization problems, including cell placement in chip design.
In-depth customer validation is also needed. Engineering teams will have to run these methods against their own models, compare them with established approaches, and judge whether the benefit justifies the cost.
They will also need software that hides the complexity. Engineers shouldn’t have to understand quantum algorithms to run a quantum-assisted simulation. Processor access, data movement, method selection, and result verification all have to sit behind the software interface, allowing engineers to use quantum computing like any other specialized computing resource.
The longer arc
Building hybrid workflows now delivers measurable gains and positions teams for the hardware still to come.
Quantum vendors have broadly suggested fault-tolerant systems will emerge in three to five years. Whether that timeline holds is unknown, but organizations that have not built the algorithms, integrated workflows, and internal expertise will not be ready to compound the gains when new hardware arrives. Those gains include larger problem sizes as well as shorter runtimes, allowing simulations to include more physics before they become impractical to run.
Simulation fidelity is largely determined by mesh size, and engineers routinely face a tradeoff: a finer mesh buys a small gain in accuracy at the cost of days they don’t have in the schedule. Past a certain point, mesh size exceeds the practical limits of classical computing altogether. Some questions never get asked as a result.
Materials science also faces computational limits. The electron interactions that govern how materials actually behave cannot be solved exactly, so researchers have spent forty years refining approximations. Feynman’s original argument for quantum computers was that simulating nature requires a quantum machine. Atomistic simulation on quantum hardware, while still early, is aimed squarely at that: solving those interactions directly rather than approximating them.
That ambition is years out. Commercial value can emerge sooner from specific workloads where the improvement is large enough and reliable enough to justify bringing quantum hardware into the loop.
Graph partitioning inside an FEA solver may be an unglamorous place for that to start. But it’s a real one.
Continue Reading
Agentic AI in Chip Design: The Rise of Autonomous Engineering (Part 2 of 4)
