> Engineers submit a full-vehicle crash simulation in the morning and come back the next day. Overnight, a supercomputer solves a system of equations with hundreds of millions of variables. > > On September 17, IonQ and Synopsys announced a joint research effort: a quantum algorithm embedded into Synopsys' Ansys LS-DYNA simulation software reduced total simulation time by 5.9% to 14.6% across four models — an automobile, an industrial drill bit component, a fluid impeller, and a jet engine assembly. Mesh sizes reached up to 35 million data points. The work won first-place Best Paper at IEEE Quantum Week 2026 (Toronto). On the day of the announcement, IonQ stock rose 9.47%, closing at $40.33.
It manages the traffic lights, not the engine
IonQ's own metaphor is a traffic cop. The quantum algorithm does not participate in the physics; it decides "who goes first" — reordering the equation system before solving to eliminate redundant computation.
Three layers to this metaphor:
- It doesn't handle transport. The actual physics is still computed by the classical solver; the quantum part touches no mechanics, material models, or contact algorithms.
- It only manages ordering. Sparse system solve cost is extremely sensitive to ordering. A good ordering means fewer fill-in nonzeros during factorization, saving memory and compute; a bad one can blow up memory entirely.
- It runs only once. The ordering happens at the start of the simulation, and the entire subsequent solve benefits, while the ordering step itself is fast.
- Which models actually ran on real hardware
- Whether third parties can reproduce results on LS-DYNA
- How ordering benefits scale with problem size (35M data points vs. 350M)
- Whether any customer puts it into production workflows
- What's missing between 36 and 150 qubits
This is why 14.6% matters more than it looks: the savings compound across hundreds or thousands of subsequent time steps. In sparse direct solvers this corresponds to equation ordering and fill-path selection — a classic performance bottleneck in explicit dynamics codes like LS-DYNA.
Four models, one range
| Test model | Structural features | Simulation time reduction | |---|---|---| | Automobile | Large volume, many contact surfaces | within 5.9%–14.6% | | Industrial drill bit component | Dense local detail | within 5.9%–14.6% | | Fluid impeller | Rotationally symmetric, regular mesh | within 5.9%–14.6% | | Jet engine assembly | Many parts, complex joints | within 5.9%–14.6% |
Only the range endpoints were published; per-model figures were not disclosed. Notably, even the worst case saved 5.9%, and the four models span very different structures — evidence of generality that is arguably stronger than the 14.6% peak. In engineer terms: a 100-hour simulation saves nearly 15 hours; such jobs queue for days at automakers and aero-engine manufacturers.
150 qubits simulated, 36 qubits real
IonQ's statement contains two adjacent sentences: numerical simulation ran on up to 150 qubits, while physical execution was verified on IonQ's 36-qubit Forte trapped-ion quantum computer.
In other words, the quantum ordering algorithm was pushed to 150 qubits only in simulation; on real hardware it ran at 36 qubits. Which models ran on hardware versus simulation was not itemized. That gap is the distance between technical and commercial feasibility. Getting effective results on 36 trapped-ion qubits is a substantial engineering feat in itself — but buyers should not read the two numbers as "150-qubit hardware."
Same-week companion: AI writes the quantum circuits
On September 16, IonQ with Oak Ridge National Laboratory, NVIDIA, and the University of Tennessee Knoxville published another Best Paper Award-winning study. Finding suitable quantum optimization circuits traditionally requires repeated parameter search: scaling from 4 to 12 qubits grew search time from 34 seconds to over 11 minutes. The new generative-AI method synthesized circuits in a constant ~28 seconds across all sizes, with doubled solution quality. All computation ran on a single NVIDIA H200 via the CUDA-Q platform — no real quantum hardware involved. Constant time matters more than "faster than humans": only constancy makes cost predictable.
Embedding beats raw speed
Both papers share the same key move: embedding, not replacement. Ansys LS-DYNA is among the most widely deployed explicit dynamics codes in engineering. A quantum algorithm requiring customers to swap their whole simulation stack would face an impassable adoption barrier. Embedded in existing software, customers change nothing in their workflow — they just invoke one extra module at the ordering step. Synopsys' Prith Banerjee emphasized that even shaving a few percentage points off simulation time can meaningfully compress product development cycles — a framing closer to the real sales pitch than 14.6%.
Stock up 9.47%, but the other side of the ledger is harder
IonQ rose 9.47% (about $500M added market cap) on a day when Nasdaq rose 1.68% and the S&P 500 1.13%. Context: full-year revenue guidance of $450–460M (raised three times this year, partly via the SkyWater acquisition, not quantum revenue); Q2 non-GAAP EBITDA loss of $120.3M — larger than quarterly revenue; $3.1B in cash; stock down 37.92% over 12 months.
The 9.47% priced in research credibility, not product revenue. Research credibility has value for a company that needs continuous financing, but it remains one step short of "quantum computing contributing revenue."
What to watch next
Who captures the saved 14.6%?
One question the material doesn't answer: who owns the saved compute time. If a simulation queued for two days and ran 100 hours now runs 85, the team saves 15 hours of machine time and a day and a half of waiting. But what engineering organizations really want is more design iterations per unit time — turning "try five designs" into "try six." That second benefit requires organizational processes to absorb it.
Quantum computing's first knock on industrial software's door was not about performance — it was about workflow. Now that the toolbox has a new tool, someone has to be willing to pick it up.
References
1. IonQ press release, "IonQ Demonstrates Computer-Aided Engineering Workload Acceleration by up to 14.6% with Quantum Technology," 2026-09-17 — https://entangled.cloud/content/go/138904176 2. Synopsys / Ansys LS-DYNA joint research, IEEE Quantum Week 2026 Best Paper First Prize, 2026-09-17 — https://newslayer.com/article/ionq-and-synopsys-report-up-to-14-6-faster-engineering-simulations-3507bc 3. Ticker Report, "IonQ's AI-Quantum Breakthrough With NVIDIA Won a Best Paper Award," 2026-09-17 — https://ticker.report/news/ionq-ionqs-ai-quantum-breakthrough-with-nvidia-won-a-best-paper-award-but-can-2026-09-17 4. Benzinga, "Why Is IonQ Stock Surging Thursday?", 2026-09-17 — https://www.benzinga.com/trading-ideas/movers/26/09/61853199/why-is-ionq-stock-surging-thursday 5. 10jqka market-move commentary on IonQ, 2026-09-17 — https://t.10jqka.com.cn/pid_694119250.shtml