Four organizations building toward industrial quantum computing — Quantinuum, Rolls-Royce, quantum error correction specialist Riverlane, and EPCC, the UK National Supercomputing Centre at the University of Edinburgh — signed a formal multi-year agreement on July 14, 2026, to test whether near-term quantum hardware can accelerate computational fluid dynamics simulation for gas turbine design, one of the most computationally stubborn problems in industrial engineering. The work will run on Quantinuum's 98-qubit Helios quantum processor, whose gate fidelity specifications — confirmed by an independent Sandia National Laboratories evaluation and published in Nature this month — represent the current frontier of commercial quantum hardware.

The announcement gives Quantinuum (Nasdaq: QNT), which completed its Nasdaq IPO at $60 per share in June 2026, its first named aerospace use case on its publicly traded platform. For engineers who spend careers waiting for simulation tools capable of handling high-Reynolds-number turbulence accurately, the partnership represents something more practical than a press release: a concrete test of whether quantum-classical hybrid workflows can contribute anything real to a class of problem that has defeated supercomputers for decades.

Gas Turbine Simulation Hits the Wall Classical Supercomputers Cannot Clear

Computational fluid dynamics governs how aerospace engineers design more efficient, quieter, and lower-emission jet engines. Every iteration of a turbine blade — its geometry, its cooling passages, its interaction with rotating components and high-pressure combustion gases — is modeled in simulation before metal is cut. The governing equations, the Navier-Stokes equations, describe fluid flows accurately. The problem is computational cost.

High-Reynolds-number turbulence — the kind inside a running jet engine — is characterized by chaotic dynamics, cascades of eddies spanning a vast range of length and time scales, and nonlinear interactions that make direct numerical simulation (DNS) intractable for any realistic engine geometry. Any flight vehicle large enough to carry a human moving faster than 72 km/h has a Reynolds number exceeding 4 million; transport aircraft wings operate at Reynolds numbers of around 40 million. DNS is computationally viable only at moderate Reynolds numbers. Above that threshold, engineers rely on turbulence models — Reynolds-averaged Navier-Stokes (RANS), large eddy simulation (LES) — that approximate the physics at the cost of accuracy.

That accuracy gap is the problem Rolls-Royce, which supplies turbofan engines to civil aviation fleets and defense operators worldwide, spends significant engineering resources trying to close. Every fraction of a percentage point improvement in turbine efficiency or an emissions prediction that better matches reality has commercial and regulatory value. The collaboration frames quantum computing as a possible path to closing that gap — not by replacing classical HPC but by augmenting it for specific subroutines inside the larger simulation pipeline.

Helios's Architecture: Why Trapped-Ion All-to-All Connectivity Matters for Turbulence

The choice of Helios as the test bed is technically motivated. Quantinuum's third-generation quantum processor is built on the quantum charge-coupled device (QCCD) architecture, which physically moves ions between specialized zones on a microfabricated chip rather than relying on fixed wiring. The hardware core uses barium-137 (¹³⁷Ba⁺) hyperfine clock states as qubits — the first commercial application of barium ions, which offer longer coherence times (on the order of seconds), faster optical transitions for gate operations, and cleaner integration with multi-zone laser setups than the ytterbium ions used in earlier generations.

Helios achieves average single-qubit gate infidelities of 2.5×10⁻⁵ and two-qubit gate infidelities of 7.9×10⁻⁴ across all operational zones, with a state preparation and measurement fidelity of 3.3×10⁻⁴ — metrics independently verified by Sandia National Laboratories and published in Nature (vol. 655, 2026). The system provides all-to-all qubit connectivity through a rotatable ion storage ring connecting two quantum operation regions via a junction, which allows any qubit to interact with any other without inserting additional SWAP gate operations.

That connectivity property matters specifically for turbulence simulation. CFD subroutines — particularly linear systems solves and pressure-correction steps — often involve densely connected mathematical graphs where many variables must interact simultaneously. Superconducting quantum processors from IBM and Google use nearest-neighbor architectures that require SWAP chains to link distant qubits, adding circuit depth and error probability for densely connected problems. Helios sidesteps this overhead by design — a concrete architectural reason why Rolls-Royce and its partners chose this hardware for this problem.

What Hybrid Quantum-HPC Actually Requires — and Where It Can Still Fail

Nobody in this partnership is claiming a near-term quantum computer will independently outrun a classical supercomputer on a full CFD workload. The actual technical bet is more specific and more difficult: that certain subroutines within a turbulence simulation can be offloaded to Helios with an accuracy or speed advantage, while the surrounding computational pipeline continues to run on EPCC's classical HPC infrastructure.

The obstacle this collaboration is designed to test is precisely the one that has made quantum CFD harder than early roadmaps suggested. Lapworth himself — Rolls-Royce's Fellow in Computational Science and the company's named lead on this project — published research in 2022 embedding the Harrow-Hassidim-Lloyd (HHL) quantum linear systems algorithm into a CFD workflow, targeting the pressure-correction solve inside the SIMPLE incompressible flow algorithm. His analysis showed that the classical preprocessing required to decompose CFD matrices into linear combinations of unitaries can dominate total runtime and, on near-term systems, potentially eliminate any quantum speedup. This is the specific known engineering problem the collaboration is now approaching with real hardware.

EPCC's role is to build the software interfaces that connect quantum and classical layers — what the partnership calls hybrid workflow integration. Its contribution includes exploring how different parts of an algorithm can be compiled, emulated, and executed across classical and quantum resources, including the pre- and post-processing steps that determine whether any quantum speedup survives the transition from theory to hardware. Oliver Thomson Brown, Quantum Group lead at EPCC, described the center's positioning: the project aligns with EPCC's mission to accelerate the effective use of novel computing across industry and academia, and with the goals of the UK's first National Supercomputing Centre.

Riverlane's Deltaflow: Error Correction as the Critical Layer

Riverlane's inclusion addresses the layer of quantum computing that most product announcements prefer to leave unexamined: the error problem. Physical qubits are inherently noisy; without aggressive error correction, computations degrade faster than they produce useful results. Riverlane has built Deltaflow, a real-time, hardware-agnostic quantum error correction stack that sits between classical control hardware and a quantum processor and corrects errors in real time. The company has raised more than $120 million and partners with more than 60% of quantum computing companies worldwide.

Deltaflow 2 is already deployed at Oak Ridge National Laboratory, connected to the Frontier supercomputer, marking the first dedicated real-time QEC system integrated with a US national lab supercomputer. In April 2026, Riverlane demonstrated QEC decoding latency 10 times faster than Google's published results, processing quantum error syndromes with sub-shot latency well below Google's published 63 microseconds. These are not abstract metrics in the context of the Rolls-Royce collaboration: latency in error correction directly determines how many error-free operations a quantum system can execute before the computation collapses, which determines whether subroutine offloading gains anything on noise-sensitive CFD calculations.

Riverlane's roadmap targets the MegaQuOp — one million error-free quantum operations — by the end of 2026, a threshold the industry has identified as the inflection point at which quantum circuit complexity exceeds what any classical supercomputer can simulate. The Rolls-Royce collaboration gives that milestone an industrial application to test against, rather than a synthetic benchmark.

Five Years of Prior Work and What Comes Next

Today's agreement does not start from nothing. The project builds on roughly five years of prior collaboration between Rolls-Royce, Riverlane, and EPCC — including participation in an Innovate UK–funded £7.5 million quantum error correction consortium targeting aerospace simulation — that laid the algorithmic, error correction, and data-requirement groundwork for tackling gas turbine CFD with commercial quantum hardware. The new agreement adds Quantinuum's hardware to that existing foundation.

The near-term test bed is Helios. The medium-term target is Sol, a 192-physical-qubit system Quantinuum is developing for a 2027 target launch that would operate approximately twice as fast as Helios and with greater error correction capacity, using a fully two-dimensional grid architecture as the "scalability launching point" for larger machines. Beyond that lies Apollo, planned for around 2029, which Quantinuum describes as its first fully fault-tolerant system with thousands of physical qubits and hundreds of logical qubits, designed to execute circuits with millions of gates.

Leigh Lapworth put the timeline rationale plainly: applications development is a multi-year activity, and for the partnership to benefit from teraQuOp-scale devices when they arrive, algorithm co-development must begin now, alongside the hardware and software.

Inside the UK Quantum Strategy This Collaboration Is Built to Serve

All four partners are either headquartered or substantially based in the United Kingdom, and the collaboration is positioned within the UK government's quantum computing mission. That mission, officially funded as part of the UK's Digital and Technology Sector Plan and backed by £670 million toward quantum computing development, targets accessible UK-based quantum computers capable of running one trillion error-free operations — the teraQuOp threshold — by 2035.

The UK committed £2.5 billion to quantum technologies across a 10-year National Quantum Strategy, with the goal of positioning the country as a quantum-enabled economy by 2033. The Rolls-Royce collaboration ties the most commercially grounded part of that mission — industrial manufacturing and aerospace engineering — to concrete hardware milestones on a publicly listed company's roadmap. That linkage between public funding ambition and private sector execution is what distinguishes this partnership from prior announcements of collaboration intent without named hardware and without named algorithms.

Frequently Asked Questions

What is the specific technical problem this collaboration is trying to solve?

Gas turbine design requires highly accurate simulation of turbulent fluid flows — specifically, high-Reynolds-number turbulence inside rotating blade geometries. Classical supercomputers cannot run direct numerical simulation (DNS) at the Reynolds numbers relevant to real aircraft engines; they rely instead on turbulence models that approximate the physics at some cost to accuracy. The collaboration is testing whether specific computational subroutines inside those simulations — particularly linear systems solves and pressure-correction steps — can be offloaded to Quantinuum's Helios quantum processor with a net accuracy or speed advantage when combined with EPCC's classical high-performance computing infrastructure.

What is quantum error correction and why does the Rolls-Royce collaboration specifically depend on it?

Physical qubits are inherently noisy; without active error correction, quantum computations collapse under accumulated errors long before producing useful results. Quantum error correction (QEC) encodes one reliable "logical" qubit in many physical qubits, then continuously measures and corrects errors mid-computation without disturbing the computation itself. For CFD subroutines to run usefully on Helios, the error rates must be low enough that the quantum calculation completes with sufficient accuracy before noise overwhelms it. Riverlane's Deltaflow stack provides real-time QEC with sub-shot latency that outperforms published benchmarks by an order of magnitude — which is why the collaboration specifically includes Riverlane rather than treating error correction as the hardware vendor's problem alone.

Can quantum computers simulate turbulence yet?

Not independently, and not at engineering-relevant Reynolds numbers. The honest answer from all four partners is that this collaboration is designed to find out whether a hybrid approach — quantum hardware handling specific subroutines, classical supercomputers handling the rest — can produce any meaningful advantage over purely classical simulation. The critical unknown is the state preparation cost: loading classical turbulence data into quantum states can, in the worst case, consume more computational time than the quantum calculation saves. The collaboration is specifically structured to test and potentially overcome this bottleneck on real hardware, using algorithms and error correction tools developed over the past five years of prior Rolls-Royce, Riverlane, and EPCC work.

What does Quantinuum's Helios hardware actually do differently from earlier quantum computers?

Helios uses barium-137 ions as qubits — the first commercial deployment of this ion species — suspended in an electromagnetic trap and physically moved between storage and processing zones on a chip. Its QCCD architecture provides all-to-all connectivity: any qubit can interact with any other without inserting additional gate operations, unlike superconducting quantum processors where qubits only communicate with nearest neighbors. It achieves single-qubit gate infidelities of 2.5×10⁻⁵ and two-qubit gate infidelities of 7.9×10⁻⁴, verified independently by Sandia National Laboratories and published in Nature this month — the lowest error rates of any commercial quantum hardware modality.