Quantum Startup Sets New Record — And Isn't Pretending It's a High Bar
A quantum computing startup and its partner say they have run the most complex fluid simulation ever on real quantum hardware. They are. The previous record, set by the same companies just months ago, was three steps. The bar is low, and the companies are being honest about that, which is more than can be said for most quantum computing announcements.
Quanscient, a Finnish startup founded in 2021, and Haiqu, a quantum middleware startup whose CTO is Mykola Maksymenko, published results on March 3 in a paper posted to the preprint server arXiv arXiv:2603.02127. Their algorithm, called OSSLBM (One-Step Simplified Lattice Boltzmann Method), ran a 15-step nonlinear fluid benchmark with an obstacle on IBM's Heron R3, the chip IBM markets as its largest available quantum processor The Quantum Insider. The paper describes the result as "the most physically complex, publicly documented variant of a Quantum Lattice Boltzmann Method hardware demonstration to date."
That claim is defensible, and not only because the bar is low. The prior result, posted to the AWS Blog in July 2025, showed three steps of quantum lattice Boltzmann dynamics on an IonQ Aria 1 processor: a 64-by-64 grid with 16 qubits and a circuit depth of 802 gates AWS Blog. Going nonlinear on top of that represents a meaningful step. The OSSLBM approach simplifies the collision operator in the underlying LBM equation, cutting the qubit count and circuit depth required compared to the full QLBM formulation. The 2025 run used advection-diffusion physics, which models how a scalar quantity like temperature moves through a fluid. The 2026 result tackles the full nonlinear Navier-Stokes regime, where the fluid velocity feeds back into the forces acting on it. That is a harder physics problem, and the fact that it runs at all on near-term hardware is the actual signal.
Computational fluid dynamics matters. Simulating how air moves around a wing or heat disperses through a combustion chamber demands enormous classical computing resources, and the difficulty scales badly with geometric complexity. Quantum computers are theoretically well suited to the underlying mathematics, but building hardware and software capable of running physically meaningful simulations has proven stubbornly difficult. The lattice Boltzmann method breaks the simulation into discrete time steps and spatial cells. The quantum version delegates specific subroutines to quantum circuits while classical hardware handles the rest. The quantum part does not carry most of the load. This is a hybrid approach, and it is worth saying so plainly because press releases sometimes imply the quantum computer is doing more than it is.
The stakes are large enough that Airbus and BMW named Quanscient and Haiqu finalists in their joint Quantum Mobility Challenge in 2024 The Next Web. Aerospace and automotive companies have strong incentives to simulate complex airflow, and any meaningful speedup in computational fluid dynamics would have obvious commercial value.
Valtteri Lahtinen, chief scientist at Quanscient, described the significance in the company's announcement Quanscient press release. "CFD is one of the most computationally difficult branches of simulation with some of the largest impact on the world's biggest sectors," he said. He did not claim the technology is ready to replace classical supercomputers.
Oleksandr Kyriienko, a professor and chair in quantum technologies at the University of Sheffield, offered a measured independent assessment in comments to The Quantum Insider The Quantum Insider. "This is an interesting and timely contribution to quantum CFD," he said. He did not say "breakthrough," "advantage," or "milestone." That restraint is itself informative.
The gap between what quantum hardware can do today and what quantum computing advocates promise it will do eventually remains vast. Fifteen steps of nonlinear fluid dynamics on a 2026 quantum chip is genuine progress within that gap. It is not the other side of it.
The result was first reported by Quantum Computing Report on April 2, 2026.