98 Qubits, One Transform: Mitsui and Mitsubishi Electric Push QFT Benchmarks on Quantinuum Helios

Scaling the Fourier transform — on hardware investors can audit

Mitsui & Co. and Mitsubishi Electric have published one of the largest approximate Quantum Fourier Transform (QFT) demonstrations to date on Quantinuum's 98-qubit Helios trapped-ion processor, running the algorithm on 98 physical qubits and extending to 12 logical qubits under Steane error correction.

Quantinuum described the work in a July 29, 2026 blog post; industry outlets reported it on August 14, 2026. The technical details appear in a white paper, Experimental Evaluation of the Quantum Fourier Transform on a Trapped-Ion Quantum Computer.

Why QFT benchmarks matter

The Quantum Fourier Transform is a foundational primitive underlying Shor's algorithm, quantum phase estimation, and many chemistry simulations. Demonstrating QFT at larger scale with measurable fidelity is one way to assess whether hardware is approaching industrial utility — rather than relying on abstract qubit counts alone.

Helios, launched commercially in November 2025, uses barium ions with all-to-all connectivity via a QCCD architecture. Quantinuum reports a typical two-qubit gate error rate of approximately 0.08%.

Physical and logical results

The joint team used Quantinuum's Guppy hybrid programming language and pytket compilation toolchains to evaluate cross-layer trade-offs between physical gate noise, code distance overhead, and post-selection acceptance rates.

On physical qubits, the white paper documents QFT scaling to 98 qubits — a clear progression from prior published records. Using the Steane code, the team formed 12 logical qubits and executed QFT with quantum error correction interwoven into the algorithm.

For logical qubits, error-detection post-selection yielded fidelity comparable to or higher than error correction alone in the tested regime — but acceptance rates fell from 83% to 8% as logical qubit count increased, illustrating the fidelity-versus-throughput trade-off.

Industrial read-through

Mitsui and Mitsubishi Electric are not academic curiosities in this context — both are industrial conglomerates evaluating quantum readiness for future finance, materials, and logistics workloads. Algorithm-level benchmarking on commercial hardware gives enterprises a concrete lens on what today's systems can execute.

Quantinuum positions Helios as achieving 99.921% average two-qubit gate fidelity, exceeding the widely cited "three 9s" threshold. The QFT work sits alongside Quantinuum's Oracle Cloud Infrastructure partnership (August 11, 2026) to deploy Helios as a managed hybrid quantum-classical service.

What this does not claim

The white paper presents quantum estimates as the most credible among approaches tested — not independently established ground truth. Classical simulation methods showed non-convergence or sensitivity in the hardest regimes, but IBM and others have repeatedly emphasized that advantage claims require community scrutiny.

Still, for procurement teams tracking fault-tolerant progress, a verified QFT at 98 physical qubits — plus a logical QFT at 12 qubits — is a measurable step beyond slide-deck milestones.

### Sources

• Quantinuum — Scaling the Signal: What a Larger QFT Says About Quantum Progress (July 29, 2026)
• Mitsubishi Electric — Experimental Evaluation of the Quantum Fourier Transform on a Trapped-Ion Quantum Computer (PDF) (2026)
• Quantum Computing Report — Mitsui & Co. and Mitsubishi Electric Benchmark Approximate and Logical QFT on Quantinuum Helios (August 14, 2026)