Quantinuum H2

Overview

Quantinuum H2 is the machine most often cited when someone wants to know what the best current physical qubits look like. Its two-qubit gates have been measured at roughly 99.9% fidelity, which means about one failure in a thousand operations, and it has repeated that performance across all pairs of qubits in the device. At 56 qubits it is large enough that a classical computer cannot fully simulate it, yet small enough to be carefully controlled.

H2 uses a design called QCCD, the quantum charge-coupled device. Instead of keeping every ion in one fixed chain, the trap physically moves ions between zones, so that qubits that need to interact can be brought together and then separated again.

History and background

Honeywell developed the QCCD approach and launched its first commercial system, H0, in 2020, followed by the 20-qubit H1 family. In 2021 Honeywell Quantum Solutions merged with Cambridge Quantum to form Quantinuum. H2 was introduced in May 2023 with 32 qubits and a quantum volume of 65,536 (2^16). In June 2024, Quantinuum upgraded it to 56 qubits and ran a random circuit sampling experiment with JPMorgan Chase at a scale a classical computer cannot fully simulate.

Several partner experiments followed. In April 2024 Microsoft and Quantinuum reported four logical qubits with an error rate about 800 times lower than the underlying physical qubits, and by September 2024 the teams reported 12 logical qubits on H2 and an end-to-end chemistry workflow. In 2025 a certified randomness experiment with JPMorgan Chase and national laboratories used H2, and in September 2025 the machine reached a quantum volume of 33,554,432. The next-generation Helios system, launched in late 2025, is now Quantinuum's flagship, and H2 remains in service for research.

How it works

Each qubit is a single ion held in an electromagnetic trap on a chip. Voltages applied to many small electrodes create moving potential wells, so ions can be shuttled along the trap, split into groups, and recombined. H2 has a racetrack-shaped loop with a junction. Gate zones are regions with laser access where two ions are brought together for an entangling operation, while memory zones keep other ions parked safely.

Because ions travel, any qubit can be moved next to any other, which gives all-to-all connectivity without the swap gates other architectures use. The price is time: shuttling and re-cooling take hundreds of microseconds, so H2's circuits run far more slowly than those on superconducting machines. Quantinuum argues that for error correction, fidelity matters more than raw clock speed, because a fast but noisy gate cannot be corrected without a huge overhead.

H2 also supports mid-circuit measurement and reset, in which some qubits are read out and re-initialized while the others keep their quantum state. Combined with real-time classical feedback, this makes it possible to run error-correcting codes, repeat-until-success protocols, and algorithms that branch on measurement results.

Connectivity

All-to-all through physical ion transport: ions are shuttled so that any pair can be brought together

Specifications in detail

Key features

• 56 fully programmable qubits with all-to-all connectivity
• Two-qubit gate fidelity around 99.9%, among the best in the industry
• Quantum volume record of 2^25 (33,554,432) reported in September 2025
• Real-time mid-circuit measurement and feedback
• Used with Microsoft to demonstrate logical qubits with error rates far below the physical qubits

Limitations

• Ion shuttling and gate operations are much slower than on superconducting chips
• 56 qubits is modest compared with some other technologies
• Scaling the QCCD architecture to thousands of qubits requires more complex traps and laser delivery
• Quantum volume is a single-number benchmark that does not capture every use case

Timeline

What it is used for

• Error-correction and logical-qubit experiments
• Quantum chemistry and materials simulation, including with the InQuanto software
• Certified randomness and cryptographic research
• Condensed-matter simulations such as Floquet dynamics, run at 56-qubit scale
• Benchmarking and algorithm development in the cloud

How it compares

• Compared with IonQ Forte, H2 has more qubits (56 versus 36) and higher reported two-qubit fidelity, and both have all-to-all connectivity.
• Compared with Google Willow, H2 has lower clock speed but fidelity per gate is several times better, and it has full connectivity instead of a nearest-neighbor grid.
• Compared with IBM Heron, H2 has about a third as many qubits but one order of magnitude lower two-qubit error and no need for swap gates.

Technical notes

Quantum volume (QV) tests the largest square random circuit a machine can run with reasonable accuracy. A QV of 2^n means an n-qubit, depth-n circuit passes the test. Quantinuum has raised H-series QV roughly tenfold per year since 2020, reaching 2^23 in spring 2025 and 2^25 in September 2025. QV is IBM's invention and a demanding benchmark, but it says little about whether a machine can run a particular application.

Why it matters

H2 shows what high-fidelity qubits make possible. Many of the first convincing logical-qubit experiments, including those with Microsoft, ran on Quantinuum hardware, and its results are a benchmark for what error correction can achieve on real machines.

It is also a reminder that quantum progress can come from quality instead of size. A 56-qubit machine with very low error has been able to run experiments that larger, noisier machines cannot match.

Frequently asked questions

Keep exploring

• Company profile: Quantinuum
• Related processor: IonQ Forte (and Forte Enterprise)
• Related processor: IonQ Tempo
• Related processor: Google Willow
• Related processor: IBM Heron (r1, r2, and r3)

This profile is written for educational purposes and summarizes company announcements, published papers, and vendor specification sheets. Last reviewed: September 30, 2026. Figures marked as company-reported have not all been independently verified, and benchmarks differ between vendors, so compare them with care.

← Back to Hardware Database