Rise of the Logical Qubit

Along with the increasing maturity of quantum computing, we are witnessing an ongoing evolution of metrics used to measure performance. Historically, physical qubit count has been the number at the forefront of headlines – building a machine with 10, 100, and 1000 qubits was a surefire way of getting the attention of the entire quantum community.

As we see QPUs with thousands and eventually millions of qubits being developed, those achievements will be no less valuable. However, now there is a clear understanding that physical qubits don’t tell the entire story. 100 qubits mean something completely different depending on if they’re superconducting, photonic, or topological.

Pairing physical qubit counts with fidelities paints a far better picture of actual performance, and that is what we’re seeing with the increased prominence of logical qubits as a metric – a way to measure not just how many qubits a system has, but also how “good” they are. As we will show below, logical qubits are becoming the next “mandatory” metric that companies need to include in their vision, or at least acknowledge.

First things first – what is a logical qubit?

Source – Nature.

In order to reliably run algorithms on a quantum system naturally prone to decoherence, quantum error correction (QEC) schemes are necessary to account for the arising physical errors. Once 2-gate fidelities are approximately at 99% and above (or error rates go below 1%), it becomes possible to group physical qubits through QEC into logical qubits.

Logical qubits are an abstraction – it’s a group of physical qubits working together to “imitate” a fault-tolerant qubit necessary for algorithms. The higher the fidelities, the fewer physical qubits are necessary to build a single logical qubit. For superconducting qubits, the typical ratio is 100s or 1000s of physical qubits to 1 logical, for trapped ions the ratio ranges from 10s to 100s. Topological qubits promise to be so perfect that only a handful of physical qubits would be equivalent to a logical one.

Companies and researchers are embracing this metric.

In the last few years logical qubits have entered the language companies use when describing their vision, in particular – their roadmaps. Below are just a few examples of roadmap changes (sourced both from company websites and the Wayback Machine).

IBM: Included logical qubits in their 2033+ milestone

BEFORE: Jun 2023

AFTER: Mar 2025

Honeywell/Quantinuum: Added a row showing logical qubit count alongside physical qubits

BEFORE: 2021

AFTER: Mar 2025

QuEra: Added logical qubits to Step 2 of the quantum roadmap

BEFORE: Oct 2023:

AFTER: Mar 2025:

The growing trend is also clearly visible in research papers: universities like Yale, Delft, and Tsinghua are more actively researching logical qubits:

Research Papers referencing logical qubits over time (2015-2024)

Source: The Quantum Insider platform. Search string – “logical” AND “qubit”

And the same growth can be seen in patent output:

Patents referencing logical qubits over time (2015-2024)

Source: The Quantum Insider platform. Search string – “logical” AND “qubit”

How many logical qubits to expect?

As more and more companies include logical qubits in their roadmaps, we get a much better picture of what scale of systems to expect. Below are sample milestones from quantum computing companies displaying both logical and physical qubit counts. The farthest milestones were taken to display the technology in its maturity.

Source: Company websites, March 2025

Most notably, a milestone of ~100 logical qubits by ~2030 is something many companies are targeting, regardless of modality.

Is that a lot?

In order to break RSA with N bit encryption, some implementations have shown that ~2N logical qubits would be needed. For RSA-2048 that puts the qubit requirement at ~4,000 logical qubits, placing the 2030 milestones still a way off.

For simulations, though, 100 logical qubits could be much more significant. The Jülich Universal Quantum Computer Simulator holds the world record for the highest number of classically simulated qubits – 48. Having a quantum machine with 49 logical qubits wouldn’t automatically signal an industry disruption, but it would open up the territory for simulations outside of the reach of classical systems, offering new insights in areas such as chemistry and materials science.

Closing thoughts

Ultimately, the best metric of a quantum computer will be performance (speed, price, availability) in real-world applications. If you’ve ever built a gaming PC and shopped for GPUs you know that all the Ray Tracing cores and high-speed VRAM mean very little until you actually boot up a game you want to play and see how well it runs (that’s assuming you can even buy a GPU in the current market – good luck if you’re trying).

That doesn’t mean that intermediary metrics can’t be helpful. Logical qubits make for a better metric than simply talking about qubit counts, and it makes the zoo of quantum computers much more comparable. The growing amount of companies using the metric further validates its value.

Other honorable mentions include:

• Quantum Volume – the maximum size of square quantum circuits that a quantum computer can implement. Used almost exclusively by Quantinuum (Honeywell) and up until recently IBM. The metric relies on classical machines to calculate, which will be increasingly problematic as quantum systems grow.
• Quantum Operations per Second (QuOps) – analogous to classical FLOPs and, as is appropriate for a great sounding word, is popular in the country with the best form of English: both Oxford Quantum Circuits and Riverlane use QuOps in their roadmaps. Though a useful metric, it is still rarely used, making comparisons challenging.

–Alan Kanapin, Analyst at the Quantum Insider