Skip to main content
KetQat

Appearance

What is measured, what is promised

Three kinds of statement get used interchangeably in quantum planning, and confusing them is the most common way a resource estimate misleads somebody. They are kept apart here, in that order.

A measurement
Somebody ran an experiment and published the numbers, on a date, with a citation. Immutable here: a newer measurement is a new record, never an edit, because a decision made against the old figure should stay auditable.
A roadmap target
A vendor-stated target, not a prediction, a guarantee, or evidence that any system exists. KetQat does not forecast whether or when a target will be met.
Your assumption
What you told KetQat to assume. It is labelled USER_PROVIDED on every figure derived from it, and it is the only one of the three that enters an estimate's arithmetic.

Measured

One entry, and that is not an oversight. A device figure transcribed without a dated primary source is indistinguishable from a guess once it is on the page, and this project has measured no hardware itself.

Google Quantum AI — Superconducting surface-code memory

Measured 2024-08-24

A distance-7 surface-code memory operating below threshold: enlarging the code reduced the logical error rate rather than increasing it.

Physical qubits in the code
101 qubitsas stated: “101-qubit distance-7 code”
Logical error per cycle
0.00143 per cycleas stated: “0.143% ± 0.003% error per cycle of error correction”
Error suppression factor per distance step
2.14 Λas stated: “suppressed by a factor of Λ = 2.14 ± 0.02”
Surface-code cycle time
1,100 nsas stated: “a cycle time of 1.1 μs”

What this does not establish

  • A memory experiment. It stores a logical qubit; it does not run a logical algorithm, and it performs no magic-state distillation.
  • 101 physical qubits is the code patch, not a machine. A computation needing many logical qubits needs many such patches plus routing space and factories.
  • The cycle time is this device's, under this experiment's conditions. It is not a property of the surface code.

Google Quantum AI and collaborators — Quantum error correction below the surface code threshold — published 2024-08-24, read 2026-08-14.

Promised

These are targets

A vendor-stated target, not a prediction, a guarantee, or evidence that any system exists. KetQat does not forecast whether or when a target will be met.

They are carried because somebody planning a multi-year programme will encounter these figures anyway, and reading them beside a plain statement of what they are beats reading them in a press release.

IBM — IBM Quantum Starling

Target 2029

IBM's stated target for a large-scale fault-tolerant system.

Logical qubits
200 logical qubitsas stated: “200 logical qubits”
Quantum gates in a circuit
100,000,000 gatesas stated: “100 million quantum gates”

What this does not establish

  • A target, not a measurement. Nothing here reports that this system exists or that this figure has been achieved.
  • The announcement states logical qubits and a gate count. It does not state a physical error rate, a cycle time, or a magic-state throughput, so a resource estimate cannot be evaluated against it.

IBM Sets the Course to Build World's First Large-Scale, Fault-Tolerant Quantum Computer at New IBM Quantum Data Center — published 2025-06-10, read 2026-08-14.

IBM — IBM Quantum Blue Jay

Target 2033

IBM's stated longer-range target.

Logical qubits
2,000 logical qubitsas stated: “2,000 logical qubits”
Quantum gates in a circuit
1,000,000,000 gatesas stated: “one billion gates”

What this does not establish

  • A target, not a measurement.
  • Stated as logical qubits and a gate count only. The parameters an estimate depends on are not given.

IBM lays out clear path to fault-tolerant quantum computing — published 2025-06-10, read 2026-08-14.

What this page will not do

No function combines an observation with a roadmap target, and none ranks vendors. They are different kinds of statement, not different amounts of one.

There is no “best available device”, no vendor ranking, and no date at which a workload becomes feasible. Producing any of them would mean deciding that one vendor's published measurement and another's announced intention are comparable quantities.

What a stated target can and cannot tell you works this through against real reference workloads.

A gap between what a workload needs and what a record states is reported as a distance, never as a schedule. Comparing a logical-qubit requirement against a stated logical-qubit figure is one requirement against one number — it says nothing about error rate, cycle time, or magic-state throughput, and on most workloads the factory is what binds.

Read how the estimates are produced, how KetQat compares to other estimators, or what KetQat will not tell you.