What Willow demonstrated: a stronger quantum memory, with a correction to keep in view

Reported evidence

What the sources report

The Nature paper reports a surface-code memory using 101 qubits on a 105-qubit Willow processor. The device count and the memory's participating-qubit count describe different things.

The authors observed lower logical error as they increased the code distance under the tested conditions. The published experiment focused on preserving quantum information.

The later author correction changes the repetition-code figure's axis and replaces misleading code-distance labels in its legend with labels identifying earlier work and this work.

IBM's architecture explanation identifies a universal set of logical operations and supporting classical decoding as requirements beyond a quantum memory.

Sources: [1] [2] [3]

Our interpretation

Investor relevance — interpretation

A useful progress record should identify the encoded information, code, error measure, workload and resource count. Putting only a processor's qubit total in a ranking hides the work those qubits performed.

This result addresses a crucial engineering question: whether protection improves as the encoding grows in the tested regime. It is a building block for a future computer, not evidence that an arbitrary useful business calculation has been completed.

Keep the surface-code memory result separate from the repetition-code experiment in the corrected figure. A source update belongs next to the explanation so readers can distinguish the two.

For investors, the next questions are how logical operations scale, what resources a useful task needs and what customers would pay for it. This paper alone does not establish product revenue or an investment return.

Sources: [1] [3] [2]

Our interpretation

What this does not establish

Peer review records expert scrutiny of the research; it is not independent replication, a general-purpose commercial advantage test or a guarantee of scalable economics.

The correction describes figure-label changes, not a retraction. Do not expand it into an unsupported claim that the headline memory result was reversed.

Sources: [1] [2]

Background

Evidence date and disclosure

Evidence event date: 2024-12-09. AI-assisted and source-linked. Issuer reports are attributed as issuer reports. The publisher may hold positions in companies covered; see Disclosures.

Sources: [1] [2] [3]