> ## Content Index
> Fetch the complete content index at: https://pvweekly.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Willow made a larger quantum memory a more reliable one
- URL: https://pvweekly.com/willow-error-correction-threshold/
- Published: 2026-10-27T19:29:57.000Z
- Updated: 2026-10-27T19:29:57.000Z
- Description: 2024 archive · Google’s December 2024 result strengthened the engineering case for error correction while leaving useful commercial computation ahead.
- Author: PV Weekly Editorial
- Tags: Technology & Defense, Archive 2024, #ledger-format, #archive-backfill-20261027

**2024 archive** · December 9, 2024Reporting prepared October 22, 2026\. This retrospective describes events and expectations from the period covered.[Browse the archive](https://pvweekly.com/archive/)

More hardware can make a quantum computer worse. Every additional qubit brings another fragile component and another opportunity for an error. Google’s December 9, 2024, Willow announcement mattered because its error-correction experiment demonstrated a way to reverse that relationship.

The accompanying Nature paper, published online that day, reported that increasing the size of an encoded quantum memory reduced its error rate. This was a specific experimental result about protecting information, with consequences for the construction of larger machines. It did not establish that a commercially useful quantum computer had arrived.

The distinction begins with what the machine was protecting. A physical qubit is a component. A logical qubit distributes quantum information across a collection of physical qubits, with additional measurements used to detect errors. The apparatus must preserve the encoded information while repeatedly checking for damage.

Google’s researchers used a surface code, arranging data qubits in square grids and using measurement qubits to check neighboring components. They expanded the data grids from three by three to five by five to seven by seven. At each step, the encoded error rate fell by roughly half. That pattern is the practical meaning of operating below the error-correction threshold: the protection gained from a larger code outweighs the extra faults its hardware introduces.

The paper’s larger memory experiment used 101 physical qubits and reported an error probability of 0.143 percent per correction cycle. Its logical memory lasted approximately 2.4 times as long as its best constituent physical qubit. Those measurements gave engineers something more informative than a chip’s raw qubit count: evidence that a group of imperfect components could preserve information better than any one of them.

There was also a conventional-computing problem inside the quantum result. Measurement records have to be interpreted by a decoder, and they arrive continuously. A machine that generates diagnostic information faster than it can process it accumulates an operational backlog. In a separate, smaller surface-code experiment, the team demonstrated real-time decoding, reporting an average latency of 63 microseconds. Keeping up with the stream and minimizing the delay of individual decisions remained distinct engineering requirements.

Google also announced an impressive performance comparison using random circuit sampling. That benchmark attracted attention because of the company’s enormous estimate for the classical computing time needed to match it. But Google explicitly acknowledged that the sampling benchmark had no known real-world applications. A benchmark advantage and a useful application advantage were separate accomplishments; the announcement established neither their equivalence nor a delivery date for the latter.

For hardware developers, the December result sharpened the next question. Could fabrication, control electronics and decoding maintain comparable performance as the machine grew far beyond the demonstrated devices? Google’s technical account identified substantial remaining overhead and decoder delays. Scaling required improvements across the system, not simply another fabrication run with more components.

Willow’s historical importance lay in making that larger engineering effort more credible. Error correction had delivered a measurable benefit as the encoded memory grew. Turning that benefit into a reliable machine capable of sustained, useful computation remained the work ahead.

## Sources

- [Quantum error correction below the surface code threshold](https://www.nature.com/articles/s41586-024-08449-y?ref=pvweekly.com) — Google Quantum AI and Collaborators / Nature · 2024-12-09.
- [Making quantum error correction work](https://research.google/blog/making-quantum-error-correction-work/?ref=pvweekly.com) — Google Research · 2024-12-09.
- [Meet Willow, our state-of-the-art quantum chip](https://blog.google/innovation-and-ai/technology/research/google-willow-quantum-chip/?ref=pvweekly.com) — Google · 2024-12-09.