Google's Willow chip shows 'magic state cultivation' working in a quantum error-correction test
Google Quantum AI researchers report in Science that growing 'magic states' on the Willow chip cut errors about 40-fold, a more efficient route to error-corrected quantum operations. Outside experts call it a real step but note that most attempts are discarded and error rates remain far from practical needs.
1 / 2
The story, neutrally told
Mixed · 2Researchers at Google Quantum AI used the company's Willow quantum chip to show experimentally that a technique called magic state cultivation can supply the special quantum states needed for error-corrected operations more efficiently than the standard method. New ScientistN “The new experiment by Rosenfeld and her colleagues proves that magic state cultivation works in practice.” Read at New Scientist ↗ Scientific AmericanLC “the researchers show how to supply magic states more cheaply using a gardening-inspired technique called cultivation.” Read at Scientific American ↗ Left · 1The study, led by Emma Rosenfeld, was published in Science on 15 October 2026, according to Scientific American. Scientific AmericanLC “In a new study published today in Science”“Emma Rosenfeld, a researcher at Google Quantum AI and the study’s lead author.” Read at Scientific American ↗ Mixed · 2Magic states are needed to run non-Clifford gates, such as the T gate, which unlock a quantum computer's advantage over ordinary computers; any flaw in the state becomes a flaw in the gate. Scientific AmericanLC ““But there is one particular gate that unlocks the quantum computer’s power,” Rosenfeld says.” Read at Scientific American ↗ New ScientistN “a special quantum state that qubits must assume before the quantum computer can run error-corrected non-Clifford operations” Read at New Scientist ↗
Mixed · 2The standard way to make them, distillation, turns many imperfect states into fewer accurate ones; Rosenfeld said up to 95 per cent of a quantum computer can be spent on this, and that it eats up a lot of qubits and time. Scientific AmericanLC ““You can have up to 95 percent of your quantum computer being spent just generating these high-fidelity magic states,” Rosenfeld says.” Read at Scientific American ↗ New ScientistN ““It eats up a lot of qubits and a lot of time,” she says.” Read at New Scientist ↗ Mixed · 2Google colleagues proposed cultivation in 2024. It grows one state within a small patch of qubits, checks it and then spreads it across more qubits for stronger protection. New ScientistN “the magic state is “grown” within a smaller patch of qubits.” Read at New Scientist ↗ Scientific AmericanLC “States that pass this test are then spread across more qubits for stronger protection” Read at Scientific American ↗ Mixed · 2In the experiment, which used 54 of Willow's 105 physical qubits, the team asked whether the state was the right magic state rather than measuring it as 0 or 1, which would collapse it. Both outlets report that errors fell about 40-fold; Scientific American adds that about four attempts in 1,000 produced a faulty state without cultivation. Scientific AmericanLC “Without cultivation, about four attempts in 1,000 produced a faulty state. With it, errors fell by roughly 40-fold. The experiment used 54 physical qubits out of Willow’s 105.” Read at Scientific American ↗ New ScientistN “Their new technique ultimately reduced Willow’s errors 40 times.” Read at New Scientist ↗
Mixed · 2Independent researchers praised the result. Kenneth Brown of Duke University said it shows cultivation works in practice. Isaac Kim of UC Davis called it an impressive demonstration that can make fault-tolerant quantum computing substantially cheaper. Dominik Hangleiter of ETH Zurich said it could be a 'gamechanger' for mid-sized machines with hundreds of logical qubits. Scientific AmericanLC ““The Google experiment shows that cultivation works in practice,” says Kenneth Brown” Read at Scientific American ↗ New ScientistN ““It's an approach that can make fault-tolerant quantum computing substantially cheaper.””“could be a "gamechanger" in terms of building intermediately sized quantum computers with hundreds of logical qubits” Read at New Scientist ↗ Mixed · 2Caveats are also reported. Only about 8 per cent of runs produced high-quality magic states, because a quality-assurance step discards 92 per cent of attempts. New Scientist notes Rosenfeld said this partly reflects a careful benchmarking method that may not be used in future applications and should improve with refinement and better hardware. Scientific AmericanLC “discarding 92 percent of attempts” Read at Scientific American ↗ New ScientistN “only 8 per cent of Willow's computations produced high-quality magic states”“It is, in this case, also affected by a very careful benchmarking technique that the team used” Read at New Scientist ↗ Mixed · 2The result is still far from practical requirements. Scientific American reports the team is at roughly one error in 10,000 attempts, while breaking RSA-2048 encryption would need states failing once in 10 billion tries, and molecule simulation one error in a million to one in 100 million. Zhenyu Cai of Imperial College London said Willow's errors must fall further before it can tackle useful problems. Scientific AmericanLC “Right now the team stands at roughly one error in 10,000 attempts.”“would require magic states that would fail only once in 10 billion tries” Read at Scientific American ↗ New ScientistN “Willow's errors will need to be reduced further before it can tackle truly useful problems” Read at New Scientist ↗
Centre · 1New Scientist adds that trapped-ion experiments have also lowered error rates with new magic-state methods. Willow's superconducting qubits reached similar error rates but run faster, which could shorten overall runtime, and Google's 2024 analysis suggested cultivation could let useful programs run with about one-tenth the qubits expected. New ScientistN “Willow’s qubits, which are based on tiny superconducting circuits, achieved error rates similar to those past experiments, but Rosenfeld says that they work more quickly”“with just one-tenth the number of qubits we had expected they would need” Read at New Scientist ↗
Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.
Left1 outlet
- Framing
- Explainer on how cultivation works, using the gardening analogy, with a sober reality check on how far the result is from useful applications.
- Emphasis
- Mechanics of magic states, the 40-fold error reduction, the 92 per cent discard rate, and the gap to RSA-2048 and molecule-simulation requirements.
- Leaves out or plays down
- Does not mention the trapped-ion comparison or the 2024 one-tenth-qubits projection, or other outside experts' enthusiasm.
- Charged language
- “magic states”
- For example
-
“The result is still far from what practical applications demand.” — Scientific American
“Google’s quantum Willow chip may be able to grow ‘magic states’” — Scientific American
Centre1 outlet
- Framing
- Presents the experiment as a more efficient way to fix quantum errors, suggesting smaller near-term machines may be more impactful than expected.
- Emphasis
- Efficiency gains, the potential for hundreds of logical qubits, the comparison with ion-based experiments, and expert quotes from Kim, Hangleiter and Cai.
- Leaves out or plays down
- Gives no absolute error rate or the distance to cryptography-breaking requirements; the 8 per cent success figure appears late.
- Charged language
- “gamechanger”“error-proofing”
- For example
-
“Their work suggests the smaller quantum computers that will be available in the next few years could be more impactful than expected.” — New Scientist
““This is the first demonstration that the concepts work. We’re not done.”” — New Scientist
Right0 outlets
No right outlet in our sources has covered this story yet.
What every side reports
- Google Quantum AI used the Willow chip to demonstrate magic state cultivation experimentally.
- Errors fell about 40-fold.
- Cultivation was proposed by Google colleagues in 2024 as an alternative to distillation.
- Only a small fraction (about 8%) of runs yielded high-quality magic states.
- Error rates are still too high for practical applications.
Google organisation
Google Quantum AI researchers present cultivation as a working, cheaper alternative to distillation, while Rosenfeld stresses it is a first demonstration and that work remains.
““This is the first demonstration that the concepts work. We’re not done.”” — New Scientist
““We’re more galvanized than ever to get to these near-term scientific applications,”” — Scientific American
Left1 article
-
Google’s quantum Willow chip may be able to grow ‘magic states’
Neutral Explanatory piece that pairs the 40-fold error reduction with candid limits on discard rate and practical thresholds.
Centre1 article
-
Google showcases way to fix quantum computer errors more efficiently
Sympathetic Positive account of an efficiency advance with supportive expert quotes, tempered by caveats on success rate and remaining errors.
Right0 articles
No coverage yet.
- 15 Oct 19:00 First New ScientistN Google showcases way to fix quantum computer errors more efficiently
- 15 Oct 19:00 First Scientific AmericanLC Google’s quantum Willow chip may be able to grow ‘magic states’
Times are when each article was published, or when we first saw it if the outlet gave no time.
Ask about this story
Loading…