Skip to content
Türkçe

Google · Chips and hardware · Research

Google demonstrates its fix for the quantum “magic state” bottleneck on its Willow chip

Published: 22 sourcesTürkçe

Google tested its fix for the “magic state” bottleneck on a real chip for the first time

Google Quantum AI ran “magic state cultivation,” a method it proposed on paper in 2024, on its Willow chip; the paper came out in Science on October 15. Magic states are used up by the operations that give a quantum computer its real power, and a useful machine could need millions or even trillions of them.

The old method is costly; the new one weeds out bad attempts early

The standard method, “distillation,” turns many flawed states into one clean state, but it eats up lots of qubits and time; some estimates cited by Google say up to 95% of a quantum computer’s capacity could go to this alone. Cultivation instead checks a single state over and over inside a small code and throws away every attempt that fails.

The result: 40 times fewer errors in cultivated states

Cultivated states had about 40 times fewer errors than states that skipped the process, with a fidelity of at least 99.99%, or about 1 error in 10,000. The team also moved a prepared state from the small code into a larger, more practical one, which Google describes as a “proof of principle” experiment.

But that result holds for only 8% of attempts

The 99.99% fidelity was measured on the roughly 8% of attempts that passed every check; the other 92% were thrown out. The 40x figure compares states made with and without cultivation; it covers the state-preparation step, not the processor’s overall performance. The paper says this level is similar to recent trapped-ion results.

This is not a code-breaking quantum computer

The experiment used at most 54 of Willow’s 105 qubits. Breaking RSA-2048 would need errors in these states to drop about a million times lower than today’s level. Craig Gidney, a co-author of the theory behind the method, wrote that its final stage needs a code of about 500 qubits, which doesn’t fit on the chip yet.

Sources

  1. Google (blog), “Bringing the magic: Google breaks open a bottleneck in quantum computing”, Google Quantum AI's announcement: magic state cultivation on Willow, ~40x fewer errors than un-cultivated states and 99.99% logical fidelity, (blog.google)
  2. Science, “Magic state cultivation on a superconducting quantum processor”, Peer-reviewed paper: 40x error reduction, fidelity of at least 0.9999(1), with only 8% of attempts retained, (science.org)
  3. arXiv, “Magic state cultivation on a superconducting quantum processor (arXiv:2512.13908)”, Open-access preprint of the paper (CC BY 4.0); the chart on slide 05 comes from here, (arxiv.org)
  4. arXiv, “Magic state cultivation: growing T states as cheap as CNOT gates”, The theory paper in which Gidney, Shutty and Jones first proposed the technique, (arxiv.org)
  5. arXiv, “Universal Quantum Computation with ideal Clifford gates and noisy ancillas”, Bravyi and Kitaev's paper introducing magic states and distillation, the starting point of Google's "20 years", (arxiv.org)
  6. Google Quantum AI (YouTube), “Demonstrating Magic State Cultivation on the Willow Processor | Google Quantum AI”, Official video in which Emma Rosenfeld explains the experiment (used on slides 02-04), (youtube.com)
  7. Scientific American, “Google’s quantum Willow chip may be able to grow ‘magic states’”, Duke's Kenneth Brown: the method works in practice, but 92% of attempts are discarded; 54 of 105 qubits used, (scientificamerican.com)
  8. The Quantum Insider, “Google Demonstrates Method to Cut Errors in Quantum Computing’s ‘Magic States’”, Notes that the 40x applies to state preparation, not the whole processor, (thequantuminsider.com)
  9. ScienceAlert, “Google’s Quantum Computer Just Achieved a Major Breakthrough With ‘Magic States’”, Explains distillation versus cultivation with a greenhouse analogy, (sciencealert.com)
  10. The Block, “Google demonstrates quantum computing breakthrough as observers weigh Q-day implications”, Crypto angle: no imminent threat, but earlier resource estimates now look more credible, (theblock.co)
  11. Decrypt (Yahoo Tech), “Google Says It Just Cracked One of Quantum Computing’s Biggest Bottlenecks”, Points out that breaking RSA-2048 would need magic states roughly a million times cleaner than today's, (tech.yahoo.com)
  12. arXiv, “Breaking even with magic: demonstration of a high-fidelity logical non-Clifford gate”, Quantinuum's trapped-ion magic state result, a benchmark for the "record" claim, (arxiv.org)
  13. arXiv, “How to factor 2048 bit RSA integers with less than a million noisy qubits”, Gidney's estimate that RSA-2048 could be factored with under a million qubits, partly thanks to cultivation, (arxiv.org)
  14. Google (blog), “Meet Willow, our state-of-the-art quantum chip”, Announcement of the 105-qubit Willow chip, (blog.google)
  15. Nature, “Quantum error correction below the surface code threshold”, The Willow paper showing errors fall as more qubits are added, (nature.com)
  16. Sundar Pichai (X), Google's CEO announced the result with a video, (x.com)
  17. Google Quantum AI (X), Official post: 40x fewer errors compared to without cultivation, 99.99% fidelity, (x.com)
  18. Google Quantum AI (X), Proof-of-principle experiment grafting the states onto standard encodings, (x.com)
  19. News from Google (X), Google's three-point summary and its "20-year bottleneck" framing, (x.com)
  20. Watcher.Guru (X), The most-viewed post about the news, past 1 million views, (x.com)
  21. Craig Gidney (X), Co-author of the theory: lots left to do, the ~500-qubit big code doesn't fit on the chip yet, (x.com)
  22. Craig Gidney (X), Poll sharing that the theory paper was desk-rejected by PRX, (x.com)

About this story

Spotted an error in this story? [email protected] · Instagram

Short link: thejarrus.com/en/google-magic-state

This story in Turkish: Google, “sihirli durum” darboğazına karşı yöntemini Willow çipinde ilk kez gösterdi

On the same topic