AdSharper Image
Multi-Functional Compact Emergency Radio
$99.99
Compact multi-function emergency radio. US only.
Shop nowGoogle's 'magic state cultivation' cut errors about 40-fold
The full reading behind my Short: why magic states are a bottleneck, and what Google showed on real hardware.
The short version
Google Quantum AI says it has shown that magic state cultivation works on its Willow processor, reducing errors about 40-fold compared with un-cultivated states, and producing T-states with a record logical fidelity of 99.99% [1]. The work was published in Science on 15 October 2026 [1].
~40xfewer errors vs un-cultivated states [1]
99.99%record logical fidelity (T-states) [1]
up to 95%of capacity spent on magic states, in some estimates [1]
2004when this bottleneck was first foreseen [1]
How cultivation works
Magic states let a quantum computer perform the operations needed for universal computing, but making high-quality ones is resource-heavy, and each can only be used once [1]. Cultivation acts like a filter: a noisy state is injected into a small region of the chip, then a series of coordinated measurements checks for errors and discards flawed attempts, growing a very pure state [1].
The team also grafted states from a distance-3 color code to a distance-5 surface code [1].
Watch the Short
About this page
The Short uses real, illustrative photos from Wikimedia Commons (credited below); they are not photos of the study itself. This page is general science news.
Why it matters, and what's next
- Smaller machines. Cheaper magic states could reduce the size of quantum computer needed for practical tasks [1].
- Still early. Google says cultivation must still be shown at larger scale, and many technical challenges remain [1].
Full narration (transcript)
Google just cut a key kind of quantum computing error by about forty times. The trick is called magic state cultivation, and it was tested on Google's Willow chip. Magic states are special ingredients a quantum computer needs to run useful programs. Researchers have seen this bottleneck coming since 2004. But making good ones is costly: some estimates say up to ninety five percent of a machine could be spent on them. And each one can only be used once. Cultivation works like a filter. A noisy state goes into a small patch of the chip, and a series of checks throws out the flawed attempts. Those checks let the team grow an exceptionally pure, low error state. Getting it to work took months of detective work, including tracking down hidden noise in the lab. On Willow, errors dropped about forty fold, with a record logical fidelity of ninety nine point nine nine percent. The team also moved these states between two kinds of error correcting code. It's published in Science, but Google says it still has to work at a larger scale.
Sources
- [1] Rosenfeld, E. & Gidney, C. (15 Oct 2026). "Bringing the magic: Google breaks open a bottleneck in quantum computing." Google blog. blog.google. Research published in Science.
Credits
Photos (illustrative, Wikimedia Commons): IBM Quantum System One (OJB Quantum, CC BY 4.0); Google Sycamore Chip 001 (Google (https://www.youtube.com/channel/, CC BY 3.0); Dilution Fridge with Quantum Processors Enclosed 008 8K (OJB Quantum, CC BY 4.0); IBM Q at CES (39660636671) (Lars Ploughman, CC BY-SA 2.0); Semiconductor clean room (Aileen Devlin/Jefferson Lab from Newport, Public domain); Superconducting Quantum Chip with Dispersive Background 2 (OJB Quantum, CC BY 4.0); Superconducting Quantum Chip with Dispersive Background (OJB Quantum, CC BY 4.0); Superconducting Metamaterial Chip 002 (OJB Quantum, CC BY 4.0); Dilution Fridge 009 8K (OJB Quantum, CC BY 4.0); Keysight InfiniiVision DSOX 4024A and digital multimeter and (Radarvector, CC BY-SA 4.0); IBM Quantum Computer Demo (Dev Jadiya, CC BY-SA 4.0); Superconducting Metamaterial Chip 004 (OJB Quantum, CC BY 4.0); Vitoria-University-Library-food-science-journals-4489 (Vmenkov, CC BY-SA 3.0). Not endorsed by Google.