On October 15, 2026, Google announced a magic-state cultivation experiment on its Willow superconducting quantum processor. The Science paper reports a 40-fold reduction in magic-state error and a fidelity bound of at least 0.9999(1), with 8% of attempts retained for that result. (Google Quantum AI; the Science paper)
The figures describe cultivated magic states, a component used to enable operations needed for universal quantum computing. They are not a general accuracy rating for Willow. The experiment demonstrates the method on hardware; Google says larger-scale cultivation remains to be demonstrated.
Why quantum computers need magic states
Quantum computers use different kinds of operations to process information. Magic states enable non-Clifford operations, which are needed for universal quantum computation. Producing reliable magic states is therefore one piece of building a quantum computer capable of a broader range of calculations.
How cultivation prepares a state
Cultivation prepares a candidate magic state and checks it through a sequence of measurements. Attempts that fail the checks are rejected; the result is a selected state that can be used for the next step. This differs from distillation, which combines a batch of imperfect states to produce higher-quality ones. The Science paper describes the checks and postselection behind the experiment’s result.
What the numbers measure
The 40-fold reduction compares cultivated magic-state error with error in uncultivated states. For the reported fidelity result, the team retained 8% of attempts. The paper gives the fidelity as a bound, at least 0.9999(1), established with a fault-tolerant measurement protocol. That number applies to the cultivated states assessed by the protocol—not to every operation or qubit on Willow.
| Result or step | Reported value | Scope |
| Magic-state error reduction | 40-fold | Cultivated versus uncultivated states |
| Fidelity bound | At least 0.9999(1) | Cultivated states; 8% of attempts retained for this result |
| Code switch | Distance-3 color code to distance-5 surface code | Transition reported by Google |
| Fidelity measurement | Fault-tolerant measurement protocol | Used to bound the cultivated-state fidelity |
The code switch and the path to scale
Google reports that the experiment moved from a distance-3 color code to a distance-5 surface code, demonstrating code switching as part of the cultivation work. The transfer itself introduced additional errors, ScienceAlert reporter Michelle Starr reported; completing the transfer did not mean the state was error-free. (Google Quantum AI; Michelle Starr’s report for ScienceAlert)
Google says larger-scale cultivation still needs to be demonstrated before the method is practical for quantum computing. The Willow result is a hardware demonstration of the approach, while that larger-scale work remains ahead.
