Google says magic state cultivation now works on its Willow chip, cutting errors about 40-fold

Google Quantum AI says it has shown that "magic state cultivation" works in practice on its Willow processor, calling it the first demonstration of a practical solution to the bottleneck, and that the experiment cut errors about 40 times compared with states made without the technique. The company published the result in Science on October 15, 2026, and described it in a post by two of its researchers, Emma Rosenfeld and Craig Gidney.

Magic states are specially prepared quantum states that a fault-tolerant machine consumes to run the harder operations a useful program needs. Each one is used up when it is applied, so they are scarce. Google says some estimates put the cost of making clean ones at up to 95 percent of a quantum computer's capacity.

The idea is not new. In 2024, Google researchers built on an earlier idea and developed cultivation, which works like a filter: a noisy state is injected into a small patch of the chip, a series of coordinated measurements checks for errors, and flawed attempts are discarded. What is new is the hardware test. Google says its protocol produced T-states, a type of magic state, with a record logical fidelity of 99.99 percent. The team also moved states from a distance-3 color code, which suits cultivation, into a distance-5 surface code, a more established error-correction scheme. Google presents this as evidence that cultivated states can survive the transfer inside its hardware.

The post also describes a setback. Early data seemed to defy physics until the team traced it to equipment drift and temperature variation that undermined its measurement method. Google says it added extra control and devised a new benchmark that uses cultivation to measure itself.

This account rests on a single source, Google's own blog post. We found no independent reporting that tested the claims, and other outlets appear to repeat Google's account. We did not review the Science paper itself, so the 40-fold reduction and 99.99 percent fidelity are the company's characterization of its results.

Google also says what remains undone. It says cultivation must still be demonstrated at a larger scale to be realistic for practical quantum computing, and that plenty of technical challenges remain.

The significance is about cost. If cultivation holds up at scale, less of a machine would need to be spent preparing magic states, which could shrink the hardware needed for useful error-corrected computation. Today, though, this is a single experiment on one chip and one protocol. It narrows a known bottleneck but does not by itself show that a practical fault-tolerant quantum computer exists.

Revised 16 October 2026: Guth re-read the original sources and rewrote this article. The earlier version remains available as revision 1.