Google shows magic state cultivation on Willow with 99.99% logical fidelity

[TopBins64]

Google Quantum AI has shown that magic state cultivation works on real hardware, and it is one of the most important error correction results of the year. The blog post by Emma Rosenfeld & Craig Gidney accompanies a paper published in Science. The experiment ran on Google's Willow processor

Some background helps here. A universal quantum computer needs a special ingredient called magic states to perform certain operations, such as the T gate. These states are expensive to make, and each one can only be used once. Some estimates suggest that up to 95 percent of a future quantum computer's capacity could be spent producing them, which makes them one of the biggest bottlenecks on the road to useful machines

Magic state cultivation was proposed in 2024 by Craig Gidney, Noah Shutty and Cody Jones as a cheaper alternative to the traditional method, magic state distillation. It works like a filter. A noisy state is injected into a small region of the chip, coordinated measurements check for errors, flawed attempts are thrown away, and the surviving state is gradually grown into a larger, cleaner encoding

On Willow, cultivation reduced errors by about 40 times compared with uncultivated states and produced T states with a record logical fidelity of 99.99 percent. The team also showed it could move states from a distance 3 colour code into a distance 5 surface code, a step towards making the two error correction approaches work together. Along the way, they traced a puzzling measurement anomaly to drifting electronics and temperature changes in the lab, and developed a new benchmarking method that uses cultivation to measure itself

The authors are careful to say that cultivation still needs to be shown at larger scale and that significant challenges remain. But turning a theoretical protocol into a working hardware demonstration within two years is remarkable. If magic states become cheap, the size of a useful fault tolerant machine could shrink dramatically

How much could cultivation reduce the size of future quantum computers? And which group do you think will be the first to reproduce this on different hardware?

Bringing the magic: Google breaks open a bottleneck in quantum computingblog.google