Google demonstrates practical magic state cultivation on Willow quantum processor

The research was published today in *Science*.

**The magic-state bottleneck**

Universal quantum computation — the kind that runs useful programs — requires a specific set of error-correctable quantum operations. The standard way to produce them is through "magic states," special quantum states that enable operations beyond classical computing. The problem: creating high-quality magic states consumes enormous resources. Estimates cited in the paper suggest up to 95 percent of a quantum computer's total capacity could be devoted solely to generating them, leaving little room for actual computation. Another constraint: a magic state, once used, cannot be reused, making it a scarce resource. Researchers have spent two decades trying to lower that cost.

**How cultivation works**

In 2024, Google Quantum AI researchers Craig Gidney, Noah Shutty, and Cody Jones developed an approach based on an idea from the literature. The concept acts as a filter: inject a noisy state into a small region of the quantum processor, run a sequence of coordinated measurements to check for errors and discard failed attempts, thereby "cultivating" an exceptionally pure, low-error quantum state that can then be converted into a larger, more useful encoding. On paper the approach looked promising, but running it on real quantum hardware required a series of technical and engineering innovations.

**The lab obstacle**

One example of the complexity emerged when Google Quantum AI research scientist Emma Rosenfeld saw data that appeared to violate the laws of physics. After months of confusion, simulations, discussions with colleagues, and consistency checks, the culprit turned out to be a faulty assumption in the team's standard measurement technique. A specific type of lab noise — caused by electrical equipment drift and temperature fluctuations — had invalidated the test. The fix demanded changes on both the theoretical and engineering fronts: the team added an extra control layer to suppress the noise's influence and invented a new measurement technique that uses cultivation itself as a self-measurement tool.

**The result and its significance**

On Google's Willow processor, the method achieved a 40-fold error reduction compared to uncultivated states and a record logical fidelity of 99.99 percent for T states. Google Quantum AI has demonstrated for the first time a practical solution to one of quantum computing's most severe bottlenecks, and a crucial step toward scaling quantum computing for practical use. The source provides no timeline for commercial deployment or additional performance details beyond those published.