Google’s Willow test cuts quantum errors 40-fold, but can it scale?

Google Quantum AI has demonstrated magic state cultivation on its Willow superconducting quantum processor, reporting a 40-fold reduction in errors and a T-state logical fidelity of 99.99%. The work was published in *Science*, but the available information does not establish its publication date.

The technique is called cultivation, not distillation. Google’s demonstration is a hardware result: it shows the method running on Willow, but does not show that the company has built a practical, large-scale fault-tolerant quantum computer.

Preparing magic states is described as a substantial resource cost in building fault-tolerant systems. A method that needs fewer resources and produces states with fewer errors could ease that engineering bottleneck. Whether cultivation delivers those benefits in a scaled-up system is not yet established.

The reported fidelity is 99.99%, with the underlying figure given as 99.99(1)%. The sources describe it as a record for T-state logical fidelity. They do not provide a previous logical-qubit performance record for comparison, so the size of the advance relative to Google’s earlier performance cannot be determined from the available information.

Google also reached a Willow below-threshold error-correction milestone in 2024. That is a separate milestone; the available evidence does not establish a direct performance comparison between it and the cultivation result.

The key test is whether the reported error reduction and fidelity can be maintained as cultivation scales, and how its resource demands compare with other methods in a larger system. For now, the demonstration addresses one costly part of fault-tolerant quantum computing, not the wider challenge of making such a computer practical.