Willow quantum chip: Google cuts errors in 'magic states' about 40-fold

Google Quantum AI says it has tackled an important bottleneck in building useful quantum computers. On the 105-qubit superconducting Willow processor, the researchers tested so-called magic state cultivation. A quantum computer needs such states to run arbitrary calculations, yet each one can only be used once. According to Google, errors in producing them fell by about 40 times compared with states made without cultivation. The team reports a logical fidelity of 99.99 percent for the so-called T-states, a record. The work was published in the journal Science.

Google researchers Craig Gidney, Noah Shutty and Cody Jones developed the method on paper in 2024. By some estimates, up to 95 percent of a future quantum computer's capacity could be needed just to produce magic states, which makes the approach economically interesting. In the experiment, the team also transferred the states from a smaller error-correcting code to a larger one, using 54 qubits. Machine learning was used to fine-tune the chip. According to Google, early measurements were misleading because of equipment drift and temperature changes, so the team developed a new measurement method.

Google itself stresses the limits: the 40-fold improvement applies only to producing these states, not to the processor as a whole. When the states were stored, the error rate was about seven times higher than simulations predicted, probably because of so-called leakage. Larger demonstrations are needed for practical use. According to The Block, experts from the crypto industry see no immediate threat to today's encryption but regard the result as confirming earlier resource estimates.