Google cuts errors 40-fold in a key quantum computing step

Google’s Quantum AI team has cut errors roughly 40-fold in preparing a key ingredient for quantum calculations, using a technique tested on its Willow processor. The results, published in Science, demonstrate a more efficient way to tackle one of the field’s major bottlenecks.

Quantum computers are prone to mistakes. To carry out certain essential operations reliably, they need specially prepared quantum states, known as “magic states”. Producing these has required large numbers of qubits, the basic building blocks of a quantum computer, as well as substantial time.

The new method, called magic state cultivation, creates these states within a smaller group of qubits. Coordinated measurements detect errors so faulty attempts can be discarded. The experiment puts into practice an approach Google researchers proposed in 2024.

Earlier theoretical work suggested cultivation could allow important quantum programs to run with one-tenth of the qubits previously expected. That potential saving has not yet been demonstrated in useful, large-scale computations.

There are also hurdles to clear. Only 8% of the experiment’s runs produced high-quality states suitable for further use, and error rates still need to fall before Willow can tackle truly useful problems. Even so, the demonstration offers a concrete route toward reducing the hardware needed for reliable quantum computers.