- Published on
Google Says It Just Cracked One of Quantum Computing's Biggest Bottlenecks
- Authors
- Name
- DeepHorus
Google Quantum AI said Thursday that magic state cultivation on its Willow chip cut errors about 40-fold. A Science paper reported the result. It produced magic states with 99.99% fidelity. This matters because magic states fuel useful quantum programs.
Quantum computers use qubits that can hold blends of 0 and 1, but they are fragile. Heat or electrical noise can flip a qubit and ruin calculations. Error correction spreads information across many qubits to catch mistakes. Simple operations are handled, but harder ones need magic states.
Magic states are single-use resources consumed when used. Producing clean ones has been expensive. Google says some estimates put the cost at up to 95% of a machine's capacity. That leaves little room for actual computing.
In 2004, Sergei Bravyi and Alexei Kitaev proposed magic state distillation, boiling many noisy states into few clean ones. It works but requires sprawling qubit factories. In 2024, Google researchers proposed cultivation as a cheaper filter-like route.
In the Willow experiment, cultivation injected rough states into a tiny patch, ran checks, discarded flawed attempts, and grew survivors. Only about 8% passed every check, so more than nine in 10 tries were discarded. The experiment used at most 54 of Willow's 105 qubits.
Cultivation underpins a 2025 estimate that RSA-2048 encryption could fall to under a million noisy qubits. That is a 20-fold drop from 20 million estimated in 2019. Breaking codes, however, would need magic-state errors near one in 10 billion. Previously, cultivation was tested only in theory and simulations.
In March, a Google Quantum AI whitepaper with Justin Drake and Dan Boneh put Bitcoin elliptic curve cryptography at fewer than 500,000 physical qubits. It follows October's Quantum Echoes experiment, the first verified quantum speedup on real hardware. The result may bring practical quantum computing closer, though not yet threatening encryption.
