In research published in Science, the Google Quantum AI team demonstrated magic state cultivation on its Willow quantum processor, reducing errors by about 40 times compared with conventional methods of creating such states. The advance addresses a major barrier to scaling quantum computers, which require substantial resources to perform complex computations with error correction.

Reducing the cost of magic states

“Magic” states are special quantum states required for universal quantum computation, including operations beyond the capabilities of the standard set of quantum gates. Producing high-quality, error-corrected magic states requires significant computational resources.

By some estimates, producing them can consume as much as 95% of a quantum computer’s total capacity, leaving only a small share for running computations. A magic state also cannot be reused after it has been consumed.

The Google Quantum AI team has worked on the problem since 2024. Researchers Craig Gidney, Noah Shattuck and Cody Jones developed magic state cultivation, a method designed to produce higher-quality quantum states through a series of coordinated measurements.

The process acts as a filter. Researchers place a noisy initial state in a small region of the quantum chip, check it for errors and discard failed attempts. Successful attempts produce a purified state with a lower error rate that can be converted into a larger quantum code.

Practical implementation required additional engineering work. Google Quantum AI researcher Emma Rosenfeld found that an incorrect assumption in the measurement methodology had distorted early experimental results. Low-frequency noise associated with temperature fluctuations and the operation of electrical equipment affected the measurements.

The team changed how it controlled the system and developed a new validation method that used the magic state cultivation process itself to measure its accuracy.

Results on the Willow processor

On Willow, the researchers achieved about 40 times fewer errors than in states created without cultivation. They also produced T-type states with a record logical accuracy of 99.99%.

The team demonstrated that the technique was compatible with the processor’s architecture by transferring purified states between two quantum error-correcting codes: the color code and the surface code. The result marked a step toward integrating cultivation into more complex quantum systems.

Google said the technology must still be scaled before it can be used for practical quantum computing. Remaining technical challenges include preventing qubit states from leaking outside the computational space and improving operation fidelity.

Cryptography and the projected quantum threat

The findings also relate to debate over the potential effect of quantum computing on cryptocurrencies. In late March 2026, Google Quantum AI warned that future quantum computers could break cryptographic systems faster than previously expected, increasing attention on the transition to post-quantum cryptography.

In May, Galaxy Digital head of research Alex Thorn said the risk of quantum attacks on early bitcoins associated with Satoshi Nakamoto remained limited and long term. According to Thorn, the assets are distributed across about 22,000 addresses holding 50 BTC each, making a potential attack more difficult.

Project Eleven’s report The Quantum Threat to Blockchains — 2026 identified 2033 as the baseline date for “Q-Day,” when a quantum computer would be able to break modern cryptography. Its optimistic scenario placed that date in 2042, while its most aggressive scenario pointed to 2030.

Ethereum developer Justin Drake separately estimated a 50% probability that Q-Day would occur by 2032. Microsoft accelerated its Quantum Safe Program and plans to complete the core transition to post-quantum cryptography by 2029. StarkWare researcher Avihu Levy later reported the successful execution of the first quantum-resistant transaction on the Bitcoin network without protocol-level changes.

Source: Incrypted