Google says it has demonstrated a way to make a key quantum-computing resource substantially more reliable on real hardware, potentially easing one of the field’s largest bottlenecks. Its Quantum AI team reported that “magic state cultivation” reduced errors by about 40 times on the Willow processor and produced states with 99.99% fidelity.
A costly resource for useful quantum computing
The result matters because quantum computers need more than qubits that can represent combinations of zero and one. They also require specialised operations to run practical programs. Those operations use magic states, single-use resources that are consumed when applied.
Quantum devices are highly sensitive to heat, electrical noise and other disturbances. Error correction spreads information across multiple qubits so that some mistakes can be detected and corrected, but the most difficult operations cannot be protected in the same way. Magic states are therefore prepared separately and supplied to the computation.
Google says some estimates place magic-state production at as much as 95% of a quantum computer’s total capacity. That leaves a relatively small share of the machine available for the calculation itself. The established approach, known as magic state distillation, combines many noisy states to create fewer, cleaner ones. However, it requires large sections of the computer to operate as dedicated “factories.”
According to Decrypt, Google’s paper in Science describes a different approach that the company proposed in 2024. Cultivation uses a small area of the chip to test an imperfect state repeatedly. Attempts that show an error are discarded, while successful states are retained and developed through further checks.
The experiment did not retain most attempts. Only about 8% passed all the checks, meaning more than nine out of 10 tries were rejected. Google nevertheless said the protocol delivered a 40-fold reduction in errors compared with uncultivated states and achieved a record logical fidelity of 99.99% for T-states, a type of magic state used in quantum operations.
What the demonstration changes—and what it does not
The cultivation technique had previously been supported by theory and computer simulations. Demonstrating it on Willow gives Google a hardware result rather than only a projected one. The experiment used no more than 54 of Willow’s 105 qubits, according to the source material.
The work is relevant to both useful quantum applications and cybersecurity, although it does not mean current encryption or cryptocurrency wallets have been broken. Quantum computers are being developed for tasks such as drug and materials design, while sufficiently capable machines could eventually threaten encryption systems used to protect online accounts and digital assets.
Magic-state cultivation is among the techniques behind a 2025 estimate by Google researcher Craig Gidney that RSA-2048 encryption could be broken with fewer than one million noisy qubits. That estimate was described as a 20-fold reduction from his 2019 estimate of 20 million qubits.
Google also said in March that a whitepaper co-authored by Quantum AI, Ethereum Foundation researcher Justin Drake and Stanford cryptographer Dan Boneh estimated that breaking Bitcoin’s elliptic-curve cryptography could require fewer than 500,000 physical qubits. Bitcoin’s cryptography prevents someone from deriving a private key from a public address.
Those estimates remain different from the capabilities demonstrated in the latest experiment. The Science paper says factoring an RSA-2048 key would require magic states with errors of approximately one in 10 billion. That is roughly one million times cleaner than the states Google produced in this test. The result therefore represents progress in reducing a bottleneck, not a demonstration that RSA-2048 or Bitcoin wallets can currently be attacked.
Google’s next step and the market picture
Google said its next objective is to operate cultivation at a larger scale. The team projects that cutting Willow’s underlying hardware noise in half would reduce magic-state errors to about two in 100,000, because errors in the process decrease with the cube of the chip’s noise level.
The company’s broader quantum programme has also included the Quantum Echoes experiment, described in the source material as the first verified quantum speedup on real hardware. The latest result adds a hardware test of cultivation to that development path, while leaving the much stricter error levels needed for code-breaking still unmet.
Moneyiar market data at 21:53-21:54 on October 17, 2026, show mixed performance among the listed assets:
- JUST (JST): $0.1354785814, down 0.16% over 24 hours and 0.96% over seven days.
- Quant (QNT): $193.1830733781, down 4.55% over 24 hours and 19.53% over seven days.
- Compound (COMP): $21.2013374966, down 3.9% over 24 hours and 14.99% over seven days.
- AI Rig Complex (ARC): $0.0388760204, up 10.31% over 24 hours but down 36.54% over seven days.
The market figures are separate from Google’s quantum-computing announcement and do not establish a direct connection between the reported chip result and the performance of any listed asset.
