Key Notes
- Google demonstrated magic state cultivation on Willow, reducing errors by roughly 40x compared with uncultivated states.
- The experiment achieved 99.99% logical fidelity while retaining only 8% of preparation attempts.
- Larger implementations are still needed before the approach can support useful fault-tolerant quantum computing at scale.
Google has demonstrated a way to produce a crucial quantum computing resource with roughly 40x fewer errors, reporting a step toward machines that can perform reliable, general-purpose calculations. The experiment used its Willow processor to prepare so-called magic states with 99.99% logical fidelity, addressing a resource bottleneck on the road to fault-tolerant computing.
The company announced the results on October 15, alongside publication in Science. The achievement concerns the quality of specially prepared quantum states; it does not mean Willow has become 40x faster or that Google has launched a commercially useful universal quantum computer.
Why Quantum Computers Need “Magic”
Magic states help supply operations needed for universal quantum computation. They are consumed during use, so a computer running a substantial algorithm needs a continuing supply. Producing those states reliably can take up a large share of the machine’s resources, making their preparation an important engineering problem in its own right.
Google’s approach, called magic state cultivation, begins with a noisy state and uses coordinated measurements to detect errors. Attempts that fail the checks are discarded. The experiment also demonstrated transferring the cultivated state between two error-correcting codes, connecting preparation with the surface-code architecture used to protect quantum information.
The 99.99% Result Comes With a Yield Trade-Off
The research paper reports a state fidelity of 0.9999(1), with 8% of attempts retained. That acceptance rate matters: the headline fidelity describes the states that survived the procedure, rather than a process that produced an equally good state on every attempt. The researchers developed a fault-tolerant measurement protocol to bound the resulting fidelity.
The distinction changes how the result should be read. Better quality and higher output are separate goals. A future machine would need to deliver enough usable states, at sufficiently low cost, to support a lengthy computation. Rejecting faulty attempts can improve the surviving output, but those rejected attempts still belong in any assessment of practical efficiency.
The study was first posted as a preprint on December 15, 2025. Today’s announcement accompanies its journal publication, rather than marking the first public appearance of every result. Larger implementations remain necessary to establish how well the approach works as the hardware and computations grow.
A Different Measure of Progress in the Quantum Race
Quantum announcements are often reduced to processor sizes or dramatic speed comparisons. This result highlights a different question: how much of a machine’s effort must go toward making its own calculations dependable? Improving an essential resource can matter even without a new chip or an immediate customer-facing application.
That engineering focus also runs through IBM’s quantum roadmap, which targets fault-tolerant computing, and Nvidia’s quantum tools, which apply AI to calibration and decoding. These efforts address different parts of the system, so their milestones should not be treated as interchangeable benchmarks.
For Google, cultivation is a promising contribution to that broader effort. The next meaningful test is whether such improvements can be combined into systems that sustain useful computations reliably and efficiently. The laboratory result strengthens the case for this approach while leaving that larger demonstration ahead.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
