Google Quantum AI said on Thursday that a technique called magic state cultivation reduced errors about 40-fold on its Willow processor, delivering magic states with 99.99% fidelity. Magic states are single-use resources needed to run useful quantum programs, and Google says some estimates put the cost of producing them at up to 95% of a quantum computer's capacity.
Google says it has turned one of quantum computing's most promising theoretical tricks into something that works on an actual chip.
In an announcement on Thursday, the company's Quantum AI team reported that magic state cultivation cut errors by about 40 times on its Willow processor, producing magic states with 99.99% fidelity.
Magic states are single-use resources consumed by error-corrected quantum programs, and their production is considered one of the field's biggest overheads. According to Google, some estimates put the cost of making them at up to 95% of a quantum computer's total capacity.
The technique is one of the methods behind Google's 2025 estimate that RSA-2048 encryption could fall to under a million noisy qubits. However, the company notes that code-breaking at that scale would require magic states with errors near one in 10 billion, a far higher bar than the fidelity reported in this demonstration.
The result is a step toward fault-tolerant quantum computing, where errors are corrected as a computation runs. Because magic states are consumed during execution and must be continually replaced, reducing the cost of producing them could materially change how soon practical quantum programs become feasible.
Analysis
Why This Matters
- Magic states are single-use resources required to run useful quantum programs, and some estimates put the cost of producing them at up to 95% of a quantum computer's capacity, so cutting that overhead directly affects how soon practical quantum computing arrives.
- The result builds on Google's 2025 estimate that RSA-2048 encryption could fall to under a million noisy qubits, with implications for the long-term security of widely used encryption.
- Quantum error correction remains one of the central engineering hurdles between today's noisy machines and machines able to run longer, useful computations.
Background
Quantum computers currently in operation are "noisy": their qubits are prone to errors, and useful programs require substantial extra resources to detect and correct those errors. Magic states are a special class of resource consumed by error-corrected quantum programs, and manufacturing them reliably has long been seen as a major cost. The new announcement claims to show, for the first time on Google's Willow chip, that a technique previously demonstrated mainly in theory can meaningfully reduce that burden in hardware.
Key Perspectives
Google Quantum AI: The company presents cultivation as a theoretical trick now working on a chip, citing a 40-fold error reduction and 99.99% fidelity on Willow, and links the technique to its roadmap toward large-scale code-breaking.
Quantum computing researchers: The practical significance will depend on whether the technique scales beyond a single demonstration and on whether error rates can be pushed toward the roughly one-in-10-billion level that Google says code-breaking would require.
Security and cryptography specialists: A credible path toward breaking RSA-2048 would increase pressure on organisations to accelerate migration to post-quantum encryption standards, a transition already expected to take years.
What to Watch
- Peer review and independent replication of the 40-fold error reduction on Willow or successor chips.
- Future Google announcements on scaling magic state cultivation to larger arrays of logical qubits.
- Whether demonstrated error rates move toward the one-in-10-billion threshold associated with attacking RSA-2048.