The headline is dramatic; the reality is conditional
The current story is not that a quantum computer has already cracked RSA-2048. It is that researchers associated with Alice & Bob have produced a new estimate for what a future fault-tolerant machine using superconducting “cat qubits” might need to factor a 2048-bit RSA integer: 22,593 physical qubits and about 3.14 days . Around Prague’s November 5 report stresses the central distinction: the result is a calculation on paper, not a demonstration on existing hardware .
That distinction matters because RSA-2048 remains one of the best-known public-key systems used to secure web connections, banking, software updates, email and digital signatures . The new estimate therefore lands in a sensitive area: if the resource requirements for attacking RSA continue to fall, the timeline for migrating to post-quantum cryptography becomes less theoretical and more operational.
The claim is based on Shor’s algorithm, the 1994 quantum algorithm that can speed up integer factorization exponentially compared with known classical methods . RSA’s security depends on the fact that it is easy to multiply two very large prime numbers together, but extremely difficult for conventional computers to reverse the operation and recover those primes . A sufficiently capable quantum computer running Shor’s algorithm would change that balance.
What Alice & Bob is actually estimating
The number now circulating is precise: 22,593 physical cat qubits to factor RSA-2048 in 3.14 days . AZoQuantum’s November 3 coverage, based on Alice & Bob’s announcement, describes the same headline result and adds that the architecture is presented as needing roughly 40 times fewer physical qubits than Craig Gidney’s 2025 Google Quantum AI estimate for the same RSA-2048 target .
The work also includes other resource estimates. The same reporting says a 256-bit elliptic-curve discrete logarithm problem, relevant to ECDSA and systems such as Bitcoin and Ethereum, is estimated at 29,588 physical qubits and 61.8 minutes . For Bitcoin specifically, current reports note that an hour remains longer than the network’s roughly 10-minute block interval, so the estimate does not imply that in-flight Bitcoin transactions are immediately vulnerable .
A faster RSA-2048 configuration is also discussed: connecting enough capacity to reach about 610,530 cat qubits could reduce the estimated RSA-2048 runtime to around 1.25 hours . Conversely, lower-qubit configurations can take longer. The central lesson is not a single magic number, but a trade-off between machine size and computation time.
Why cat qubits change the arithmetic
Cat qubits are not a brand name for ordinary qubits. They are a hardware approach named after Schrödinger’s cat, using quantum states of photons in a resonator to encode information . Their appeal is that they can strongly suppress one major class of error, bit flips, while leaving phase flips as the main error to correct .
That error bias is the reason the estimate is so much smaller than many earlier resource projections. If the hardware itself suppresses bit-flip errors, the error-correction layer can focus more narrowly on phase-flip errors, reducing overhead . Alice & Bob also credits LDPC-cat codes and a redesigned adder circuit as key contributors to the lower qubit count .
This is why the estimate should not be compared to every other quantum-computing headline by qubit count alone. A noisy qubit, a logical qubit, a cat qubit and a qubit in a neutral-atom array are not interchangeable units of computing power. What matters is the complete architecture: error rates, cycle time, decoding latency, gate design, layout, and how long the qubits must preserve useful information.
The hidden assumption: almost no bit flips
The strongest caution in the current reporting concerns bit-flip stability. Quentir’s November 3 analysis says the 22,593-qubit figure treats bit flips as negligible and equates that to a characteristic bit-flip time of about 100 years . That is a severe assumption, not a measured property of a current multi-thousand-qubit processor.
Quentir also reports that if the system is allowed to have a bit-flip time closer to one hour, an extra protection layer would be needed and the estimate rises to about 45,000 cat qubits . That figure is still strikingly low compared with older million-qubit-scale discussions, but it is a very different engineering target from the headline number.
ScienceAlert’s report, republished by Yahoo Tech on November 5, makes the same practical point: no one currently has a 23,000-qubit quantum computer capable of performing this computation . It also notes that multiple hurdles remain, including temperature control, scalability, stability and linking qubits or chips over useful distances .
A blueprint, not a breach
The most responsible way to read the result is as a blueprint-level estimate. AZoQuantum reports that, unlike rough estimates that multiply logical qubits by an average overhead, this work attempts to map major parts of the computation down to what physical qubits do and when, producing chip layouts and schedules . That makes the work more concrete than a back-of-the-envelope calculation.
But concrete does not mean built. Around Prague’s current report says the study has not yet been peer reviewed and that the authors acknowledge major engineering hurdles remain . ScienceAlert likewise frames the work as a preprint resource estimate, not a finished machine or a live cryptographic attack .
This matters for public interpretation. A reader could easily see “break RSA-2048 in 3.14 days” and infer that encrypted bank accounts, software updates and private messages are suddenly exposed. The current evidence does not support that conclusion. What it supports is more subtle: if the assumptions can be engineered at scale, the size of a cryptographically relevant quantum computer may be smaller than many security planners expected.
What security teams should take from it
For cybersecurity leaders, the important part is not panic; it is lead time. AZoQuantum’s November 3 article quotes the post-quantum migration message clearly: replacing cryptographic infrastructure across large organizations can take years, and US and UK guidance calls for completing the transition to post-quantum cryptography by 2035 .
That deadline can sound comfortably distant, but large systems do not replace cryptography overnight. Certificates, embedded devices, payment systems, identity platforms, VPNs, archival data, industrial equipment and government systems all have different refresh cycles. The “harvest now, decrypt later” risk also means encrypted data captured today could become readable in the future if it remains valuable long enough.
The new estimate therefore strengthens the argument for crypto-agility: organizations should know where RSA and elliptic-curve cryptography are used, be able to swap algorithms without rebuilding entire systems, and prioritize long-lived secrets first. The story does not prove that RSA-2048 will fall in the early 2030s, but it does make complacency harder to defend.
The broader cryptographic signal
The history of quantum resource estimates is one of compression. Current coverage notes that estimates for Shor’s algorithm have dropped from tens of millions of physical qubits in older studies to below one million in 2025, and now to fewer than 23,000 under Alice & Bob’s cat-qubit assumptions . Such comparisons must be handled carefully because the hardware assumptions differ, but the direction of travel is clear: better algorithms, codes and architectures keep reducing the projected cost.
That does not mean every current encryption system is equally exposed. Symmetric encryption and hash functions are affected differently by quantum algorithms than RSA and elliptic-curve systems. The urgent target is public-key cryptography used for key exchange and digital signatures.
Nor does it mean that cat qubits have won the quantum-computing race. It means Alice & Bob has presented a compelling architecture-specific case for a lower-resource Shor machine, and current reporting is treating it as an important benchmark rather than a completed capability . Other platforms may also improve, and their practical advantages could appear in manufacturing, control electronics, connectivity or error correction.
Bottom line
As of the latest reports within the past 72 hours, the story is a sharp warning, not an emergency breach. The 22,593-cat-qubit, 3.14-day figure is a resource estimate for a hypothetical fault-tolerant quantum computer, not evidence that RSA-2048 has been broken in practice . The estimate depends on demanding assumptions, especially extremely rare bit flips, and relaxing those assumptions raises the hardware requirement to about 45,000 cat qubits .
Still, the result is important because it narrows the imagined gap between today’s cryptography and tomorrow’s quantum hardware. For governments, banks, cloud providers, software vendors and infrastructure operators, the prudent response is not fear. It is inventory, migration planning and post-quantum testing now, while there is still time to replace the locks before the key exists.