Alice & Bob's 22,593 Cat Qubit Claim: What the New RSA-2048 Estimate Actually Shows
Alice & Bob's 22,593 Cat Qubit Claim: What the New RSA-2048 Estimate Actually Shows
French quantum computing company Alice & Bob has drawn fresh attention with a claim that Shor's algorithm, run on 22,593 physical cat qubits, could factor RSA-2048 encryption in roughly 3.14 days. It's important to be precise about what this is: a theoretical resource estimate, not a demonstrated attack. No quantum computer has actually broken RSA-2048. The figure comes from a mix of an unreviewed preprint posted to IACR ePrint and company blog posts and newsroom materials.
Despite that caveat, the claim is drawing coverage because it represents a significant reduction from prior leading estimates. Many observers in the quantum computing field note that resource estimates like this one tend to shrink over time as researchers refine error-correction schemes and algorithmic assumptions, and this latest number continues that trend.
How This Compares to Prior Estimates
Alice & Bob presents its new figure in contrast to Google's 2025 estimate from the Gidney et al. research, which reportedly required roughly 40 times more physical qubits to achieve a similar result. The company also points to its own earlier internal estimate of 49,649 qubits, based on repetition codes, as a baseline it says it has now improved upon.
Alice & Bob attributes the drop to a newer error-correction approach it calls "LDPC-cat" codes. A companion estimate for breaking ECDLP-256, a different encryption scheme, puts the figure at 29,588 qubits and about 61.8 minutes. Taken together, these numbers are being promoted as evidence of fast-moving progress in reducing the hardware overhead theoretically required for cryptographically relevant quantum attacks.
What's Actually Been Verified — and What Hasn't
It's worth separating two distinct claims here. The first is that LDPC-cat error-correction codes are a promising approach to reducing qubit overhead in cat-qubit architectures. This idea has some grounding in academic work, including research published through an institutional research portal and related work appearing on arXiv and in Nature Communications on cat qubit error correction more broadly.
The second, separate claim is the specific RSA-2048 resource figure: 22,593 qubits and 3.14 days. This number currently rests solely on the unreviewed IACR preprint and the company's own communications. The broader error-correction research does not independently confirm this specific estimate. There has been no peer review of the RSA-2048 figure in a journal, and no hardware validation demonstrating the claimed performance. A recurring concern among industry observers is that general advances in error-correction research can be used to lend unearned credibility to specific, unverified headline numbers.
Reading the Endorsements and Framing Critically
Coverage of the estimate includes quoted endorsements from a Gödel Prize laureate and the CEO of PQShield, a commercial post-quantum cryptography company. These endorsements lend an appearance of authority, but it's worth noting that PQShield has a direct commercial interest in heightened urgency around migration to post-quantum cryptography, which is the market it serves.
The original coverage of this estimate has also been described as carrying characteristics of a press release, with promotional language such as claims of setting a "record low" and framing that suggests cryptographic risk is both real and rapidly approaching. Some reporting on the topic includes a notable gap between attention-grabbing headlines stating that the work "breaks RSA-2048" and a more measured caveat found later in the same coverage, acknowledging that the figures are theoretical resource projections rather than demonstrated attacks. Readers are encouraged to weigh headline framing against these buried qualifications.
What This Means for Post-Quantum Cryptography Migration
Resource estimates like this one, even when unverified, tend to fuel arguments for faster migration to post-quantum cryptographic standards. Many in the security community view this as a reasonable precaution regardless of whether any single estimate holds up, since transitioning cryptographic infrastructure takes years.
At the same time, it's worth being cautious about conflating a theoretical reduction in required qubit counts with an imminent, real-world cryptographic threat. The field currently includes multiple competing estimates from Google, Alice & Bob, and others, reflecting an active and unsettled area of research rather than a settled scientific consensus.
A measured takeaway is to treat the 22,593 qubit, 3.14-day figure as a company-reported projection worth watching rather than an established fact. Independent peer review, replication by other research groups, and eventually hardware demonstration would be needed before this specific claim could be considered verified.