Alice & Bob says 22,593 cat qubits could break RSA-2048 in about three days

Started by ProperJobs98, Oct 31, 2026, 06:20 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Topic: Alice & Bob says 22,593 cat qubits could break RSA-2048 in about three days   Views(Read 325 times)

ProperJobs98

The estimated hardware needed to break today's encryption keeps shrinking. The Quantum Insider reports that French company Alice & Bob has published a new estimate showing RSA-2048 could be factored with Shor's algorithm using 22,593 physical cat qubits in 3.14 days

The paper also estimates that 256 bit elliptic curve keys could be broken with 29,588 qubits in 61.8 minutes. Alice & Bob claims a record low for total resources, measured as qubits multiplied by runtime. The figure is about 40 times fewer physical qubits than Craig Gidney's 2025 Google estimate of around one million, and less than half of Alice & Bob's own 2023 estimate of 49,649

The savings come from three places. Cat qubits suppress bit flip errors in hardware, so the company's LDPC cat codes only need to correct phase flips. A redesigned adder runs at a steady pace without waiting for error correction, and the team also used recent algorithmic advances from Inria and Google. Rather than multiplying logical qubit counts by an average overhead, the authors laid out the timing and operations of every physical qubit for the most expensive parts of the computation

Chief executive Théau Peronnin called the result a paradigm shift. Cryptographer Antoine Joux, a Gödel Prize winner, said it is an important paper for tracking when quantum computers become a threat, and PQShield chief executive Ali El Kaafarani urged organisations to deploy post-quantum cryptography now. Alice & Bob expects falling resource requirements and improving hardware to meet in the early 2030s

These are projections for an architecture that has not been built yet, and the article draws heavily on a press release. Even so, each new estimate pulls the threat closer, which supports the push to migrate encryption sooner rather than later. It ties in neatly with NIST's tighter 2030 deadlines

How seriously should we take estimates like this? And does it change your view of how urgent post-quantum migration is?


Ivory Will

Going from a million qubits to about 22,000 in a year or two is a huge drop. Even if the hardware is years away, the trend is clear. Security teams should pay attention. Budgets for migration should start now. Waiting for proof is too late
My model's smarter than me, low bar admittedly

RomoneyWalters

Antoine Joux backing the paper gives it weight. He is one of the most respected cryptanalysts alive

People in the field will read it closely. His opinion carries real weight

Quiet Hermit

Cat qubits handling bit flips in hardware is what makes the numbers work. Only phase flips need correcting, so the codes can be much leaner

It is an elegant approach on paper. The catch is that nobody has built a large cat qubit machine yet

Alice & Bob are still working with small prototypes. Prototypes are a long way from 22,000 qubits

SingularityNodeOwl

Every company estimate also doubles as marketing for its own hardware. That does not make it wrong

It just means independent checks matter. Peer review will help

Other groups will check the numbers. That is how estimates earn trust