Home / Cryptography & Math / Breakthrough Research Details Shor’s Algorithm Compilation on Cat Qubits, Breaking RSA-2048 in Just 3.14 Days Using 22,593 Qubits

Breakthrough Research Details Shor’s Algorithm Compilation on Cat Qubits, Breaking RSA-2048 in Just 3.14 Days Using 22,593 Qubits

A team of researchers from quantum computing firm Alice & Bob, alongside academic institutions including Univ Paris-Saclay, CEA, CNRS, Univ Rennes, Inria, and the University of Edinburgh, has released a comprehensive new paper detailing an end-to-end resource estimation for running Shor’s algorithm on superconducting cat qubits. Designated as Paper 2026/2665 on the Cryptology ePrint Archive, the study outlines a complete compilation pathway—ranging from high-level mathematical descriptions down to the precise physical allocation and scheduling of qubits on a hardware chip layout.

The research focuses heavily on modern iterations of Shor’s algorithm, extending its application not only to RSA encryption schemes but also to the Elliptic Curve Discrete Logarithm Problem (ECDLP) and the Finite Field Discrete Logarithm Problem (FFDLP). By meticulously mapping out logical circuits and evaluating the inherent trade-offs, the authors have compiled dominant cryptographic subroutines directly into physical-qubit operations. This achievement bridges the long-standing gap between abstract quantum algorithms and the harsh physical constraints of near-future hardware architectures.

Architecture, Primitives, and Circuit Compilation

To achieve a realistic physical resource estimation, the research team evaluated two primary hardware architectures: one based on repetition cat codes and another leveraging Low-Density Parity-Check (LDPC) cat codes. Cat qubits, which are superconducting oscillators that encode quantum information in coherent states of light, offer distinct advantages in error suppression, particularly regarding phase-flip noise. However, successfully compiling complex algorithms like Shor’s requires entirely new operational primitives tailored to these unique hardware characteristics.

To this end, the researchers designed a parity-Cnot gate specifically for cat qubits, extended a multi-target Cnot gate to the LDPC-cat architecture, and successfully adapted an active injection procedure. Building directly upon these newly developed primitives, the team modified the advanced adder circuit from Gidney’s prominent 2025 resource estimate for factoring RSA-2048 to fit cat-qubit hardware constraints.

Furthermore, the compilation strategy integrates this customized adder with a coarser resource estimate for the phaseup operation, alongside the factory designs and lookup compilation methodologies derived from the FeMoco resource estimation previously conducted by Larissa Kroell and her co-authors. Combined, these rigorous compilation steps generate concrete chip layouts and execution schedules capable of tackling heavy cryptographic workloads.

Factoring RSA-2048 and Benchmarking Additional Cryptographic Problems

The paper’s primary benchmark centers on the factorization of RSA-2048, a widely used cryptographic standard that underpins secure internet communications. Under specific baseline assumptions—including a 1-microsecond error-correction cycle time, a 10-microsecond reaction time, and circuit-level phase-flip noise with a probability of $p_Z = 5 times 10^-4$ in the absence of bit-flips—the configuration successfully compiles to approximately 22,593 cat qubits with a total execution time of 3.14 days.

Beyond standard RSA encryption, the researchers demonstrated the versatility of their compiled operations by applying them to alternative cryptographic problems. Solving the Elliptic Curve Discrete Logarithm Problem on secp256k1 requires at most 30,000 cat qubits and completes in roughly one hour by reusing the operations compiled for RSA-2048. Additionally, the team introduced FFDLP-512 as a viable intermediate benchmark, showing that it can be solved using about 6,700 cat qubits in just 21 minutes.

The research also explores significant operational trade-offs, particularly between spatial allocation and execution speed. By trading space for time, the resource requirements shift dramatically; RSA-2048 can be factored in a mere 1.25 hours, though this configuration demands approximately 611,000 cat qubits.

Addressing Hardware Constraints and Error Correction Layers

A critical consideration in physical quantum computing is managing noise and hardware limitations. The baseline RSA-2048 compilation relies on an assumption of zero bit-flips during the operational window, which mathematically corresponds to an exceptionally long bit-flip time of roughly 100 years.

To make the hardware requirements more experimentally feasible, the authors evaluated how relaxing this constraint impacts overall resource costs. By introducing an additional concatenation layer into a distance-3 bit-flip error-correcting code, the required bit-flip time can be safely reduced to approximately one hour. Under these relaxed conditions, the estimated physical cost for factoring RSA-2048 increases moderately to about 45,000 cat qubits, demonstrating the resilience and adaptability of the proposed cat-qubit compilation framework.

The research team behind Paper 2026/2665 includes Axel Pappalardo, Dahmun Goudarzi, Victor Hertz, Hugo Jacinto, Larissa Kroell, André Schrottenloher, Peter Shanahan, Élie Gouzien, Rémi de La Vieuville, and Vivien Londe. Their findings provide a vital roadmap for the quantum computing community, offering precise blueprints for executing foundational quantum algorithms on superconducting hardware platforms.

Tagged: