Quantum error correction is the "Achilles' heel" of the entire industry. Without solving this problem, scaling quantum processors and their practical application, including the potential threat to cryptography, remain an unattainable goal. However, recent achievements by Microsoft Quantum and IBM Research demonstrate that we are moving from theory to real engineering solutions.
Microsoft and Quantinuum: 800x Error Reduction
A joint study by Microsoft Quantum and Quantinuum, published in the prestigious journal Nature, represents a significant step forward. In their work, the teams demonstrated a reduction in the logical error rate on the Quantinuum ion processor by a factor of 11 to 800 compared to equivalent physical circuits.
The success is based on the use of two specialized designs: a 12-qubit code inspired by the Knill scheme, and a 16-qubit tesseract color code. The former encodes two logical qubits, the latter four. During experiments involving computations with up to 12 logical qubits, impressive metrics were achieved. When preparing the Bell state, the error rate dropped from 0.8% to 0.001% — an 800-fold improvement.
This is not just a laboratory success. Microsoft emphasizes that these results prove that modern quantum devices are already capable of using fault tolerance to suppress errors in non-trivial circuits. Previously, the partners had already demonstrated over 14,000 error-free experiments and stable operation of 12 logical qubits.
IBM Research: AI Guarding Quantum Stability
In parallel, IBM Research presented an innovative approach to finding new correction codes using large language models (LLMs). Using the open-source library OpenEvolve, the system generated 465 candidates for new correction codes.
Of particular interest is the found code [[288,50,8]], which allows packing 50 logical qubits — a significant leap compared to the previous record of 16 for this family of bivariate bicycle codes. A compact code [[72,4,8]] was also discovered, which may be more efficient for certain architectures.
Although the practical applicability of these candidates still requires verification, the approach itself opens new horizons. IBM has already announced plans to build the IBM Quantum Starling quantum computer with 200 logical qubits by 2029, the architecture of which will rely on these developments.
Expert opinion: These two events are not just scientific publications. They signal that the industry has found two fundamentally different but equally effective ways to solve the error problem: engineering (Microsoft) and algorithmic (IBM). For the crypto industry, this is both a challenge and a reminder: post-quantum cryptography is not far off, and preparation needs to begin now, not when the first quantum computer cracks RSA-2048.