Quantum error correction (QEC) is the technique that will make large-scale quantum computers possible. Without it, errors accumulate faster than computations complete, making the machine useless. In 2024–2025, multiple laboratories have now demonstrated logical error rates below the fault-tolerance threshold for the first time.
"The fault-tolerance threshold is not a single number — it depends on the error model, the code, and the decoder. For the surface code with realistic noise, the threshold is roughly 1% physical error rate. Getting below it means logical errors decrease as you add more physical qubits."
The Surface Code
The surface code is the leading candidate for fault-tolerant quantum computing on superconducting platforms. It arranges qubits on a 2D grid and uses syndrome measurements — non-destructive parity checks — to detect errors without collapsing the quantum state. The code distance d determines the number of errors that can be corrected: a distance-d surface code corrects up to ⌊(d−1)/2⌋ errors.
| Code Distance | Physical Qubits | Logical Error Rate (p_phys=0.1%) |
|---|---|---|
| d=3 | 17 | ~10⁻⁵ |
| d=5 | 49 | ~10⁻⁸ |
| d=7 | 97 | ~10⁻¹¹ |
| d=11 | 241 | ~10⁻¹⁷ |
Recent Experimental Results
Google's Willow chip demonstrated logical error rate suppression of 2.914× per code distance step, achieving a logical error rate of ~10⁻⁶ for a distance-7 code. IBM's Heron processor showed similar scaling with their own surface code implementation. IQM in Finland demonstrated a distance-3 surface code with a logical error rate below 10⁻³ — below threshold for their hardware.
The Path to Fault-Tolerant Quantum Computing
Estimates suggest that running Shor's algorithm on a 2048-bit RSA key requires approximately 4,000 logical qubits at a logical error rate of 10⁻¹². With current surface code overheads (~1,000 physical qubits per logical qubit), that means roughly 4 million physical qubits. Google's 2030 target is 1 million physical qubits.
Key Takeaways
- Below-threshold error correction is now experimentally verified on multiple hardware platforms.
- The surface code is the leading approach, but alternatives (colour codes, LDPC codes) may offer better overhead.
- Millions of physical qubits are still required for practically useful fault-tolerant computation.
- Fast, accurate classical decoders are increasingly the bottleneck — not the qubits themselves.