Quantum error correction has always demanded a cruel tradeoff: protect one logical qubit by spreading its information across many physical qubits, but each added physical qubit introduces more noise. If the physical error rate is too high, adding more qubits makes things worse, not better. The surface code threshold — roughly 1% error per operation — marks the boundary where error correction transitions from helpful to harmful.
Google's 101-qubit Willow processor (arXiv:2408.13687, Nature 2025) crossed that boundary. A distance-7 surface code achieved 0.143% error per cycle, and increasing code distance from 5 to 7 suppressed logical error by a factor of 2.14. The logical memory exceeded the lifetime of its best physical qubit by a factor of 2.4 — the first demonstration of break-even quantum error correction on a surface code.
The result matters not for its magnitude but for its direction. Previous surface code demonstrations showed that more qubits helped, but didn't reach the point where the logical qubit outlived its physical components. Breaking even means the error correction isn't just slowing degradation — it's providing net positive protection. The curve bent the right way.
Harvard's neutral-atom team and Quantinuum's ion-trap system achieved parallel milestones, suggesting that below-threshold operation isn't a quirk of one hardware platform but a capability emerging across multiple architectures simultaneously. The convergence from three independent hardware approaches — superconducting, neutral atom, trapped ion — within months of each other implies the limiting factor was engineering maturity, not fundamental physics.
The transition from above-threshold to below-threshold isn't a continuous improvement — it's a phase transition. Below threshold, scaling works. Above threshold, scaling fails. The distance between 1.1% and 0.9% error is the difference between a technology that can never scale and one that can.