Fault-Tolerant Quantum Computing: The Leap from Feasible to Useful

For much of the past decade, the dominant story about quantum computing has been that more qubits mean more computing power. Experimental systems progressed from a handful of qubits to tens and then hundreds, while reported control fidelities rose from 99% to 99.5%, 99.9%, and 99.97%. Yet meaningful universal quantum computation will require circuit depths in the millions or tens of millions. Even at 99.999% fidelity, the probability of an entirely correct computation after one million steps is only 0.0045%; at ten million steps, the probability begins only after 43 zeros following the decimal point.

In other words, a present-day system may execute a limited quantum circuit at high fidelity, but accumulated errors can destroy the computation as its depth increases. More qubits or better devices alone cannot solve this underlying problem. Fault-tolerant quantum computing is the real entry ticket to universal quantum computing.

Error correction is not a recent idea. In the 1990s, the fragility of quantum states raised a fundamental question: if even a small environmental interaction causes decoherence, can large-scale quantum computation ever work? The impasse was broken around 1995. Shor proposed the first complete quantum error-correcting code, and Steane and others developed more systematic encoding methods. Their work showed that quantum information can be encoded in entangled structures across multiple physical qubits to resist local noise. The threshold theorem then established that, provided the physical error rate remains below a threshold, quantum error correction can suppress logical errors arbitrarily far. Together, these results demonstrated that stable and scalable quantum computation is possible in principle despite fragile physical qubits.

Experimental technology advanced rapidly during the 2010s. Superconducting qubits, trapped ions, and other leading platforms achieved control over tens and then hundreds of qubits, ushering in the noisy intermediate-scale quantum, or NISQ, era. These devices can execute circuits of limited scale and show potential advantages on selected shallow-circuit tasks. But without error correction, even hundreds of physical qubits support only limited circuit depth; as depth grows, the probability of success trends toward zero.

This does not mean researchers forgot the importance of error correction. Qubit count and fidelity are prerequisites for it. Error correction uses many physical qubits to protect one logical qubit, the unit that performs the computation. Without enough physical qubits, a stable logical qubit cannot be encoded. Without sufficient fidelity, correction operations introduce more errors than they remove.

NISQ is therefore not the key that opens the door to quantum computing, but it is a necessary transitional stage. It has demonstrated controllability and potential quantum advantage while building the prerequisites for fault tolerance. A complete correction mechanism and system-level implementation are the final pieces needed before the door can open fully.

If error-correcting codes and the threshold theorem established large-scale quantum computing in principle, NISQ systems have demonstrated precise control over tens or hundreds of qubits, and error-correcting codes have existed for three decades, why is fault tolerance still an open engineering challenge?

The nature of the problem changes when the objective shifts from more qubits, higher fidelity, or a better code toward a complete fault-tolerant quantum-computing system. The central challenge is no longer a single technical breakthrough. It is organizing physical implementation and error correction under common constraints so that the entire system can operate continuously and reliably. The difficulty emerges in the transition from coding theory to an engineered system.

Resource requirements are severe because stable logical qubits need substantial physical-qubit redundancy. More importantly, quantum error correction cannot be handled after the fact. It is a continuous real-time loop: the system must repeatedly measure error information, infer faults, and immediately apply feedback. If this loop cannot keep pace with the physical system, errors accumulate and spread. Typical correction cycles run on microsecond timescales, creating stringent real-time control requirements.

Fault tolerance is therefore neither only an algorithm problem nor only a device problem; it is a system-engineering problem constrained by time, resources, and coordination. It couples precise physical-qubit control, code optimization, decoding strategy, real-time control, compilation, and mapping. Delay or mismatch at any layer can destabilize the entire system. The real challenge is constructing a whole system that operates reliably under shared constraints.

If the past decade answered whether quantum computing is worth pursuing, the question now is when it can be made to work. Progress in algorithms and hardware has shown that quantum computing can outperform classical computing on selected tasks. Its long-term potential is no longer the central point of debate.

At the same time, these advances identify a common bottleneck: without fault tolerance, quantum advantages cannot be amplified or sustained. Fault tolerance is no longer one research direction among many; it is the decisive threshold for practical quantum computing. Before that threshold, progress primarily demonstrates capability. Beyond it, quantum computing becomes a scalable and usable computing system.

Classical computing offers a useful analogy. Early storage media were unreliable, and random bit flips accumulated as systems grew. Error-correcting-code memory added redundancy and verification above the physical bits, allowing systems to detect and repair errors during operation and turn unreliable hardware into dependable large-scale storage. Communication systems similarly rely on error correction across physical and link layers to transmit information over noisy channels.

Quantum and classical error correction, however, are fundamentally different. Classical storage and communication can compare redundant data and repair a detected error directly. Quantum computing cannot make an equivalent direct comparison because measurement collapses the quantum state and destroys the computation. This is why Shor’s quantum error-correcting code was such a foundational breakthrough: without it, quantum computing cannot scale or become practical. Classical error correction supplements classical computing; quantum error correction is a core foundation of quantum computing itself.

Fault tolerance is therefore both the first and the final step from feasible quantum computing to useful quantum computing. Without it, there is no practical quantum computation. Once the threshold is crossed, quantum computing gains access to scalable general-purpose computing power.

From an industry perspective, an equally important question is who is approaching this threshold that will shape the future of quantum computing.

Competition in fault-tolerant quantum computing has entered a new stage. International teams including Google and IBM continue to advance surface-code experiments, logical-qubit construction, and real-time correction systems. Their progress shows fault tolerance moving from theory toward early system validation. Chinese teams are also advancing in error-correction theory, system simulation, control systems, and engineering implementation. In decoding algorithms, architecture, and hardware–software co-optimization, domestic work is on the same technical trajectory as the international frontier and is beginning to develop distinctive approaches.

As engineering investment grows, competition is shifting from isolated technical results toward system capability. Future differentiation will depend not only on a leading individual metric but on the ability to build a complete and stable fault-tolerant system. Fault-tolerant quantum computing is therefore both a core technical challenge and a strategic focus of global technology competition.