The Path to Fault Tolerance

Throughout this site, you've encountered a recurring theme: today's quantum computers are in the NISQ era, too noisy and small-scale for the most dramatic promised applications. This article pulls together what's scattered across many other pages into one clear picture of what fault tolerance actually requires, and how close the field really is.

Step one: get below the fault-tolerance threshold

The fault-tolerance threshold theorem proves that if physical qubit error rates can be pushed below a certain critical threshold, then adding more physical qubits to an error-correcting code (increasing what's called the code's "distance") can suppress the resulting logical qubit error rate arbitrarily — in principle enabling computations of any length, given enough physical qubits.

This is precisely what made Google's Willow processor (covered in our Hardware Database) significant: it was among the first demonstrations that increasing surface code distance actually does decrease logical error rate in practice, not just in theory — confirming the field has crossed this critical threshold, at least for the specific code and hardware tested.

Step two: scale up logical qubit count

Crossing the threshold tells you error correction can work in principle — it doesn't yet mean you have enough logical qubits to run useful algorithms. Current estimates for codes like the surface code suggest needing several hundred to over a thousand physical qubits per reliable logical qubit, depending on the target error rate.

Running Shor's Algorithm to break real-world RSA encryption is generally estimated to require thousands of logical qubits — meaning potentially millions of physical qubits, far beyond even the largest processors covered in our Hardware Database today.

Step three: reduce the overhead

Given how steep the physical-to-logical qubit ratio is, much of current fault-tolerance research focuses on reducing this overhead — through better decoders (like the neural network decoders covered in our Research Papers), more efficient error-correcting codes than the standard surface code, and hardware-level improvements to gate fidelity that reduce how much redundancy is needed in the first place.

The bridge technology: error mitigation

While the field works toward full fault tolerance, quantum error mitigation techniques offer a practical, lower-overhead alternative for extracting useful results from today's NISQ hardware — accepting some residual error rather than eliminating it entirely, in exchange for not needing the massive qubit overhead full error correction demands.

This is why algorithms like VQE and QAOA — both designed to be relatively shallow and error-mitigation friendly — remain the dominant near-term algorithms, rather than deeper algorithms requiring full fault tolerance to run reliably.

Different hardware platforms, different paths

As detailed throughout our Hardware Database, different qubit technologies are pursuing fault tolerance with different tradeoffs: superconducting platforms emphasize manufacturing scalability and gate speed, trapped-ion and QCCD platforms emphasize gate fidelity and connectivity, and approaches like Xanadu's GKP qubits pursue error correction encoded directly within continuous-variable photonic states rather than using many discrete physical qubits per logical qubit in the traditional sense.

What "useful, fault-tolerant quantum computing" will actually look like

Based on current trends, most realistic roadmaps anticipate a gradual transition rather than a single dramatic milestone: NISQ algorithms with error mitigation continuing to improve and find narrow practical applications, while logical qubit counts climb slowly, eventually enabling small-scale fault-tolerant demonstrations, and only later, large-scale applications like full cryptographically-relevant factoring.

Frequently Asked Questions

How many years until fault-tolerant quantum computing arrives?

Industry roadmaps from major hardware companies generally target meaningful fault-tolerant milestones within the next 5–15 years, though predictions in this field have historically run optimistic, and the scaling and overhead-reduction challenges described above remain substantial.

Does crossing the fault-tolerance threshold mean we're basically done?

No — crossing the threshold is a necessary proof-of-concept, but scaling from a small experimental demonstration to the thousands of logical qubits needed for major applications like Shor's Algorithm remains a substantial, separate engineering challenge, likely to take additional years even with continued threshold confirmations.