Quantum Fault Tolerance Roadmap
Practical quantum computing sits on a four-stage ladder:
- Below threshold — adding physical qubits exponentially suppresses logical errors. Google Willow crossed this in 2024 with Λ=2.14 per 2 distance steps.
- Break-even — error-corrected computation beats raw physical hardware. Quantinuum crossed this in March 2026 (94 logical qubits, iceberg codes); IBM's <480ns qLDPC decoder enables it for superconducting.
- Quantum advantage — solving a useful problem faster than any classical system. IBM targets end of 2026 for verified advantage.
- Full fault tolerance — running arbitrarily long algorithms reliably. IBM targets 2029 for the first large-scale fault-tolerant quantum computer.
The roadmap is no longer speculative: hardware is shipping, below-threshold is proven, break-even crossed, and the engineering problem is fab + decoder + scale. Three architectures are now running parallel paths to fault tolerance, and all three have formal commercial backing in 2026:
| Architecture | Anchor company/lab | 2026 milestone | Backer |
|---|
| Superconducting | IBM (Nighthawk, qLDPC) | Verified quantum advantage end-2026 | IBM |
| Superconducting (2nd lane) | Google Willow | Below threshold + lattice surgery primitive | Google + Alphabet |
| Trapped-ion | Quantinuum (iceberg codes) | 94 logical / 98 physical @ break-even | Honeywell / Cambridge Quantum |
| Neutral atom | QuEra, Atom Computing, Pasqal | 100k atoms/chamber target; first logical qubits demos | Google (QuEra), Microsoft (Atom), independent (Pasqal) |
Key Claims
- IBM targets verified quantum advantage end of 2026 — Evidence: strong (IBM)
- IBM targets large-scale fault tolerance by 2029 — Evidence: strong (IBM)
- Google Willow crossed below-threshold in Aug 2024 — Λ=2.14 per 2 distance steps; 0.143% error/cycle at d=7. Evidence: strong (Willow)
- Quantinuum crossed break-even at scale in March 2026 — 94 logical qubits on 98 physical. Evidence: strong (Quantinuum)
- First superconducting lattice surgery demonstrated Jan 2026 — Nature Physics; compute primitive on surface-code qubits. Evidence: strong (Besedin)
- Google adopts dual-modality quantum strategy Apr 2026 — adds neutral atom via QuEra alongside Willow. Evidence: strong (Google+QuEra)
- Neutral-atom scaling targets 100,000 atoms/chamber — QuEra + Atom Computing. Evidence: moderate (Google+QuEra)
IBM Roadmap
| Milestone | Target |
|---|
| Nighthawk delivery (120 qubits, 5,000 gates) | End of 2025 |
| Verified quantum advantage | End of 2026 |
| Nighthawk+ (7,500 gates) | End of 2027 |
| 15,000 gates, 1,000+ qubits | 2028 |
| Large-scale fault-tolerant quantum computer | 2029 |
Quantinuum Trajectory
| Milestone | Date |
|---|
| Beyond break-even at scale | March 2026 |
| 94 logical qubits, 98 physical | March 2026 |
| Partially fault-tolerant (postselection-based) | 2026 |
| Scaling iceberg codes to larger logical sets | 2026-2028 |
Google Trajectory (Two Lanes)
| Lane | Milestone | Date |
|---|
| Willow (superconducting) | Below threshold, Λ=2.14 | Aug 2024 |
| Willow (superconducting) | Lattice surgery primitive | Jan 2026 (Besedin et al.) |
| Atlantic Quantum (fluxonium) | Integrated into Willow roadmap | Oct 2025 |
| QuEra (neutral atom) | Strategic investment + internal program | Apr 2026 |
| Google internal neutral atom | Led by Adam Kaufman (CU Boulder) | Apr 2026 |
Neutral Atom (Third Architecture) Trajectory
| Company | Milestone | Date |
|---|
| Pasqal | 1,000 qubits reached | 2024 |
| Pasqal | 2 logical qubits demonstrated (European first) | 2025 |
| Pasqal | 250-qubit QPU for quantum advantage attempt | First half 2026 |
| Pasqal | 10,000 qubits target | 2026 |
| QuEra | Error-correction-ready machine to AIST Japan | 2025 |
| QuEra + Google | Strategic partnership | April 2026 |
| Atom Computing + Microsoft | Phoenix system on Azure Quantum | Prior 2026 |
| Industry | 100,000 atoms / chamber target | "Next few years" |
Open Questions
- What is the first "verified" quantum advantage problem — cryptography (factoring, discrete log), materials simulation, optimization, sampling?
- Do historical AI-style roadmap timelines apply (usually pushed back)?
- Which architecture reaches 2029 fault tolerance — superconducting, trapped-ion, or neutral atom?
- Does Google's dual-modality (Willow + QuEra) hedge pay off, or does it dilute focus?
- At what qubit count does classical simulation become infeasible enough to verify advantage?
- Is 100k atoms/chamber achievable with sub-threshold error rates, or does scaling degrade fidelity?
Related Concepts
Backlinks
Pages that reference this concept:
Changelog
- 2026-04-17 (initial) — Compiled from Quantinuum + IBM sources.
- 2026-04-17 (update) — Added Google Willow below-threshold data, Besedin lattice surgery, Google+QuEra neutral atom expansion. Expanded from 2-architecture to 3-architecture race (+ Google as a second superconducting lane). Added neutral atom trajectory table.