Fault Tolerance Without Thousands of Physical Qubits
Nord Quantique, a quantum hardware company based in Sherbrooke, Canada — which raised a CAD 30 million growth round from Fidelity Canada in spring 2026 at a valuation of roughly USD 1.4 billion, making it Canada's fourth quantum unicorn after D-Wave, Xanadu, and Photonic — announced a key milestone in July 2026: SPAM (state preparation and measurement) error rates for a single-mode grid-state qubit have been pushed below 0.1%, roughly a 100x improvement over comparable GKP systems and now on par with mainstream transmon platforms.
Why "Bookend" Errors Matter
Every quantum computation involves setting the qubit to the correct initial state (state preparation) and reading out the final answer (measurement). As CEO Julien Camirand Lemyre put it plainly: if either of these steps fails, even flawless computation in between yields unreliable results. SPAM errors are "bookend" errors — occurring *before* and *after* error-correction logic — so they can silently undermine even the most sophisticated correction protocols.
A Different Route: 1 Physical Qubit per Logical Qubit
IBM and Google rely on surface codes, requiring hundreds to thousands of physical qubits per logical (error-corrected) qubit — a widely acknowledged scaling bottleneck. Nord Quantique takes the opposite approach: bosonic encoding in superconducting microwave cavities, specifically the Gottesman–Kitaev–Preskill (GKP) grid state and its multimode extensions, embedding one logical qubit directly into one superconducting cavity. This targets a potential 1:1 physical-to-logical ratio, dramatically reducing the hardware, power, and infrastructure needed for fault tolerance. The company describes it as "embedding error correction into every qubit and rebuilding computation from the qubit up."
How It Works
The core is a repeat-until-success stabilization protocol: prepare a state, verify success, and keep it if successful or discard and retry if not. This avoids the overhead of real-time error correction and complex classical control systems while preserving low logical error rates. The same protocol is also used to prepare magic states — the non-Clifford resource states required for universal quantum computing — where high-fidelity preparation is considered one of the most resource-intensive challenges across all architectures.
Timeline
- 2021: First grid-code implementation in a superconducting cavity
- 2023: Industry-first QEC demonstration with a bosonic code
- 2024: Demonstrated 1:1 physical:logical encoding in PRL (best encoding ratio for this class of system)
- 2025: Demonstrated the Tesseract multimode bosonic code maintaining quantum information across 32 consecutive error-correction cycles
- 2026 (July): SPAM errors below 0.1%
Why It Matters
Today's quantum narrative centers on surface codes (Google Willow, IBM Starling) and AI decoders (QXL / Quantum X Labs). Nord Quantique's GKP bosonic route offers an underappreciated counterexample: fault tolerance doesn't only come from stacking qubits — it can be built into every qubit itself. Breaking the 0.1% SPAM barrier fixes the long-standing "bookend error" weakness of the GKP route, and if the 1:1 encoding ratio holds, it could rewrite the hardware economics of fault-tolerant quantum computing. Lemyre admits the result isn't "flashy" — but it's exactly these unsexy fixes that are required homework before fault-tolerant quantum computing becomes practical.
In one sentence: when error correction is welded into every qubit, fault tolerance may not require thousand-qubit arrays.
Sources: Nord Quantique research paper (2026-07), BetaKit, QuantumWay company analysis, niconaut technical briefings; public DARPA QBI / CQCP information.