Charged excitons in monolayer MoSe₂ interfaced with a magnetic substrate carry negative valley polarization that flips sign as temperature rises. The result is counterintuitive, and the mechanism behind it points to a deeper truth about how information encoded in quantum states survives its environment. That same architectural insight is now driving the global race to deliver fault tolerant quantum computing, and it is being mirrored, accidentally, in a cloud storage product shipped in July 2026. Quantum error correction, the engineering discipline that turns fragile physical qubits into reliable logical qubits, sits at the center of both stories, and 2026 is the year it stops being a laboratory curiosity.
Why These Two Signals Belong Together
This matters because both signals expose the same engineering principle that quantum error correction formalizes from opposite ends of the computing stack. In a 2021 paper on monolayer MoSe₂ interfaced with bismuth iron garnet (DOI: 10.1088/2053-1583/ac3887), a research team showed that charged excitons carry a valley polarization that can be stabilized — and even reversed — through the choice of magnetic substrate. In July 2026, encrypted cloud provider NordLocker, owned by Nord Security, shipped "Project Renaissance," an atomic metadata logging architecture that protects file integrity through append-only writes. The two technologies operate at vastly different scales, but the design logic is identical to what quantum error correction demands at the qubit level: protect encoded information by choosing the right interface, whether that interface is a ferrimagnetic crystal or a metadata log. The timing is not coincidental — both efforts respond to the same observation that information, in any substrate, is only as durable as the architecture around it.
How It Works
Monolayer MoSe₂ possesses a valley degree of freedom — the K and K′ points of its hexagonal Brillouin zone — that physicists treat as a binary label for electrons, much like spin-up and spin-down. When the semiconductor sits on a thin film of bismuth iron garnet, a ferrimagnetic insulator, the out-of-plane magnetization of the substrate breaks the symmetry between the two valleys. The result, observed in circularly polarized photoluminescence, is a population imbalance that the team captured as negative valley polarization in charged excitons, or trions. That imbalance flips sign with temperature, an effect the authors attribute to a combination of magnetization-induced carrier redistribution and valley-switching scattering from bright to dark excitons. Trions, in particular, carry longer valley lifetimes than neutral excitons, which is why the MoSe₂ community has converged on them as a probe for valley physics.
The paper's central claim, drawn directly from its abstract, is that "interfacing atomically thin van der Waals semiconductors with magnetic substrates enables additional control on their intrinsic valley degree of freedom." That sentence reads like a basic observation. It is not. The ability to tune a valley state by choosing the right interface is the same lever that quantum error correction pulls when it uses surface codes to encode a single logical qubit across roughly one thousand physical qubits. The surface code, formalized in the late 1990s by Daniel Gottesman — now at the University of Maryland's Joint Center for Quantum Information and Computer Science — and refined by MIT's Peter Shor, has become the de facto standard. Caltech's John Preskill, who coined the term "quantum supremacy," has argued for years that fault tolerant quantum computing requires substrate-level protection of the kind the MoSe₂ paper now demonstrates.
For practitioners, the analogy is direct. Quantum error correction in its surface-code form is, at heart, a substrate problem: pick the right lattice, and the encoded information survives its environment. In International Business Machines' 1,121-qubit Condor processor, unveiled in December 2023, and the 156-qubit Heron that followed, syndrome measurement — repeatedly querying neighboring physical qubits to detect errors without collapsing the encoded Logical Qubit state — is the operational analogue of a magnetic substrate enforcing valley order. Neither approach eliminates decoherence outright. Both contain it. And both, crucially, are designed so that the protective interface can be swapped out as the underlying hardware improves. The quantum error correction community now talks about "below threshold" the way the MoSe₂ paper talks about negative polarization: a regime in which adding more substrate does not introduce more noise, but absorbs it.
Who's Moving
The hardware side of the quantum error correction race is dominated by publicly traded American firms. International Business Machines Corporation (NYSE: IBM) operates the Condor and Heron superconducting platforms under Jay Gambetta, IBM Quantum's Vice President, and has published a fault tolerant quantum computing roadmap targeting more than 200 logical qubits by 2029. Alphabet Inc. (NASDAQ: GOOGL) crossed the fault tolerance threshold in December 2024 with the 105-qubit Willow chip, the first demonstration that adding physical qubits reduces the logical error rate exponentially. IonQ Inc. (NYSE: IONQ) pursues the same milestone with trapped ions, with its Tempo and Forte systems now reporting two-qubit gate fidelities above 99%. Microsoft Corporation (NASDAQ: MSFT) is the outlier, betting on Topological Qubits based on Majorana zero modes — an approach that has drawn sustained skepticism from independent physicists and has not yet produced a verified logical qubit at scale.
Private capital is moving just as aggressively. PsiQuantum, a photonic quantum computing company based in Palo Alto, has raised more than $700 million in disclosed equity funding across multiple rounds to build a million-qubit machine that uses quantum error correction at every node. Quantinuum, the trapped-ion spinoff of Honeywell, received a $300 million strategic investment from JPMorgan Chase in 2024 to accelerate its fault tolerant research, with a stated target of a universal fault tolerant system by 2030. The closest direct competitor to the surface code itself is the family of bosonic codes used by Amazon Web Services and Yale's Robert Schoelkopf, which encode a logical qubit inside a single high-quality oscillator — fewer physical qubits, but tighter control demands. The trade-off between surface codes and bosonic codes is one of the central architectural questions in quantum error correction today. And NordLocker, the encrypted cloud storage provider owned by Nord Security, is the uninvited guest at the quantum table: its Project Renaissance, deployed in July 2026, replaces a rigid file tree with an atomic metadata logging model. NordLocker reports improvements in metadata write reliability after the migration. The mechanism is classical. The philosophy — protect encoded data by writing to a log no one can rewrite — is the same one IBM and Google apply when they perform syndrome measurement on a Surface Code lattice.
Why 2026 Is Different
Three timelines for quantum error correction are now converging. In the next twelve months, IBM, Google, and Quantinuum will publish the first logical qubit counts that exceed proof-of-concept, with IBM's roadmap explicitly targeting a multi-logical-qubit demonstration by the end of 2026. Within three years, by mid-2029, at least two vendors expect to deliver systems on which a logical qubit outperforms its best physical qubit by a measurable margin — the operational definition of fault tolerant quantum computing. Within five years, on a 2030 horizon, cryptographically relevant quantum computers will begin to threaten the RSA-2048 baseline that protects most internet traffic, and the first commercial quantum key distribution networks will go live to replace it. McKinsey, in its 2024 Quantum Technology Monitor, sized the addressable market for quantum hardware, software, and services at $106 billion by 2040. The roadmap to fault tolerant quantum computing is no longer speculative. Real-time classical decoders that process syndrome measurements faster than qubits decohere are operating in research settings, and modular quantum architectures that connect smaller QEC modules into larger ones are under active development at IBM and PsiQuantum. The MoSe₂/BIG work and the NordLocker launch both point to the same conclusion: the next decade of computing will be defined less by transistor counts and more by the architecture of error correction at every layer of the stack.
The lesson from MoSe₂ and from NordLocker is identical: a quantum of information, whether it lives in a valley or a vault, is only as durable as the architecture that surrounds it. That architecture is finally arriving.
In short: quantum error correction crossed the fault tolerance threshold in December 2024 with Google's 105-qubit Willow chip, and 2026 is the first year in which IBM, Google, and Quantinuum will publish logical qubit counts that matter for cryptography, materials science, and optimization alike.