An arXiv preprint shows single-molecule ODMR in a designer carbene with 2.2 ms coherence at 4.5 K. Two days later, NVision announced a $55 million Abbott-led round and a photonic molecular-qubit architecture still demonstrated at just one emitter.
A German quantum-imaging company best known for POLARIS, a hyperpolarization platform being installed at major cancer and research centers, announced on May 13 that it had raised $55 million and was expanding into quantum computing with a platform built from organic molecules. Two days earlier, an NVision-led team had posted a preprint to arXiv reporting optically detected magnetic resonance on a single triplet carbene molecule, with a 21 millisecond spin relaxation time and 2.2 millisecond coherence under dynamical decoupling at 4.5 K.
The two facts are connected. The coherence numbers are an order-of-magnitude improvement over prior molecular-ensemble measurements, and they come out of the same molecular-spin engineering program NVision built while developing POLARIS. The corporate move is the company's argument that the electron-spin control it used to enhance hyperpolarized MRI tracers is now mature enough to be a qubit modality.
The experiment is significant. The company is mid-stage. The quantum-computer architecture it describes is, on the bench, one molecule.
NVision's qubit is a triplet ground-state carbene, an organic molecule with a divalent carbon whose two unpaired electrons form a triplet, S = 1, ground state. Embedded in a structurally matched ketone host crystal, the molecule emits photons at a precise frequency and carries an electronic spin that can be initialized, microwave-controlled, and read out optically. The control playbook resembles the one used for nitrogen-vacancy centers in diamond: optical initialization and readout coupled to microwave control of an electronic spin.
The molecule is not an NV center; the point is that it offers a chemically designed analogue of an optically addressable spin defect. An NV center is a specific defect in diamond. A carbene can be chemically designed. The host, the ligands, the symmetry, and therefore the spin Hamiltonian are chemistry knobs. The arXiv paper frames this as "synthetic control" over an optically addressable spin qubit. The team grows ketone host crystals doped with a stable diazo precursor, cools selected crystals to 4.5 K in a cryogenic confocal microscope, and photoactivates the precursor with light to generate the carbene in situ.
The figures of merit reported in the preprint:
Metric | Value |
|---|---|
Operating temperature | 4.5 K |
Spin relaxation T₁ | 21 ± 2 ms |
Hahn-echo T₂ | 12.2 ± 0.6 µs |
Dynamical-decoupling T₂ (XY8-NN) | 2.2 ± 0.3 ms |
Excited-state lifetime | 24 ± 2 ns |
Spectral diffusion / spectral stability (one hour) | σ = 2.6 MHz |
Addressability | single-molecule ODMR demonstrated |
That is the milestone: a single molecule, read out optically, with a millisecond-class coherence time after pulse protection. For dynamical-decoupling coherence, the preprint claims more than an order-of-magnitude improvement over prior molecular-ensemble measurements, and it clears the bar that color-center physicists set when they argue organic molecules can compete with diamond NV centers as the unit cell of a photonic quantum architecture. A field-overview paper from Mann and Bayliss at Glasgow reviewed the case for optically addressable molecular spin qubits in 2026; the NVision result is arguably the strongest single-molecule data point for that case so far.
What the preprint does not contain is also part of the story. There is no two-qubit gate. There are no reported single-qubit or two-qubit gate fidelities. There is no error-correction demonstration. It also does not report two-emitter interference, deterministic molecule placement, photonic-chip coupling efficiency, or entanglement generation between molecular nodes. The dynamical-decoupling T₂ is pulse-protected; the bare Hahn-echo T₂ is twelve microseconds. The 4.5 K operating point is closed-cycle helium, not room temperature, despite some commentary framing organic-molecule platforms as inherently warmer than the alternatives. The paper was submitted to arXiv on May 11 and has not been peer-reviewed.
A working multi-qubit chip has not been disclosed.
NVision was founded in Ulm in 2015 as NVision Imaging Technologies GmbH, a spinout from the University of Ulm's institutes of Theoretical Physics and Quantum Optics. The company traces back to Ulm quantum-optics and theoretical-physics groups associated with Fedor Jelezko, Martin Plenio, Alex Retzker, and Ilai Schwartz; its current leadership presents Sella Brosh as co-founder and CEO. Several of the same scientific founders and Ulm-linked researchers appear on the preprint a decade later.
The company's original product, POLARIS, is NVision's hyperpolarization platform for metabolic MRI. NVision says it can boost signals from metabolic imaging agents such as [1-¹³C] pyruvate by more than 10,000 times, potentially making early metabolic treatment response visible much sooner than anatomical imaging. But the regulatory and clinical framing matters: NVision currently describes POLARIS as a preclinical research system, with investigational clinical-system development ongoing and hyperpolarized agents not approved for routine clinical use. POLARIS is being installed at Memorial Sloan Kettering Cancer Center, the University of Cambridge, and Technical University of Munich, with a year-end 2026 target of roughly 20 installations across the US, Europe, and Asia, per the company press release.
NVision says PIQC, the new quantum-computing platform, marketed as "Pixie," emerged from the molecular-spin engineering it developed while building POLARIS. The hyperpolarization work uses electron-spin control to transfer polarization onto metabolic imaging agents such as hyperpolarized [1-¹³C] pyruvate. PIQC proposes to use the same family of electron spins on organic molecules as the qubits themselves. The corporate continuity is the same chemistry team, the same instrumentation, and the same regulatory relationships. The customers, the products, and the unit economics are entirely different.
The $55 million is structured as equity plus a $17 million venture loan from the European Investment Bank. Abbott led as the sole strategic diagnostics investor; CEO Sella Brosh's "quantum-enhanced drug discovery" framing in the release reads as the pitch Abbott bought. Playground Global, Matterwave Ventures, Entrée Capital, and CDP Venture Capital participated, according to Quantum Insider's coverage. Cumulative funding now sits at roughly $120 million. The company began presenting itself as NVision Quantum Technologies on May 13.
As a financing event, that is plausible few-qubit-demonstrator money, not fault-tolerant-system money. It is less than half of IonQ's reported 2025 annual revenue.
The closer parallel is not any one modality, but the broader pattern: new quantum hardware stories become investable before they become fault-tolerant systems. The companies that survive that gap are the ones that move quickly from physics milestone to multi-qubit operations, publish fidelity numbers, and show a credible manufacturing path.
That is the context the announcement lands inside. Quantinuum's H2 recently used 56 trapped-ion qubits for quantum-magnetism simulations that severely challenged several leading classical simulation approaches, with reported native two-qubit-operation fidelities around 99.94 percent. Microsoft and Atom announced 24 entangled logical qubits on neutral atoms in late 2024, with the system positioned for 2025 availability. The contrast is that the leading modalities are now publishing or roadmapping at far larger physical-qubit counts: Google has shown below-threshold surface-code memories, IBM is pushing modular 120-qubit-class processors, and IonQ is selling an aggressive path toward fault-tolerant trapped-ion systems. NVision is at month one of being a quantum-computing company.
The case for molecular is not qubit count or fidelity. It is upstream of both.
NVision's claim is that designer organic molecules can be deposited as thin organic layers directly onto photonic chips using standard semiconductor processes, with quantum operations performed via optical addressing through the photonic circuit. That is a fabrication-path story, not an applications story.
The comparison is to the engineering each incumbent modality has to do to scale. Trapped ions require vacuum chambers and laser arrays. Neutral atoms require tweezer optics. Diamond NV centers require single-crystal CVD growth and ion implantation with controlled placement. Superconducting qubits require dilution refrigeration below twenty millikelvin and Josephson-junction fabrication at angstrom-scale tolerances. Each maps onto a different supply chain, and most of those supply chains are bespoke.
A platform that deposits a designer molecule onto a photonic waveguide through wafer-scale chemistry would, in principle, inherit the existing CMOS and silicon-photonic supply chain. The missing device metrics are placement yield, spectral uniformity across emitters, photon indistinguishability after integration, coupling efficiency into the photonic mode, cavity/waveguide loss, two-node entanglement rate, and gate fidelity. That is the argument that makes molecular-qubit work interesting as a Chips-layer story rather than as a one-off physics result.
It is also the argument with the longest list of unstated engineering problems behind it. Photolyzing precursor molecules in situ to create carbenes at controlled positions on a chip, at single-molecule density, with registration to optical waveguides, with yields high enough to support multi-qubit devices: none of those are demonstrated. The press materials do not address them, and the preprint reports one molecule.
NVision's molecule-on-photonics thesis sits naturally beside DARPA's HARQ program's broader heterogeneous-integration logic, even if the company is not being presented here as a HARQ awardee.
The networking case is the secondary thesis. Single-photon emission with narrow zero-phonon lines and 2.6 MHz spectral stability over an hour is a promising spin-photon-interface story, though not yet a networked-node demonstration. That matters for distributed quantum architectures the way Xanadu's March public-market debut signaled the photonic-quantum thesis as networking-native. NVision is not the only group working on this. A separate team reported erbium-based molecular spin-photon interfaces at telecom wavelengths in 2025. The molecular-qubit field is broader than NVision; NVision is the well-capitalized commercial entrant.
The European Investment Bank putting $17 million of venture debt into a $55 million round is unusual outside of public-policy-aligned deep tech. The DLR Quantum Computing Initiative lists NVision as a partner; the DLR QCI runs an innovation center in Ulm. The arXiv preprint acknowledges funding from the State of Baden-Württemberg, the German Research Foundation, BMWK's COMIQC project, and ERC grant QRES #770929. Germany's High-Tech Agenda Deutschland targets at least two error-corrected quantum computers at European-top-level by 2030. The EU Quantum Strategy emphasizes industrialisation, supply-chain resilience, and startup scale-up, while the EU Chips Act has backed six complementary quantum pilot lines, including the SPINS pilot line for semiconductor spin qubits. That makes NVision unusually well-aligned with Europe's desire for differentiated quantum hardware, not just local versions of US-led modalities. That is why the round closed where it did.
The next milestone is either heralded entanglement between two molecular nodes or an entangling two-qubit gate on an integrated photonic device, with a fidelity and rate the company is willing to publish. That is the threshold separating a credible single-molecule physics demonstration from a quantum-computing company. The $55 million is roughly the right amount of money to attempt it. The team and the chemistry are roughly the right ones to attempt it. Whether the manufacturability thesis survives contact with the second qubit is the question the next paper has to answer.
