Molecular Qubit Breakthrough Moves to HyperLight Photonic Chips

A new kind of qubit tries to get onto a chip

Ulm-based NVision made two announcements on October 28. The first is a peer-reviewed paper in Nature reporting what the company calls the first organic molecular spin-photon interface that can be addressed one molecule at a time. The second is a partnership with HyperLight, a maker of thin-film lithium niobate (TFLN) photonics, to put those molecular qubits directly onto photonic integrated circuits.12 The paper is titled A Single-Molecule Spin-Photon Interface. NVision's main claim is that molecular spins could not previously be controlled individually and coherently, so they could not work as real qubits, and that this barrier is now gone.2

The two announcements are being treated as one story, and they should not be. The Nature result is a real physics milestone, backed by measured numbers. The HyperLight deal is a plan for building hardware, and no combined chip has been tested yet. Most of the coverage, apart from the press releases, makes the same distinction.319

What was actually shown

The experiment comes from NVision Imaging Technologies, Ulm University and the Hebrew University of Jerusalem. The team used synthetic carbene molecules, which are organic molecules built around a carbon atom with two free bonding sites. They showed they could prepare the molecule's spin state with light, control it coherently, and read it out optically.5 The carbene sits inside a closely matched organic host crystal. A focused laser removes nitrogen from a precursor compound, which creates each individual qubit inside the crystal.14

The headline numbers were all measured at 4.5 kelvin:

• Coherence: 2.2 milliseconds, extended using dynamical decoupling (a sequence of microwave pulses that shields the spin from noise)
• Spin relaxation time: 21 milliseconds
• Readout light: 595.04 nanometres
• Single-photon purity: a g²(0) value of 0.142, well below the 0.5 threshold, which confirms the light comes from one isolated molecule519

One practical point: 4.5 K can be reached with ordinary helium cryostats. Superconducting machines need much more expensive dilution refrigerators.5

NVision says it reached this molecule by screening more than 150 candidates on computers and synthesizing 40 of them in the lab.1 That method matters for the bigger argument. CTO Ilai Schwartz says chemistry lets designers tune memory, gate speed and photon emission within a single molecule, which fixed semiconductor defects cannot do.4

Where the reporting diverges

The more independent coverage pulls apart several of the company's claims.

Temperature. The release says the carbenes work at "accessible cryogenic temperatures" of 2K to 4K.7 The reported Nature measurements were taken at 4.5 K.519 The gap is small, but one figure is measured and the other is a company description.

Photon-loss tolerance. This is the bigger issue. The press materials say the platform "tolerates up to 70% photon loss," which would let it run on standard silicon photonics instead of specialized ultra-low-loss optics.9 One analysis traces that figure to a separate architecture paper from May 20. That paper describes proposed protocols. It is not a result from the new experiment.8 Another outlet calls the figure a company-supplied claim with no independent integrated-device measurements behind it.3

Optical efficiency. The same analysis finds a less flattering number. The paper reports a 10.8% photoluminescence quantum yield, and only 19% of that light falls in the narrow emission line used for readout. Multiplied together, that is about 2.1% emission efficiency before any collection losses.8 The optical linewidth of 38 MHz is also about 5.8 times wider than the lifetime-derived limit. That matters because photons from separate molecules have to be nearly identical before they can entangle distant qubits.8

Scale language. The GlobeNewswire version says the approach opens a path toward "millions of qubits" on foundry-compatible chips.13 No independent account repeats that figure.

Taken together, my reading is that NVision has shown a strong single-qubit building block whose network-facing optics still need serious work. The 70% loss figure describes what the architecture is designed to tolerate. It does not describe what the hardware has been shown to do.

Why HyperLight, and why thin films

The business case depends on manufacturing. NVision wants to deposit molecular qubits as thin films, much as OLED display materials are deposited, directly onto photonic chips. That would turn existing photonics fabrication into a quantum platform.13 One financial outlet sums it up as OLED know-how turned into a quantum chip blueprint, and notes that thin-film deposition is something the electronics industry already does at scale.10 Elsewhere the company lists film thicknesses of roughly 20 to 100 nanometres.8

HyperLight sells TFLN electro-optic circuits for AI data centers and telecom and metro networks. The deal extends its TFLN Chiplet line into quantum applications.2 It already offers shared and dedicated wafer runs, custom designs and bare-die delivery. That gives NVision a named manufacturing route, though molecular-qubit yields still have to be proven.8 HyperLight co-founder Kevin Luke said photonic interconnects are essential for scaling quantum processors, and that the combination draws on decades of semiconductor investment.1

The design argument is "connectivity without proximity." Most chips pack qubits tightly together, which brings crosstalk, complicated control wiring and heat problems. NVision's Photonic Integrated Quantum Circuit (PIQC) architecture instead links qubits with optical waveguides, within a chip and between chips.25 The idea is sound. Other photonic-interconnect efforts are betting on the same principle.

The obstacles are also clear. Quantum Computing Report lists three:

• depositing molecular film uniformly across large wafers
• keeping coupling losses low between molecules and waveguides
• showing high-fidelity entangling gates between separate qubits5

A separate trade publication adds that deposition has to keep the spin and optical properties seen in the crystal, and that no multi-qubit processor or chip-based entanglement has been shown so far.19 NVision's own technical account says the next test is whether the molecule's properties survive on fabricated chips.8

The roadmap and the error-correction gap

NVision expects its first prototype chips in the fourth quarter of 2026. It is aiming for a live on-chip demonstration by year-end, once foundry partners deliver.1 After that, the target is 1,000 physical qubits, then 1,000 logical, fault-tolerant qubits on chip by 2030, in a form that fits in a data center.1

The jump from the first target to the second is huge. A logical qubit is built from many physical qubits using error correction. Going from 1,000 physical to 1,000 logical qubits in about four years would mean either very efficient error-correcting codes or a fast increase in physical qubit count, and the public materials explain neither.3 One outlet points out that nothing has yet shown the error correction fault-tolerant logical qubits would need.3 Another says the 2030 goals depend on future progress in integration, entanglement, control and error correction.19 I think those caveats are right. Until two molecular qubits have been entangled on a chip, 2030 is a statement of intent, not a forecast.

The investor quotes in the release are, unsurprisingly, more optimistic. Backers include Playground Global, DCVC, Entrée Capital and the European Investment Bank.17 Entrée Capital's Avi Eyal says the result puts NVision among the few companies with a credible, manufacturable path to scale.9 Quantum Machines CEO Itamar Sivan says his company's control systems are helping NVision move toward working systems.2 Finimize offers a more useful benchmark. It argues investors should watch process metrics, such as whether the film can be deposited the same way every time across many chips, rather than any headline quantum-performance claim.10

The cryptography backdrop

The timing makes a pointed contrast. One day before NVision's announcement, NIST finalized IR 8547, its roadmap for moving to post-quantum cryptography.2122 Under Executive Order 14412 and OMB memo M-26-15, federal high-value and high-impact systems must adopt post-quantum key establishment by December 31, 2030, and post-quantum digital signatures by December 31, 2031.24 NIST will deprecate RSA, ECDSA, EdDSA and conventional Diffie-Hellman after 2030 and disallow them after 2035. It recommends ML-KEM-768 as the default key-establishment setting.24

The reason for the urgency is the "harvest now, decrypt later" threat: data collected today can be decrypted once a powerful enough quantum machine exists.21 NIST also notes that past cryptographic transitions have taken 10 to 20 years.

NVision's 2030 logical-qubit target is not a code-breaking forecast. A thousand logical qubits from one startup would not be a cryptographically relevant machine by itself. Still, the matching dates show the shape of the field. The policy community is planning against a deadline set by quantum progress across the whole industry, and dozens of hardware companies, now including molecular-qubit developers, are each publishing roadmaps aimed at that same window. An October GAO report found federal agencies have not adequately prepared for the transition.23 For security teams, the message from both stories is the same: do not plan around any one vendor's timeline.

The verdict

The Nature paper makes a credible case that designed molecules can be real, optically addressable qubits with millisecond coherence. That is a meaningful addition to the range of quantum hardware approaches. The HyperLight partnership gives that science a realistic manufacturing route. Everything after that, including loss tolerance on real chips, entanglement between molecules, error correction and the 2030 logical-qubit target, is still unproven. The year-end on-chip demonstration is the first real test. It will show whether molecular qubits keep their properties once they are on a wafer, and that will matter more than any number in the press release.