Qluster, a quantum computing startup in Paris, announced on November 17, 2026 that it has raised €4.7 million (about ¥820 million at €1 = ¥175.05) in a pre-seed round led by Quantonation, an investment firm specializing in quantum technologies. The company aims to develop a mid-scale quantum computer by 2032 and a fault-tolerant, general-purpose quantum computer by 2036. Its foundation is a light-source technology the founders have researched at the Laboratoire Kastler Brossel (LKB) in Paris. The company uses a "continuous-variable" approach, which encodes information in continuous physical quantities of light, such as amplitude and phase. However, Xanadu of Toronto, Canada, which leads in the same approach, acknowledges in its paper on the Aurora machine that reducing optical loss to the level fault tolerance requires remains a major challenge. Qluster is thus entering a field with significant technical hurdles, with practical use targeted for 2036.

AD

€4.7 million to set up its own lab, with thousands of logical qubits targeted by 2036

According to Quantonation's announcement, this is Qluster's first funding round, with Quantum Launchpad and Bpifrance also participating. The funds will mainly go toward establishing the company's own research and development facility and expanding its team. The company says it will hire experimental physicists and experts in quantum computing theory and architecture. Qluster was founded only a few months ago by four people. CEO David Fainsin and CTO Ambroise Boyer both hold PhDs in quantum optics, COO Clarisse Thibault comes from CNRS Innovation, and CSO Valentina Parigi is a professor at Sorbonne University.

The development plan has two stages. By 2032, the company plans to build a mid-scale quantum computer capable of solving initial industrial problems. By 2036, it aims to evolve this into a large-scale, fault-tolerant, general-purpose quantum computer with thousands of logical qubits that can run billions of operations. Fault tolerance is the ability to keep computing correctly through error correction even when physical errors occur during computation. A logical qubit is a qubit protected by error correction across multiple physical qubits and treated as a single one.

Qluster calls this development approach "Architecture First." The idea is to translate an industrial problem into a quantum circuit, estimate the required number of qubits, circuit depth, execution speed, and energy consumption, and then design the whole machine accordingly. In the November 17 announcement, Fainsin pointed out that the recent race to build quantum computers has focused too heavily on improving individual qubits, and said the current challenge is to build a computing architecture that can meet industry demand.

How computing with quantum states of light works, and the founders' research on more than 21 modes

The continuous-variable approach Qluster uses encodes quantum information in continuous physical quantities of light's electromagnetic field. It generates "cluster states," in which many optical modes are quantum-mechanically entangled, and uses them as a computational resource. The company explains that because many modes can be handled at room temperature, the approach is easier to scale and to keep stable in operation.

Three technologies are key to understanding this approach: squeezed light, cluster states, and frequency multiplexing. Squeezed light is light in which one of the quantum fluctuations corresponding to amplitude and phase is suppressed below the ordinary quantum noise level. A cluster state is one in which many such optical modes are entangled; computation proceeds by measuring each mode in sequence and adjusting later measurements according to the results. Frequency multiplexing uses different frequency components within a single beam of light as independent modes. Because it increases the number of modes without simply adding more light sources, it is one way to generate large cluster states.

In 2022, Parigi's research group reported simultaneously generating squeezed light with 21 spectral modes at a repetition rate of 156 MHz. The team described this as the first combination of frequency multiplexing and time multiplexing of squeezed light. In 2024, a paper co-authored by Fainsin and Parigi was published in Physical Review Research (6, 043113). That study used nonlinear waveguides, channels that guide light, to generate multimode squeezed light in the 1550 nm band also used in fiber-optic communications. It observed squeezing in more than 21 frequency modes, suppressed quantum fluctuations by up to more than 2.5 dB, and also succeeded in generating cluster states with a small number of nodes.

According to Qluster, its core technology is a compact, patented light source developed by Parigi and Fainsin at LKB. The company says its principle has been demonstrated experimentally and published in a paper, but the announcement does not identify which paper. In an interview with Sorbonne University, Fainsin explained that during his doctoral research, he used a nonlinear crystal integrated into a photonic chip to generate a new kind of quantum state of light, which led to the patent application. LKB, the base for this research, has also produced three Nobel laureates in physics: Alfred Kastler in 1966, Claude Cohen-Tannoudji in 1997, and Serge Haroche in 2012.

AD

From cluster-state generation in 2019 to Qluster's founding in 2026: a timeline

Continuous-variable photonic quantum computing has built up important results, including the generation of two-dimensional cluster states in 2019 and, in 2025, Xanadu's modular machine and on-chip GKP state generation. In 2026, following the NTT–OptQC partnership, Qluster announced its funding. These are selected milestones, however, and the studies are not directly linked in a single technical lineage. Generation of a one-dimensional cluster state with one million modes had also been reported before 2019.

In 2019, two papers in Science demonstrated the deterministic generation of two-dimensional cluster states: work by Asavanant et al., including the Furusawa laboratory at the University of Tokyo (Science 366, 373), and work by Larsen et al., including Andersen of Denmark (Science 366, 369). Photonic quantum computers also include a discrete-variable approach that uses single photons as qubits, being developed by companies such as Quandela in France and PsiQuantum in the US. These two studies, by contrast, showed that the continuous-variable approach can also experimentally generate the two-dimensional cluster states that serve as computational resources. Later, in 2022 and 2024, Parigi's group at LKB published results on multimode squeezed light using frequency multiplexing.

In 2025, Xanadu published two papers in Nature. The Aurora paper, published on January 22, reported a machine housing 84 squeezers (devices that generate squeezed light) in four server racks. The device acquired data continuously for two hours at 1 MHz and synthesized and measured a cluster state of 86.4 billion modes. On June 4, Xanadu also reported generating GKP states in the 1550 nm band on a silicon nitride chip fabricated from a 300 mm wafer. A GKP state is a quantum state that encodes a qubit in a single optical mode and has a structure suited to error correction.

In Japan, NTT and OptQC announced a capital and business alliance on August 3, 2026. OptQC is a company built on 25 years of photonic quantum computing research at the University of Tokyo. The two companies aim to develop a fault-tolerant photonic quantum computer on the scale of one million qubits by fiscal 2030. Qluster's funding announcement came about three and a half months later.

Company Published results Stated goal Funding in this announcement
Xanadu Aurora machine (Nature, January 2025); on-chip GKP state generation (Nature, June 2025) Not compared Not compared
OptQC Not compared With NTT, develop a fault-tolerant, million-qubit-class computer by fiscal 2030 Not compared
Qluster Not compared Mid-scale machine in 2032; fault-tolerant general-purpose machine in 2036 €4.7 million pre-seed

Note: "Not compared" indicates items not covered in this table and does not mean that results or plans do not exist. All development goals are plans announced by each company.

The three companies are at different stages. Xanadu has published research results using actual hardware for Aurora and its GKP state generation chip, and OptQC is aiming for the million-qubit class by fiscal 2030 through its partnership with NTT. Qluster, by contrast, is at the stage of using the funds it has just raised to set up its own lab, building on its founders' light-source research. It has not yet published any demonstration as a quantum computer, and its fault-tolerant target of 2036 is later than the fiscal 2030 goal set by NTT and OptQC.

Aurora's optical loss of over 95% at most, and the gap to what fault tolerance requires

Xanadu's Aurora paper specifically shows the gap between the current device and the performance needed for fault tolerance. According to the paper's discussion, the best architecture currently known requires an optical loss budget of about 1% and about 10 levels of multiplexing to achieve fault tolerance. By contrast, the optical loss measured in Aurora was about 56% on the heralding path and just over 95% on the heralded path. The loss budget here means the upper limit on the fraction of light that can be lost as it passes through each component while remaining tolerable.

The two kinds of paths, heralding and heralded, exist because Aurora generates the special quantum states needed for computation probabilistically. Some modes of multiple squeezed-light sources are measured with photon-number-resolving detectors, and when a particular photon number is detected, the desired quantum state is judged to have been generated in the remaining modes. A photon-number-resolving detector can tell how many photons arrived. The measurement-side path that signals state generation is the heralding path, and the path whose state is confirmed and used in computation is the heralded path. Xanadu's paper on GKP state generation likewise explains that reducing optical loss can improve effective squeezing performance.

The Aurora paper also spells out the conditions behind the roughly 1% tolerable loss. It is not a strict upper limit but a value said to be achievable with a particular circuit configuration. On the hardware side, the paper estimates that if the insertion loss of each component could be improved by a factor of 20 to 30 in decibel terms relative to state-of-the-art optical components, fault-tolerant operation could be achieved without relaxing the architecture's requirements. The June GKP paper from the same year also notes that further reduction of optical loss is needed for fault tolerance.

Room-temperature operation, often cited as an advantage of photonic quantum computers, also deserves caution. In Aurora, the photon-number-resolving detectors are housed in a cryostat, a cryogenic container, while the other modules run at room temperature. The roughly 1% loss budget is also a value based on the architecture Xanadu examined, and whether Qluster would need the same level depends on the configuration it adopts.

AD

From the new lab to a mid-scale machine in 2032: the remaining technical challenges

Qluster's Fainsin also acknowledges that adopting the continuous-variable approach does not solve all technical problems. In the November 17 announcement, he said the company's approach does not eliminate the remaining scientific and technical challenges to building a reliable quantum computer.

Quantonation, which led the investment, rates the approach's prospects highly. Ylan Tran, who handles the investment, said scalability is the central challenge of quantum computing, and expressed the view that continuous-variable architectures using nonlinear optics could lead to quantum computers with millions of qubits operating at room temperature. This, however, is an investor's technical outlook. In Aurora, the photon-number-resolving detectors require cryogenic cooling, and Qluster has not disclosed in this announcement which detectors or quantum error-correction schemes it will use.

The first thing Qluster will do with the funds is set up its own lab and secure researchers for both experiment and theory. After that, it will need to raise the multimode squeezed light that its founders demonstrated in papers to a level of performance usable as a component of a quantum computer. Xanadu's Aurora paper likewise estimates that approaching the level needed for fault tolerance will require improvements in both architecture and hardware.

In assessing Qluster's progress, key factors will be the squeezing performance and optical loss of the light source developed in the new lab, the detectors used, and the concrete configuration of the error-correction scheme. Which industrial problems the 2032 mid-scale machine will target, and how much computational performance it will achieve, will also be telling. Beyond that, the company plans to develop by 2036 a fault-tolerant, general-purpose quantum computer that can run billions of operations using thousands of logical qubits and tackle complex industrial problems.