Bringing photonic quantum computing from the lab to the data centre
Images by Quandela
French quantum computing firm Quandela and Taiwanese semiconductor equipment supplier Hermes-Epitek have signed a memorandum of understanding to jointly develop hardware for light-based quantum computers, marking a new step in international photonics cooperation.
Under the agreement, signed on 12 November at Taiwan’s Hsinchu Science Park, the two companies will combine complementary technologies required to build an optical quantum system. Paris-based Quandela develops sources that emit single photons alongside the circuits used to guide them, while Hermes-Epitek manufactures photonic components and detectors sensitive enough to register individual particles of light.
Initially, the partnership will focus on building an experimental platform in Taiwan. Engineers will link the detectors using optical fibre before investigating whether they can be integrated more closely onto the circuits themselves. Any move to install a full-scale quantum computer will require a separate accord detailing funding, specifications, a host site and a timetable. For now, researchers are tackling a more fundamental engineering hurdle: measuring how much light actually survives the journey through the hardware.
How photonic quantum computing works
Spun out of research at France’s National Centre for Scientific Research (CNRS) and the Centre for Nanoscience and Nanotechnology, Quandela was founded in 2017 by Pascale Senellart, Niccolò Somaschi and Valérian Giesz. The company’s technology relies on quantum dots: these are microscopic regions of semiconductor material that emit a single photon when stimulated by a pulse of light. For the system to function properly, photons generated by successive pulses must share nearly identical properties so they can interfere with one another as intended.
In this architecture, a quantum bit, or qubit, is encoded in the two possible paths a photon might take through an optical circuit. When those paths converge, quantum interference alters the probability of the photon emerging at a given output. By repeating the process thousands of times, researchers gradually build a statistical distribution of results. However, as calculations grow more complex and demand more photons per run, cumulative signal loss across the system makes complete data increasingly difficult to capture.
Much of that light is lost at the joints between separate components. Tighter hardware integration could curb those losses, provided the detectors and circuits operate harmoniously. Hermes-Epitek’s detectors rely on a supercooled nanowire; when struck by a photon, the wire’s superconductivity is momentarily disrupted, generating a detectable electrical pulse. The planned fibre-coupled platform will give engineers a vital baseline against which to measure future, more tightly integrated designs.
From working devices to repeatable production
In May, co-founder and chief executive Niccolò Somaschi noted that “execution matters as much as breakthrough science”. To back that up, the company appointed Cyril Dujardin as chief operating officer and brought in Michel Zecri to oversee manufacturing, quality and customer support. Its roadmap to 2033 outlines a hybrid production model in which partner foundries fabricate the optical circuits, leaving Quandela to assemble the final systems in-house.
The strategy also envisages linking smaller processors across optical networks. Researchers have already demonstrated one key building block in the laboratory, using the spin of a particle trapped inside a quantum dot to generate a stream of photons with linked quantum states. Such patterns are designed to help future machines shield information from errors, though preserving those delicate states as photons travel between separate modules remains a major engineering hurdle.
Lucy brings photonic quantum computing to a supercomputing centre
Quandela delivered its first MosaiQ system to the French cloud provider OVHcloud in 2023. Two years later, in October 2025, Quandela and the German cryogenics specialist attocube delivered Lucy, a 12-qubit photonic machine procured by the European High Performance Computing Joint Undertaking. Formally inaugurated in April at the French Alternative Energies and Atomic Energy Commission’s computing centre south of Paris, Lucy sits alongside the Joliot-Curie supercomputer at Bruyères-le-Châtel.
On 5 November, a team from Oslo Metropolitan University announced that a European pilot programme had awarded it 15 hours of compute time on Lucy to search for signs of quantum chaos in dissipative systems; these are environments that lose energy to their surroundings. Having previously worked with superconducting machines, the researchers will use the allocation to test how their experiment can be programmed onto a photonic architecture and which hardware features they can exploit.
OVHcloud has also added cloud access to Quandela’s 12-qubit Belenos system. While Quandela lists a power consumption of eight kilowatts for the core Belenos unit, experts note that any meaningful comparison of energy efficiency would require measuring the power drawn by the entire installation to complete a comparable task.
Fitting quantum processors into existing workflows
Quandela is actively pursuing further international installations. Last month, the Vietnamese company Quanova selected a MosaiQ system for a research and education programme at Quang Trung University in Quy Nhơn, backed by Nam A Bank; a deployment Quandela says would mark Vietnam’s first quantum computer. Closer to home, the firm was one of five awarded framework agreements in 2024 under PROQCIMA, a French defence procurement initiative aimed at developing large-scale quantum systems.
The company is also exploring how photonic processors might slot into existing supercomputing workflows, handling specialist steps within larger jobs run on conventional hardware. In June, Quandela reported testing a low-latency link between a physical photonic processor and Nvidia’s accelerated computing infrastructure. That was followed in October by a demonstration in Berlin showing how an application could exchange data with its system controller using Nvidia’s NVQLink and a simulated quantum processor. Meanwhile, its MerLin software allows developers to experiment with hybrid quantum and machine-learning models, though proving a definitive gain in speed or accuracy on an industrial workload will require separate benchmarking.
Ultimately, experiments such as OsloMet’s offer a concrete test of what Quandela’s industrial strategy can deliver. Over those 15 hours, researchers must translate a complex physics question into an optical circuit, run it enough times to gather usable data and interpret the results. While the work in Hsinchu and the Nvidia integration tackle different pieces of that puzzle, the real commercial test will be whether such scarce, specialist compute time can become an ordinary, routine part of scientific research.
Further reading on MoveTheNeedle.news:
Q-CTRL raises US$290 million to make quantum computing useful
A different future for quantum computing