On November 12, 2026, at Hermes-Epitek’s premises in Hsinchu Science Park, Dr. Valérian Giesz, co-founder and chief scientific and product officer of Quandela, and Dr. David Chen, president of Hermes-Epitek Corporation, signed a memorandum of understanding to develop and industrialize photonic quantum technologies. The announcement, published by Hermes-Epitek the same day, pairs two companies that occupy opposite ends of a photonic quantum system: Quandela builds the sources that produce single photons, and Hermes-Epitek builds the detectors that count them.
Read the announcement carefully and it says something narrower than the headline suggests. This is a memorandum of understanding, not a contract, an order, or a product. Its first phase is a proposed joint research and development programme. Its second phase — the installation and deployment of a photonic quantum computer in Taiwan — is explicitly subject to a separate agreement that has not yet been negotiated. The system specifications, the installation site, the funding, and the timetable for that phase all remain subject to mutual agreement, which is to say undecided. Nothing in the announcement quotes a performance figure, a qubit count, a delivery date, or a sum of money, and neither should any coverage of it.
That narrowness is not a criticism. It is the point. An MoU between a source company and a detector company is a statement about where the bottleneck in photonic quantum computing actually sits, and the bottleneck is not in any single component. It is in getting the components to work together, reliably, at manufacturing scale, inside one system.
The two halves of a photonic machine
A photonic quantum computer is, in outline, three subsystems: sources that produce single photons on demand, circuits that route and interfere them, and detectors that register the result. Quandela supplies the first two. Its technology is built on semiconductor quantum dots — the “artificial atoms” that emit one photon at a time when excited — embedded in micropillar cavities that use the Purcell effect to accelerate emission and direct it into a usable mode. The company’s Prometheus source packages that dot-and-cavity device into a standalone, fiber-coupled unit, and its MOSAIQ platform integrates the sources with programmable photonic integrated circuits, readout devices, and the Perceval software stack into a full-stack photonic quantum computer. The lineage runs back to more than 25 years of research at the Centre for Nanoscience and Nanotechnology (C2N), the CNRS and Université Paris-Saclay laboratory where co-founder Pascale Senellart’s group developed the in-situ fabrication technique that places a dot deterministically at the centre of its cavity. Quandela, founded in 2017 and headquartered in Massy, France, is the spinout that turned that laboratory technology into an industrial product, and it has been developing its photonic quantum computer since 2020.
Hermes-Epitek supplies the third subsystem, and its background is the less glamorous half of the quantum supply chain. Founded in 1977, the company is a provider of semiconductor equipment and technologies — ICP etchers, ion beam implanters, MOCVD systems, silicon carbide substrate manufacturing — with a workforce of more than 1,200 people and operations across Taiwan, the United States, Japan, Singapore, and Malaysia. Its quantum business, built on that equipment and process base, covers superconducting and semiconductor epitaxial thin films, single-photon detectors based on epitaxial superconductor–semiconductor–metal heterostructures, quantum foundry services, cryogenic systems, and photonic component manufacturing.
The detector technology at the centre of the MoU is the superconducting nanowire single-photon detector, or SNSPD. An SNSPD is a meandering strip of superconducting material, cooled well below its critical temperature and biased just below its critical current. When a single photon is absorbed, it breaks Cooper pairs locally, creates a resistive hotspot, and the bias current is diverted into a readout circuit, producing a voltage pulse. The device then recovers as the hotspot cools. The result is a detector with high detection efficiency, very low dark-count rates, and timing jitter measured in tens of picoseconds — the performance that makes SNSPDs the standard readout for photonic quantum computing and quantum communication experiments. The price is cryogenics: the nanowire must sit at a few kelvin, which is why the announcement lists Hermes-Epitek’s cryogenic capabilities alongside its detectors rather than as an afterthought.
The pairing, then, is complementary in a way that goes beyond a press-release symmetry. Quandela’s sources are also cryogenic devices — the quantum-dot emission that makes them bright and indistinguishable requires cooling to a few kelvin — so a system built from both companies’ hardware shares a single thermal environment rather than two. The source produces the photon, the circuit routes it, and the detector, sitting in the same cryostat, registers it. The announcement’s framing of the joint programme as combining “Quandela’s quantum computing architecture and system integration expertise with Hermes-Epitek’s capabilities in detector technology, semiconductor processing, and manufacturing” is, for once, a reasonably literal description of what each side brings.
Why fiber coupling comes first
The programme’s sequencing is the most informative part of the announcement. The two companies plan to begin by testing a joint experimental platform in Taiwan using fiber-coupled detectors, before exploring closer integration of detectors and photonic circuits.
Fiber coupling is the conservative first step, and that is its virtue. A fiber-coupled detector is a standalone device: the SNSPD sits in its own cryostat, the photon arrives through a standard optical fiber, and the interface between the two companies’ hardware is a connector. Neither side’s hardware has to change to accommodate the other. The test platform can measure what matters — whether the detector reads the source’s photons with the efficiency and timing resolution the system needs — without committing either side to a redesign.
The second step, closer integration of detectors and photonic circuits, is where the real difficulty lives. Integrating an SNSPD onto a photonic chip means fabricating a superconducting nanowire directly on the waveguide, so the photon never leaves the chip. That eliminates the fiber coupling loss and the packaging complexity of getting light from a chip into a fiber and back into a cryostat, and it is the architecture a scalable photonic processor would eventually want. But it also means the detector’s fabrication — superconducting films, nanometer-scale patterning, tight thermal budgets — must coexist with the photonic circuit’s fabrication on the same substrate, and a failure in either process step ruins both. Several research groups have demonstrated on-chip SNSPDs, but moving from a laboratory demonstration to a manufactured, reproducible component is exactly the kind of step this MoU is about, and the announcement is careful to describe it as something to be explored after the fiber-coupled platform, not delivered alongside it.
The same discipline applies to the programme’s stated goal of assessing the combined technologies’ “suitability for future photonic quantum systems,” subject to successful validation. Validation here is not a formality. A source and a detector that each perform well in isolation can fail together in ways neither specification sheet predicts: the detector’s dark counts can swamp the source’s heralded events, the cryostat’s vibration can shift the fiber alignment, the detector’s recovery time can limit the system’s clock rate. The joint platform is the instrument for finding out.
Industrialization, in the Taiwanese sense
The word the announcement uses for its goal — industrialize — deserves attention, because it means something specific in Taiwan’s context. Taiwan’s semiconductor ecosystem is not just a collection of fabs; it is a manufacturing culture built around yield, reproducibility, and process control at volumes no laboratory can match. Hermes-Epitek’s contribution to this partnership is not only its detectors but its position inside that ecosystem: decades of equipment and process expertise, epitaxial thin-film capability, and foundry services that know how to turn a device design into a repeatable process.
That is the capability photonic quantum computing most conspicuously lacks. The field’s sources and detectors are still largely hand-built, characterized one device at a time, and tuned by the researchers who made them. Quandela’s Prometheus source is a deliberate step away from that — a packaged, fiber-coupled, ready-to-use device — but a packaged source is not yet a manufactured system, and a system that needs dozens or hundreds of sources and detectors, all performing to specification, is a manufacturing problem before it is a physics problem. The MoU’s bet is that Taiwan’s process discipline can be applied to photonic quantum hardware the way it was applied to classical photonics and semiconductors.
It is worth noting that this is not Quandela’s only move in Asia. The company has been pursuing a parallel strategy in South Korea, where chief executive Niccolo Somaschi said at a Seoul press conference that Quandela intends to identify partners within two years for joint development spanning production, integration, and real-world operation, naming Samsung, SK hynix, and LG Innotek as potential collaborators alongside Korea’s three quantum clusters, and where a partnership with KAIST has established a Quandela Hub for education and industry–academia collaboration. The Taiwan MoU and the Korea strategy are the same industrialization logic applied to the two semiconductor ecosystems most capable of delivering it. Neither has yet produced a manufactured product, and both should be read as options being opened, not outcomes being promised.
Taiwan, for its part, has been building momentum around quantum technology more broadly. The week before this MoU, Taiwan and the Netherlands announced four joint photonics research and development partnerships under their bilateral Technology Bridge Program, targeting photonic packaging, 5G and 6G communications, AI data centre connectivity, and quantum computing. The Quandela–Hermes-Epitek agreement lands in that current, not in isolation — one more piece of an ecosystem that is assembling itself around photonic hardware.
What is still undecided
The honest way to cover an MoU is to list what it does not decide, because that list is the story’s actual content. The announcement is explicit about four open items for the deployment phase: system specifications, installation site, funding, and timetable. Each of those is a substantial negotiation on its own.
System specifications means, among other things, how many photons the machine would work with, what computation it would perform, and what performance it would be judged against. A photonic quantum computer installed in Taiwan could be a small research system for algorithm development or a larger machine aimed at industrial use cases, and those two machines differ by orders of magnitude in source count, detector count, and cryogenic load. Nothing in the announcement narrows that range.
Installation site matters because a photonic quantum computer is not a server that plugs into a rack. It needs cryogenic infrastructure, vibration isolation, clean power, and technical staff — and the site determines who operates it, who has access to it, and what it is used for. Funding matters because no figure has been attached to any phase, and a joint R&D programme and a machine installation are different kinds of expenditure, likely involving different funders. Timetable matters because the deployment phase is contingent on the R&D phase: the announcement says the companies will assess their technologies’ suitability “subject to successful validation,” which makes the deployment schedule downstream of results that do not yet exist.
There is also a structural fact about MoUs that should be stated plainly. A memorandum of understanding is not binding in the way a contract is. It records an intention to collaborate and a framework for doing so. Companies sign them to create a working relationship, to justify spending engineering time together, and to signal to funders and customers that a partnership exists. They also let them walk away. The correct reading of today’s announcement is that Quandela and Hermes-Epitek have agreed to spend time and effort finding out whether their technologies work well together, and have agreed to talk about a machine if the answer is yes. That is a meaningful commitment of engineering effort. It is not a machine.
What to watch next
The first thing to watch is the most concrete: the joint experimental platform in Taiwan. When it is built, the interesting numbers will be the ones the announcement does not quote — the end-to-end detection efficiency of Quandela’s photons through Hermes-Epitek’s fiber-coupled detectors, the dark-count rate at operating temperature, and the timing jitter of the combined system. Those measurements, if published, will say more about the partnership’s prospects than the MoU itself.
The second is the integration question. If the fiber-coupled platform validates, the next public signal would be work on detector–circuit integration: on-chip SNSPDs, co-packaged cryogenic assemblies, or foundry processes that produce both. That is the step where Hermes-Epitek’s manufacturing expertise either proves decisive or proves insufficient, and it is the step that will take the longest.
The third is the separate agreement for a machine in Taiwan. Its negotiation will be the real test of whether this partnership is a research collaboration or an industrial programme. A signed deployment agreement with a named site, a funding source, and a specification would be news of a different order from today’s MoU. Until then, the honest summary is narrower — and also more interesting than a routine partnership announcement. Two companies that each own one end of a photonic quantum system have agreed to find out, together and in Taiwan, whether the ends can be joined into something manufacturable. The physics says it is possible. The manufacturing says it is not yet done. The MoU is the price of finding out which of those two statements matters more.
Sources and further reading
- Hermes-Epitek Corporation, “Quandela and Hermes-Epitek Join Forces to Advance Photonic Quantum Technologies,” November 12, 2026. https://www.hermes.com.tw/en/quantum-pics-2026/ Primary source for the MoU, the signatories, the joint programme, the fiber-coupled platform, the planned deployment agreement, and all company background figures in this article.
- Quantum Zeitgeist, “Quandela, Hermes-Epitek To Advance Photonic Quantum Technologies,” November 12, 2026. Trade coverage; confirms the MoU terms and sequencing.
- Quandela, “The Power of Single Photon Sources.” https://www.quandela.com/technology/the-power-of-single-photon-sources/ Source for the Prometheus source, the quantum-dot and micropillar-cavity fabrication process, and the Purcell-effect design.
- Quandela, Sepocq project page. https://www.quandela.com/scientific-collaborations/sepocq/ Source for the MOSAIQ platform, the Perceval software stack, and the C2N research lineage.
- C2N, Université Paris-Saclay, “How is a quantum computer built? A report from inside C2N with Pascale Senellart,” October 29, 2026. Source for the in-situ fabrication technique and Quandela’s development of a photonic quantum computer since 2020.
- Quantum Zeitgeist, “Quandela Eyes Korean Firms For Quantum Chip Production.” Source for the Korea partnership strategy, the two-year partner-identification plan, and the KAIST Quandela Hub.
- BTQ Technologies Corp., “BTQ Technologies Signs MOU With Quandela To Advance Quantum Proof-Of-Work Protocols,” PR Newswire. Source for Quandela’s founding in 2017 and its full-stack photonic quantum computing platform.
- Digitimes, “Taiwan-Netherlands launch four photonics R&D partnerships,” November 5, 2026. Source for the Taiwan–Netherlands Technology Bridge Program partnerships.
- Note on verification: all MoU terms, signatory names and titles, programme sequencing, and company background figures in this article are taken from the Hermes-Epitek announcement of November 12, 2026. No performance figure, qubit count, delivery date, funding amount, or system specification is claimed for the planned deployment, because none has been published. The Korea strategy details are taken from the Quantum Zeitgeist coverage cited above. General descriptions of SNSPD operation and quantum-dot source physics are standard device physics, not claims about either company’s products.