[Volume 38. Quandela — Europe's Photonic Quantum Computing Champion and the Seoul Bet]

Executive Summary

While exploring the quantum computing landscape, I came across Quandela, and the platform left enough of an impression that I decided to write about it.

Founded in 2017 as a CNRS spinout near Paris, Quandela builds on two decades of single-photon source research led by co-founders Pascale Senellart-Mardon, Valérian Giesz, and CEO Niccolo Somaschi.

Today, with roughly 100 employees, the company operates as a full-stack quantum player—spanning hardware, a cloud platform, its open-source SDK (Perceval), and a quantum ML environment called MerLin.

In November 2025, it also signed an MoU with the Seoul Metropolitan Government worth up to ~$57M to establish an R&D and manufacturing hub in Korea.

Why Photonic?

What drew me in first was the infrastructure argument. IBM and Google's superconducting qubits require near absolute-zero cooling, fundamentally incompatible with standard data center operations. Quandela's photons run at room temperature, fit into 19-inch racks, travel natively over commercial fiber, and consume well under 1MW. For anyone who has sold enterprise infrastructure, that pitch lands differently.

The deeper differentiator: a February 2025 paper showed Quandela's hybrid architecture reduces fault-tolerant components by 100,000x versus purely photonic competitors, from roughly one million per logical qubit down to 12. This is a theoretical result; hardware validation has not been publicly confirmed. But the IP implications are significant.

Hardware Lineup

Model	Qubits	Status
MosaiQ (Arcturus)	6	OVHcloud, 2023
Belenos	12	Launched May 2025; 1,200+ users, 30 countries
Lucy (EuroHPC)	12	Delivered CEA TGCC Oct 2025; full operation early 2026
Canopus	24	OVHcloud, mid-2026

Belenos delivers 4,000x more computing power than MosaiQ. Lucy will connect to France's Alice Recoque exascale supercomputer in 2026, opening a new class of hybrid HPC-quantum workloads.

Belenos

Quandela Cloud & Software Ecosystem

Quandela Cloud

I created an account, generated an API token, and had Perceval connected in under five minutes. The WebIDE launches in-browser with no local setup. The platform serves 1,200+ users across 30 countries, with business users now at 25% of total usage. Key capabilities: qpu: prefix for real QPU access, sim: prefix for the Exqalibur GPU emulator (up to 20 qubits, 150 billion parameter spaces, kilohertz rates), and QPU reservation for dedicated time slots.

Quandela Hub, Platforms status

Perceval: Open-Source Photonic SDK

The design principle that stood out most: identical code runs on local simulation, cloud emulator, and real QPU. I tested this directly. The notebook I ran applied Boson Sampling to HER2-Trastuzumab interface analysis, encoding the protein contact matrix (PDB: 1N8Z) into a 12-mode photonic circuit.

import perceval as pcvl

import perceval.components as comp

from scipy.linalg import expm

# Step 1: Protein contact matrix → Unitary encoding

H = adj_matrix * (np.pi / (2 * max_eigenval))

U = expm(1j * H)

u_svd, s, vh = np.linalg.svd(U)

U = u_svd @ vh # SVD-corrected unitary

# Step 2: Build photonic circuit (12 modes, 4 photons)

circuit = pcvl.Circuit(12)

circuit.add(0, comp.Unitary(pcvl.Matrix(U)))

# Step 3: Run Boson Sampling — swap "SLOS" → "qpu:belenos" for real QPU

simulator = pcvl.Simulator(pcvl.BackendFactory.get_backend("SLOS"))

simulator.set_circuit(circuit)

input_state = pcvl.BasicState([1,1,1,1,0,0,0,0,0,0,0,0])

probs = simulator.probs(input_state) # 554 output states

# Step 4: Map photon occupancy → residue importance scores

mode_occupancy = np.zeros(12)

for state, prob in probs.items():

for i in range(12):

mode_occupancy[i] += prob * state[i]

residue_importance = mode_occupancy / mode_occupancy.sum()

Switching to the real QPU is one line. The 12-qubit constraint is real; anything beyond PoC scale relies on the sim: backend. But for a biopharma domain user, this is currently the most accessible entry point into photonic quantum computing.

HER2 with Boson Sampling code

MerLin: Quantum Machine Learning Environment

Unveiled in summer 2025, MerLin integrates directly with PyTorch and scikit-learn, letting data science teams add a quantum layer without rewriting classical code. OVHcloud will host it from mid-2026. Early adopters include Imperial College London, Mila, and NYUAD's QML Lab.

Korea Strategy

On November 12, 2025, Quandela and the Seoul Metropolitan Government signed an MoU at Seoul City Hall, attended by Deputy Mayor Kim Tae-kyun, CEO Niccolo Somaschi, French Ambassador Philippe Bertoux, and Quandela Korea head Kim Yoo-seok. The ~$57M framework covers a Quantum Technology Development Center as Quandela's R&D, manufacturing, and commercialization hub for Asia.

One month earlier, Seoul had signed a similar MoU with Pasqal, co-founded by Nobel laureate Alain Aspect. Seoul is running a deliberate strategy to become Asia's quantum hub. Korea's semiconductor infrastructure at Samsung and SK Hynix is structurally compatible with silicon photonics; Korean fabs could eventually supply components for Quandela's hardware roadmap, a supply chain positioning no photonic competitor has matched.

Technology Roadmap 2024–2030

Year	Milestone	Status
2025	First logical qubit; Belenos launch; 4 QC/year capacity	Belenos; logical qubit unconfirmed
2026	Canopus on OVHcloud; MerLin on OVHcloud; Lucy fully operational	In progress
2027	Second quantum computer factory	Planned
2028	50 logical qubits; quantum networking	Planned
2030	Hundreds of logical qubits; fault-tolerant QC	Planned

Performance target: 400 QOPS today to 10,000 QOPS (25x). The February 2025 paper de-risks the fault-tolerant path on paper; hardware validation remains the critical next milestone.

Competitive Positioning

	Quandela	IBM	Google	IonQ
Modality	Photonic	Superconducting	Superconducting	Trapped Ion
Cooling	None	~-273°C	~-273°C	Room temp
Physical qubits (2025)	12	100+	70+	35
Logical qubit path	12 components (theoretical)	~1M	~1M	N/A

Direct photonic competitors: Xanadu (Canada), ORCA Computing (UK), PsiQuantum (US). The 100,000x component reduction is Quandela's core differentiation claim, pending hardware validation.

SWOT

Strengths: Only photonic company with paying customers across two continents; 100,000x IP moat; full-stack ecosystem from SDK to QPU; Seoul MoU with government co-investment; anchor partners in French government, EDF, and OVHcloud.

Weaknesses: 12 qubits vs. IBM/Google scale; logical qubit unconfirmed; small photonic developer community; 4 QC/year assembly capacity; revenue undisclosed.

Opportunities: Korea semiconductor synergy for photonic components; QKD demand surge; MerLin capturing AI engineering teams; Alice Recoque hybrid workloads in 2026.

Threats: IBM and Google capital scale; PsiQuantum and Xanadu accelerating; Seoul MoU subject to government budget timelines; logical qubit delay undermines 2030 roadmap credibility.

AI Paul's View

Having run Perceval notebooks on Quandela Hub, tested the sim: to qpu: transition firsthand, and studied the Korea MoU in detail, my read is this: Quandela has built something real. The platform works. The SDK is genuinely accessible. The hardware has paying customers. That foundation matters more than qubit count comparisons at this stage.

The Seoul MoU is the strategic signal I watch most closely. Positioning Korea as the intersection of European photonic quantum IP and the world's most advanced semiconductor manufacturing is a bet that makes structural sense. MerLin shifts the sales conversation from physicists to AI engineering teams at Samsung Research, LG AI Research, and SK Telecom, a fundamentally faster sales cycle. The 2025 logical qubit target remains unconfirmed and is the single most important near-term proof point to watch.

Risks are real: 4 QC/year capacity, undisclosed revenue, unconfirmed logical qubit, and government co-investment timelines. But the France-Korea quantum partnership that began taking shape in late 2025 has genuine momentum on both sides. France brings deep quantum IP and manufacturing expertise; Korea brings semiconductor infrastructure, R&D investment capacity, and a clear government commitment to becoming Asia's quantum hub. This collaboration is still in its early chapters, and I expect it to deepen and expand in ways that will matter well beyond the quantum computing sector.

This research is based on publicly available information and represents personal analysis only. Not investment advice.