Quandela

Phase A — Understand the business

Lens 1 · Company Overview

Quandela is a photonic quantum computing company — registry-bucketed under optical-computing but it builds quantum computers, not optical interconnects or photonic AI accelerators (it is closer to PsiQuantum than to Lightmatter/Ayar Labs in that same bucket). Founded 2017, spun out of the CNRS / Centre for Nanosciences and Nanotechnologies (C2N) at Université Paris-Saclay; HQ in Massy / Paris, with subsidiaries in Munich (Germany), Canada, and planned Korea.

The business model is unusual for quantum and is the single most important fact about this company: Quandela sells and ships physical quantum computers to customers' own data centers. In 2023 it became "the first European player to sell and deploy a quantum computer to a private client" (OVHcloud) — most QC peers (IonQ, Rigetti, IBM, Google) sell cloud access to machines they keep in-house. Three revenue motions:

1. On-prem machine sales — full quantum computers delivered to supercomputing centers and clouds (CEA/EuroHPC, OVHcloud, Exaion/EDF). The €8.5M Lucy contract is the anchor ASP datapoint (Lens 5).
2. Cloud access — the MosaiQ platform; 1,200+ users across 30 countries on the Belenos 12-qubit system.
3. Components — standalone Prometheus single-photon source (the underlying tech sold as a product), and MerLin, a quantum-machine-learning software platform (launched Feb 2026).

Product ladder (physical qubits): MosaiQ-2 (2022) → MosaiQ-6 → Belenos / MosaiQ-12 (12q, May 2025), billed as "the world's most powerful in its category, ~4,000× the prior 6-qubit generation" → Canopus (24q, 2026) → MerLin QML layer (claims 18 photonic QML models, 20,000× acceleration via NVIDIA CUDA-Q). "Lucy" is the productized MosaiQ-12 delivered to EuroHPC.

Lens 2 · Supply Chain

Photonic-QC supply chain, upstream → company → end customer, named where sourceable:

• Upstream — III-V semiconductor / quantum-dot epitaxy. The core IP is InAs/GaAs quantum-dot single-photon sources. Quandela has vertically integrated the most chokepointed input: in spring 2024 it inaugurated its own cleanroom for single-photon-source production and a semiconductor pilot line. This is the deliberate moat — it does not buy its photon sources on a merchant market (there isn't one).
• Photonic integrated circuits (PICs) / optics. Linear-optical processing chips, waveguides, and the cryogenic enclosure. Partner attocube (German) is named as the co-vendor on the EuroHPC Lucy machine — cryo/nano-positioning hardware. Detectors (superconducting nanowire SPDs) are a likely external input — not disclosed.
• Assembly. A dedicated quantum-computer assembly & service facility in Massy (opened June 2023), described as "the first of its kind in Europe". Target throughput cited as ~4 machines/year from 2025.
• Downstream — buyers: CEA/GENCI/EuroHPC (sovereign HPC), OVHcloud (sovereign cloud), Exaion (EDF subsidiary) in Canada, 1,200+ cloud users. Heavy government / public-procurement concentration (see Lens 13).

Chokepoint read: the single-photon source is the whole-industry chokepoint, and Quandela owns it — that is the bull case in one line. The single-source risk runs the other way (customer concentration on EU state procurement), not on inputs.

Lens 3 · Competitive Advantages (moats)

• Deterministic single-photon sources — the real moat. Quandela's quantum-dot sources are deterministic / on-demand ("a stable stream of highly indistinguishable photons" at GHz clock rates), versus the probabilistic photon generation (SPDC heralding) used by most photonic peers. Deterministic generation is the hardest unsolved problem in photonic QC — "generating individual photons on demand is hard; most sources are probabilistic". Founder Pascale Senellart's 2008 C2N breakthrough (position-controlled, high-purity sources) is the scientific bedrock — a ~15-year, CNRS-deep IP head start that is genuinely hard to replicate.
• Architecture — hybrid spin-photon (SPOQC). Quandela's fault-tolerant path merges linear-optical photons with semiconductor spin qubits to act as a quantum memory / reduce component counts and enable error correction over non-local optical links. The headline claim: a "100,000× reduction in the number of components needed for fault-tolerant calculations". Photons that don't need to be physically adjacent to interact is a structural advantage if the memory works.
• Vertical integration (own cleanroom + assembly line) → switching costs once a machine is on a customer's floor + a real manufacturing capability European rivals lack.
• Sovereignty moat (policy, not physics). As a French/EU-sourced machine ("80% European-sourced components" on Lucy), Quandela is the structural beneficiary of EU "European preference" procurement and France 2030 (Lens 8). This is a real, durable, non-technical moat in its home market.

Bargaining power: weak-to-moderate. It needs the buyers (state HPC centers, OVHcloud) more than they need it — there are substitute QC vendors (Pasqal neutral-atom, Alice & Bob/IQM superconducting) competing for the same sovereign euros. Its leverage is being the photonic national champion in a multi-modality national strategy.

Lens 4 · Segments

No audited segment data exists (private, header-only segments.csv). Qualitative split ``:

• Hardware (machine sales + components): the majority of recognized revenue — the €8.5M-class on-prem deals and Prometheus sources. n/a — not disclosed by segment.
• Cloud/Software (MosaiQ access + MerLin QML): early, recurring, low-base; 1,200+ users but pricing undisclosed.
• Geography: France/EU-dominant (CEA, OVHcloud, EuroHPC), with North America (Exaion/Canada) and an Asian (Korea) expansion vector. Government / publicly-funded buyers dominate — the segment risk is customer type (state procurement) more than product line.

---

Phase B — Measure performance

Lens 5 · Earnings Result (private overlay → Traction & unit economics)

No P&L, no earnings print. What is sourceable:

• ⚠ Provenance correction (load-bearing): the figure circulating as "Quandela 2025 revenue €31M, 18 quantum computers delivered" is IQM's, not Quandela's — a direct follow-up search returned: "this information appears to refer to IQM Quantum Computers, not Quandela". Do not attribute it to Quandela. It is excluded from this dossier.
• Revenue (Quandela, estimates only, conflicting): third-party aggregators diverge wildly — getlatka cites ~$23.8M ARR; prospeo/others cite ~$5.5M annual revenue. Both unaudited, neither corroborated by a primary source — treat as n/a — not reliably sourced, range $5–24M. The honest read: low-single-to-low-double-digit $M, pre-scale.
• ASP anchor (the one hard commercial number): Lucy's total acquisition cost = €8.5M, co-funded 50% EuroHPC JU / 50% France. That is a real, contracted price for one 12-qubit machine — the single most useful unit-economics datapoint on the company.
• Throughput: stated capacity ~4 machines/year from 2025. At €8.5M/machine that frames a single-digit €M to ~€30M+ hardware run-rate ceiling at full utilization.
• Cash burn: undisclosed; ~144–152 employees against ~$70–75M lifetime raised implies the Series B (Nov 2023) is being consumed and a raise is plausibly needed within ~12–24 months (Lens 11).

Lens 6 · Earnings Calls (private overlay → founder interviews / strategy signal)

No earnings calls. Management's stated focus, from 2026 communications:

• From "qubit-count theater" to fault tolerance + utility. Quandela's own "Four 2026 trends" piece and DARPA framing both center logical qubits and cost-vs-value, not raw physical-qubit headlines. The DARPA QBI brief is explicitly to "detail a concept for a quantum computer where computational value exceeds its cost".
• Industrialization / scale-up is the new narrative. 2026 governance hires (Lens 9) and an inaugurated pilot line signal a deliberate pivot from "lab champion" to "manufacturer." Recurring phrases: sovereignty, fault-tolerant, deterministic, hybrid spin-photon, scalable.
• Tone: confident, roadmap-anchored (2024–2030 → restated 2026–2033), increasingly commercial. The thing they've stopped leading with: pure physical-qubit one-upmanship.

Lens 7 · Comps (private overlay → cap table & photonic-QC peer marks)

Photonic / fault-tolerant-QC peer set. Multiples are meaningless pre-revenue → compared on capital, valuation, and stage:

The comp table is the whole bear/bull tension in one frame: Quandela is capitalized at ~1% of PsiQuantum and ~2% of Xanadu's implied value while claiming a better photon-source technology. Either it is dramatically undervalued, or the market has correctly priced that capital — not cleverness — wins a race that needs ~$1B+ to reach utility scale. Syndicate quality (IPO-proximity tell): Bpifrance, Quantonation, Omnes, Serena, Crédit Mutuel Innovation, EIC Fund — strong sovereign/strategic European backing but NO tier-1 US crossover (no Fidelity / T. Rowe / Coatue / BlackRock). That absence is itself a signal: Quandela is a European-capital story, and a near-term US-style mega-round or IPO is not visible.

Lens 8 · Stock-Price Catalysts (private overlay → funding & validation events)

No stock. Value-inflection events that have moved/will move the private mark:

• Nov 2023 — €50M Series B + first commercial machine sale (OVHcloud). Founding commercial proof point.
• May 2025 — Belenos 12-qubit launch (4,000× prior gen).
• Oct 2025 / Apr 2026 — Lucy delivered to & inaugurated at EuroHPC/CEA TGCC (€8.5M, French-German vendor partnership w/ attocube) — the flagship sovereign reference customer.
• Nov 2025 — OVHcloud deal to host Belenos + Canopus on sovereign cloud from mid-2026.
• Feb 2026 — MerLin QML platform launch (CUDA-Q / NVIDIA tie).
• May 2026 — Macron's +€1B France-2030 quantum funding + "European preference" procurement — the macro tailwind.
• June 2026 — DARPA Quantum Benchmarking Initiative Stage A selection — the highest-credibility external, US validation to date; 6-month system-concept report on the SPOQC architecture toward utility-scale by 2033. This is the most important recent catalyst.

Pattern: the mark moves on sovereign reference wins, named-architecture validation (DARPA), and capital events — not on physical-qubit numbers. The next re-rate is gated on (a) a new funding round at a disclosed valuation, and (b) the first logical-qubit / fault-tolerance demonstration.

---

Phase C — Judge people & books

Lens 9 · Management

• Niccolò Somaschi — Co-founder & CEO. PhD in semiconductor photonics / nanostructured materials; ex-CNRS postdoc, FORTH (Greece) PhD researcher; worked directly on solid-state quantum light sources. Deep-domain technical founder — the science is his, not bolted on.
• Pascale Senellart — Co-founder & Scientific Advisor / CSO. Professor & Research Director at CNRS C2N, internationally-recognized authority on single-photon sources — her 2008 position-controlled, high-purity source technique is the company's foundational IP. Scientific-founder credibility here is genuinely top-tier — this is the rare deep-tech where the Nobel-adjacent science is in-house.
• Valérian Giesz — Co-founder & COO. Institut d'Optique engineer (photonics/nanoscience) + HEC Paris (Challenge+ entrepreneurship) + Paris-Saclay PhD — the commercial/operational founder bridging lab→business.
• 2026 governance build-out (the scale-up tell): appointed Cyril Dujardin as COO and Michel Zecri as VP of Industrialization — explicitly "for its next phase of industrial scale-up". Bringing in industrialization leadership is exactly the right move to take a lab champion to a manufacturer — and a signal the founders know operations, not science, is now the binding constraint.
• Capital allocation: disciplined-by-necessity — vertically integrated the chokepoint input (cleanroom/pilot line) rather than buy-vs-build outsource; mix of equity + non-dilutive (EIC, government). No buybacks/dividends (pre-profit). The risk is under-capitalization vs. a $1B+ race, not value destruction.
• Archetype: classic deep-tech scientific-founder trio, now layering professional operators on top — the healthy maturation pattern, executed on schedule.

Lens 10 · Forensic Red Flags

Private, unaudited — no income statement, balance sheet, or cash-flow statement to forensically examine. Standard forensic lenses (revenue recognition, receivables vs. revenue, SBC, goodwill) are n/a — no audited financials disclosed. The structural caution flags for a company of this profile:

• Revenue quality / mix: an unknown share of "revenue" is government grants and co-funded contracts (EuroHPC 50%-funded, France 2030, EIC) rather than arm's-length commercial demand. The €8.5M Lucy machine was 100% publicly funded. Grant-dependent revenue can look like product-market fit when it is industrial policy — the key thing to verify on any future raise.
• Estimate dispersion: the 4× spread in third-party revenue estimates ($5.5M vs $23.8M) signals no reliable public financials — discount all of them.
• Roadmap slippage risk: the "first logical qubit by 2025" milestone against the DARPA framing of utility-scale by 2033 suggests the timeline has already stretched (2024–2030 → 2026–2033). Watch for further slip.
• Component-claim verifiability: the "100,000× component reduction" is a theoretical architecture claim, not a demonstrated result — flag as marketing until peer-reviewed / DARPA-benchmarked.

Regulatory findings (required sub-section):

• SEC (EDGAR LR + AAER): None possible — "Quandela has no CIK — it is private and not required to file with the SEC. No EDGAR enforcement search is possible.".
• Non-SEC (FTC/DOJ/FDA/CFPB/EU/etc.): web search for "Quandela" (FTC OR DOJ OR FDA OR consent decree OR settlement OR fine OR penalty) enforcement returned no material enforcement, litigation, or penalty findings.
• Item 3 (Legal Proceedings): n/a — no 10-K exists (private, non-filer).
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER, both zero/not-applicable), web search, and the absence of any filer obligation, as of 2026-06-24. (Standard for an early private — absence of findings ≠ audited clean books.)

---

Phase D — Project & stress-test

Lens 11 · IPO-Readiness & Path-to-Tradeable (private overlay)

No EPS model (pre-profit, private). The question that matters: how far is this from a tradeable security, and does cash reach the next inflection?

• Stage / readiness (on private-watch.json's 1–5 scale): ~2 (growth), trending toward 3. Series B in Nov 2023 (~$70–75M lifetime), ~150 employees, real revenue but pre-scale. No S-1, no secondary market activity, no crossover-fund round — an IPO is not on a 1–2-year horizon. The realistic path-to-tradeable is 3–5+ years and most likely runs through either (a) a much larger growth round at a disclosed valuation, (b) consolidation into a better-capitalized platform, or (c) an EU-sovereign / strategic outcome — not a near-term US listing like Xanadu's.
• Milestones that unlock the next mark: (1) first logical qubit / fault-tolerance demonstration on the spin-photon platform; (2) a Series C at a disclosed up-round (the single cleanest readiness signal — and its absence to date is notable); (3) DARPA QBI Stage B/C progression (external de-risking); (4) recurring commercial (non-grant) machine bookings.
• Runway-to-catalyst: undisclosed burn vs. ~$70–75M raised + ~150 heads + a capex-heavy cleanroom → a raise is plausibly needed within ~12–24 months to fund the fault-tolerance push. This is the key risk to track. ``
• Comparable-exit framing: Xanadu ($3.6B SPAC) and PsiQuantum ($7B) set the photonic-QC ceiling; Quandela at ~1–2% of those marks is either a deep-value option or correctly priced for being out-capitalized. No forecast.ts EPS line logged (private, no EPS; and per --watchlist rules, no forecast created in the sweep).
• Write-back: added a private-watch.json entry (stage 2, readiness "growth→late-stage", catalyst = first logical-qubit demo + Series C) with this dossier path, so privates.ts shows the name dossier-warm.

Lens 12 · Bull vs Bear

Bull case. Quandela owns the hardest input in photonic quantum computing — deterministic, on-demand single-photon sources rooted in 15 years of in-house CNRS science no rival can quickly copy — and is the only company in its bucket actually selling and shipping machines to customers' data centers, with a real €8.5M ASP and blue-chip sovereign references (CEA/EuroHPC, OVHcloud, EDF). It sits at the center of a €1B+ French/EU industrial-policy tailwind with structural "European-preference" procurement protection, and just earned the field's most credible external validation (DARPA QBI Stage A). If the hybrid spin-photon architecture's "100,000× component reduction" holds, it could leapfrog rivals on the path to fault tolerance — the only metric that ultimately matters. Capitalized at ~1% of PsiQuantum, it is the cheapest credible option on photonic fault-tolerance.

Bear case. Three things that could permanently impair it: (1) Capital asymmetry. Utility-scale QC is a ~$1B+ race; PsiQuantum has $2.1B and a $7B mark, Xanadu is going public at $3.6B — Quandela has ~$75M. In a race won by capital + manufacturing scale, a 100× funding gap is often decisive regardless of who has the better physics. (2) Grant-dependent demand. Much of its "traction" is publicly-funded (EuroHPC 50%-funded machines, France 2030, EIC) — strip the subsidy and arm's-length commercial demand for NISQ-era photonic machines is unproven. (3) Modality risk. Photonic QC may simply lose to superconducting (Google/IBM/IQM) or neutral-atom (Pasqal/QuEra) before it reaches fault tolerance; deterministic photon sources don't matter if the whole modality is a dead end.

Pre-mortem (18 months out, thesis broke): Quandela ran low on cash in late 2027, raised a flat/down round (no US crossover appeared), the logical-qubit demo slipped again, a superconducting rival hit ~100 logical qubits first, and the French state quietly concentrated its quantum bets — leaving Quandela a well-engineered acquisition target rather than the photonic champion.

Contrarian view (what the market refuses to see): in a sovereignty-driven industry, being European and shipping real machines may matter more than winning the global fault-tolerance race outright — Quandela's most likely "win" is a protected, subsidized, mid-scale European-champion business (or a strategic exit), not beating PsiQuantum. The market either over-discounts that floor or over-credits the moonshot ceiling.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration is political, not just commercial. The biggest "customers" are the French state and EU bodies. If French quantum policy concentrates around a single national champion (and a multi-modality strategy cannot fund five companies to utility scale forever), Quandela's lifeline is a political decision it doesn't control.
• The moat may be the wrong moat. Deterministic single-photon sources are a genuine advantage within photonics — but if the market reads photonics as structurally behind superconducting/neutral-atom on logical qubits today (rivals citing 90–100 logical qubits in 2026; PsiQuantum util-scale 2027; Quandela's own utility horizon 2033), then Quandela is winning a sub-race that doesn't get to the finish.
• Most dangerous competitor bulls underestimate: PsiQuantum. Same photonic family, 30× the capital, a 2027 utility-scale claim, BlackRock/NVIDIA-backed, breaking ground on Brisbane + Chicago sites. If photonic QC works, the best-capitalized photonic player most likely wins — and that is not Quandela.
• Capital-allocation / dilution risk: a capex-heavy vertical-integration strategy (own fab/cleanroom) on a ~$75M base is admirable but cash-hungry; the next raise is dilution or distress, and the absence of a disclosed up-round since Nov 2023 is itself a yellow flag.
• The killer scenario (and plausibility): a superconducting or neutral-atom vendor demonstrates commercially-useful fault tolerance first (plausible by ~2028–2030) → photonic-QC capital dries up → Quandela can't fund its 2033 roadmap → forced sale at a modest mark. Plausibility: moderate-to-high. This is the real risk, and it is largely exogenous to how good Quandela's photons are.

Lens 14 · Management Questions (ordered by information value)

1. What is your current cash runway in months, and does it reach your first demonstrated logical qubit without a new raise? If not, what's the gap and the plan?
2. What share of 2025 recognized revenue was arm's-length commercial vs. government grant / co-funded (EuroHPC, France 2030, EIC)? What's the trajectory of the commercial line specifically?
3. You're capitalized at a small fraction of PsiQuantum and Xanadu. Why does Quandela win — or even survive — a utility-scale race that appears to need $1B+? What's the realistic end-state (independent champion / consolidation / strategic)?
4. The "100,000× component reduction" — is that a demonstrated result, a simulated architecture claim, or a target? What has DARPA's Stage A benchmarking actually validated so far?
5. Your roadmap moved from 2024–2030 (first logical qubit by 2025) to a 2026–2033 / utility-by-2033 framing. What slipped, and why should we trust the new dates?
6. When is the next funding round, what size, and are you targeting a US crossover investor (Fidelity/T. Rowe/Coatue)? Why hasn't one come in yet?
7. Walk me through the unit economics of an €8.5M machine — gross margin, and how that scales as you move from ~4 machines/year toward volume.
8. If photonic QC reaches fault tolerance after superconducting or neutral-atom, what is your honest contingency — does the company survive that, and how?
9. How dependent is your moat on Pascale Senellart and the C2N relationship? What happens to the single-photon-source edge if key scientific founders step back?
10. Why should a sovereign buyer choose photonic Quandela over French peers Pasqal (neutral-atom) or Alice & Bob (cat-qubit) for the same France-2030 euros?
11. What's the real status of the spin-qubit quantum memory in the SPOQC architecture — the linchpin for the non-local-link error correction? Demonstrated, or still on the bench?
12. What is your manufacturing yield on single-photon sources, and is the pilot line a genuine volume capability or a demonstrator?
13. How do you defend the on-prem machine-sale model as cloud-access economics (IonQ/IBM/PsiQuantum) potentially commoditize NISQ-era access?
14. What's the IP defensibility beyond the founding patents — how much of the source advantage is patent-protected vs. trade-secret / know-how, and when do the foundational patents expire?
15. Five years out, is the most likely outcome an IPO, an acquisition, or continued private/sovereign operation — and what milestone would flip you toward each?

---