Optical Computing
The best-capitalised, hardest-physics bet in quantum — $4B+ in private + sovereign money betting that silicon-photonic fusion-based computing leapfrogs straight to a million error-corrected qubits by ~2027, with the one fact that matters still unproven (a utility-scale machine running) and a freshly installed operator-CEO signalling the company is now an execution problem, not a science one.
Research
The verdict
The best-capitalised, hardest-physics bet in quantum — $4B+ in private + sovereign money betting that silicon-photonic fusion-based computing leapfrogs straight to a million error-corrected qubits by ~2027, with the one fact that matters still unproven (a utility-scale machine running) and a freshly installed operator-CEO signalling the company is now an execution problem, not a science one.
PsiQuantum is a Palo Alto-based quantum-computing company founded in 2016 by four physicists out of the University of Bristol and Imperial College London — Jeremy O'Brien (CEO until Feb 2026, now Executive Chairman), Terry Rudolph (Chief Architect), Pete Shadbolt (Chief Scientific Officer) and Mark Thompson (Chief Technologist). Its founding premise is a contrarian bet on scale: that quantum computing only becomes commercially valuable once you have error correction, which requires on the order of one million physical qubits — so building anything smaller is a detour. Everyone else (IonQ, Rigetti, IBM, Google) ships small, noisy machines today and scales up; PsiQuantum has spent a decade building nothing publicly useful while it engineers a straight shot to the utility-scale endgame.
The business model in plain terms. PsiQuantum is not (yet) selling quantum computers or cloud access at scale. It is a capital-intensive deep-tech builder whose "product" today is a roadmap, a manufacturable chip, and two government-anchored construction projects. Revenue to date is a blend of (a) government R&D contracts (DARPA US2QC, a $10.8M AFRL contract, the US2QC program), (b) sovereign infrastructure deals where governments fund the data centres, and (c) a nascent applications/collaboration layer (e.g. a research agreement with the National Cancer Center Japan in oncology). The end-state model is utility-scale Quantum Compute Centers — datacenter-sized machines sold/leased for drug discovery, materials science, AI and personalised medicine.
Customers / suppliers / competitors.
Contract structure. No take-or-pay, no recurring SaaS at scale yet. The defining structural feature is that PsiQuantum has externalised its single biggest cost — the data centres — onto sovereign balance sheets (Australia funds Brisbane, Illinois/US funds Chicago). That is a moat-shaped financing structure and a concentration risk in one (see Lens 3, Lens 13).
Map: upstream materials/equipment → PsiQuantum design → GlobalFoundries fab → assembly/cryo integration → Quantum Compute Center → end user. Named stakeholders along the chain:
Chokepoints: (1) GF fab yield on exotic quantum components — the binding constraint; (2) BTO modulator supply (partly internalised); (3) cryoplant scale-up — cooling tens of thousands of chips is itself an industrial-engineering problem, not just a physics one. This lens is named, not generic — the chain is essentially SOI wafers → GF Malta + GF Dresden → PsiQuantum cryo-integration → sovereign-funded data centre.
Bargaining power: strong over governments (they need a flagship; PsiQuantum is the most-funded option), weak over GlobalFoundries (PsiQuantum needs GF more than GF needs PsiQuantum's quantum volume today). Over future customers: undetermined — there is no installed base yet.
segments.csv is an empty stub — no segment-level revenue is disclosed (private company).. Qualitatively, the business decomposes into three not-yet-monetised segments, each ``:
Trend: spend is accelerating (two simultaneous data-centre builds + a $1B Series E deployed into them); revenue remains contract/grant-shaped. n/a — private, not disclosed for any segment margin or growth figure.
PsiQuantum has raised ~$2.32B across 8 private rounds from ~38 investors, on top of ~$1.7B in government funding — roughly $4B+ in total capital, making it the most-funded pure-play quantum computing company in the world.
| Round | Date | Amount | Valuation | Lead / notable | Source |
|---|---|---|---|---|---|
| Series C | Nov 2019 | (BlackRock first invests) | — | BlackRock | |
| Series D | Jul 2021 | $450M | ~$3.15B | BlackRock; +Baillie Gifford, M12 (Microsoft), Blackbird, Temasek | |
| (reported) | Mar 2025 | ~$750M | ~$6B pre-money | — | |
| Series E | Sep 2025 | $1.0B | ~$7B post-money | BlackRock, Temasek, Baillie Gifford; +NVentures (NVIDIA), Macquarie, Ribbit, Adage, QIA, Type One, Counterpoint Global (Morgan Stanley), 1789 Capital, S Ventures (SentinelOne) |
Use of the Series E: break ground in Brisbane + Chicago, deploy large-scale prototypes to validate the systems architecture, advance the photonic chips & fault-tolerant architecture.
Burn signal. Two simultaneous datacenter-scale builds + a decade of pre-revenue R&D imply a very high burn; the $1B Series E plus ~$1.7B sovereign money is the runway to first utility-scale machines (~2027). Exact burn/cash position: n/a — private, not disclosed. The valuation step-up ($3.15B → ~$6B → ~$7B in ~4 years) tracks the broader quantum-stock mania (Lens 8), not realised revenue.
No earnings calls exist (private). Substitute: the public-communications arc.
Syndicate quality — strong, and increasingly IPO-shaped. PsiQuantum's cap table reads like a pre-IPO book: BlackRock (since 2019), Temasek, Baillie Gifford (a classic crossover/pre-IPO growth investor), Microsoft's M12, NVIDIA's NVentures (strategic — the chip incumbent buying optionality), Qatar Investment Authority (sovereign), Counterpoint Global / Morgan Stanley and Adage (crossover public-markets money). The presence of crossover funds (Baillie Gifford, Counterpoint Global, Fidelity-style money) is the standard IPO-proximity tell. NVentures' entry is doubly notable — NVIDIA validating a non-superconducting approach.
Public comps (context for the $7B private mark) — all ``:
| Company | Approach | Status | Mkt cap / valuation | ~Revenue | EV/Sales | Source |
|---|---|---|---|---|---|---|
| Quantinuum | Trapped-ion | Public (Jun 2026 IPO) | ~$15.7B | building | ~505x | |
| IonQ | Trapped-ion | Public | (largest pure-play by rev) | ~$130M ('25); $225–245M ('26 guide) | ~179x | |
| Rigetti | Superconducting | Public | — | $4.4M (Q1'26); $569M cash, no debt | n/a | |
| PsiQuantum | Photonic | Private | ~$7B | ~grant/contract only | n/a — not meaningfully revenued |
Read: at ~$7B private with no commercial revenue, PsiQuantum is valued below public peers Quantinuum ($15.7B) and well below where the public quantum names trade on sales — but those public multiples (179x–505x sales) are themselves a mania signal, not a floor. A 2026–27 PsiQuantum IPO into a hot tape could mark it far above $7B regardless of fundamentals; a tape reset (Lens 13) cuts the other way. No P/E, no EV/EBIT — the company has no earnings; any such multiple would be fabricated, so: n/a.
PsiQuantum has no stock, so the relevant "price" is its private mark and the public quantum index that drives sentiment and its eventual IPO multiple. The benchmark S&P Kensho Global Quantum Computing index was +69.3% YTD through May 2026 vs S&P 500 +10.7%. Events that have re-rated PsiQuantum / the sector:
Pattern: this name (and the sector) reacts to (1) hardware/manufacturing proof points, (2) capital events (mega-rounds, government money, IPOs), and (3) policy (CHIPS, sovereign deals) — far more than to revenue, which barely exists. It is a narrative-and-capital-driven complex, highly sensitive to a risk-on/risk-off turn.
The headline development for this dossier: the CEO transition (Feb 10, 2026). Co-founder Jeremy O'Brien stepped from CEO to Executive Chairman; Victor Peng — former CEO of Xilinx (which he transformed into the adaptive-computing leader, sold to AMD for ~$49B in 2022) and President of AMD (Feb 2023–Aug 2024) — was named Interim CEO while the board runs a permanent search.
n/a — private, not disclosed. Peng is interim — a caveat (see red flags).n/a (no returns yet).No financial statements exist — PsiQuantum is private with no CIK. So classic forensic-accounting tests (revenue recognition, receivables vs revenue, SBC flattering non-GAAP, goodwill) are n/a — no audited financials disclosed. What can be assessed as risk:
n/a).Regulatory findings (required). From regulatory/regulatory-findings.md: No SEC findings possible — PsiQuantum has no CIK and does not file with the SEC (0 LRs, 0 AAERs). Non-SEC web search ("PsiQuantum" (FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine OR penalty) enforcement) returned no material enforcement actions, settlements, or fines against PsiQuantum across the searches run for this dossier. No 10-K Item 3 exists (private). Conclusion: no material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER — none possible, private) and web search as of 2026-06-23. The only material government interactions are funding relationships (CHIPS LOI, DARPA, AFRL, Australian grants), not enforcement.
No EPS projection is possible (no revenue, no financials) — building one would be fabrication. Instead, per the +private overlay, the question that matters: what unlocks an S-1, and when?
ipo_readiness (1–5 scale): 3 (growth/late-stage) — not yet pre-IPO/secondary-active (4) on public evidence: there is no disclosed IPO timeline, and the value-inflection event (a working utility-scale machine) is still ahead.Forecast (Brier-tracked): per --watchlist rules, no forecast.ts create is run in breadth mode. A candidate binary to log later: "PsiQuantum demonstrates a fault-tolerant / utility-scale machine (>1k logical-qubit-equivalent or first error-corrected utility milestone) by YE2027 — p ≈ 0.35 " and "PsiQuantum files an S-1 / goes public by YE2028 — p ≈ 0.40 ". Both deferred to a human-gated /thesis pass.
Bull case. PsiQuantum is the only quantum company that, if the physics works, plugs straight into the world's most powerful scaling engine — the semiconductor supply chain — and skips the painful intermediate-NISQ era entirely. It has the most capital ($4B+), the deepest sovereign entanglement (Australia + Illinois + US defence have staked their quantum strategy on it), a manufacturable chip with credible fidelities (99.98% SPAM, 99.22% two-qubit fusion-gate ), the right strategic investor (NVIDIA), and a freshly installed operator-CEO (Peng) to industrialise it. If a utility-scale machine comes alive ~2027, PsiQuantum could be the fault-tolerant leader, IPO into a roaring tape, and re-rate far above $7B — the Quantinuum $15.7B IPO is the proximate proof the market will pay for the category.
Bear case (permanent-impairment risks).
Pre-mortem (18 months out, thesis broke): Brisbane/Chicago commissioning slipped past 2027 on cryoplant/yield problems; loss thresholds proved harder than the Nature paper implied; the quantum tape rolled over; follow-on capital got expensive; the "interim" CEO search dragged; the $7B mark looked stale. The bet didn't blow up — it ran out of runway before the science finished.
Contrarian view — what the market is refusing to see: the consensus frame is "quantum is a bubble; PsiQuantum is the most over-funded part of it." The contrarian read is the inverse on a long horizon — PsiQuantum is the only player whose architecture is even capable of the million-qubit endgame, and it has externalised its biggest cost onto governments that politically cannot let it fail. The risk isn't that it's over-funded; it's that it's under-de-risked on physics — those are different bets, and the market conflates them. Are current marks too high? On revenue, absurdly. On optionality-to-the-endgame, defensibly — if you believe the loss problem is engineering, not physics.
Dismantling the bull case. The structural break: there is no product, and there may never be one on this architecture. A decade and $4B in, the company's "proof" is a chip with good component fidelities and a Nature paper — not a working error-corrected computer. The bull case rests entirely on extrapolating from 99.22% two-qubit fusion-gate fidelity and ~80% optical transmission to a million-qubit fault-tolerant machine — a leap that requires loss to cross ~99.5%, which no one has demonstrated at scale. Revenue concentration: ~$1.7B of the capital is government — and governments change. A new Australian government, a CHIPS-policy reversal, or a DARPA de-scope removes the floor. The most dangerous competitor bulls underestimate: not another photonic player but the trapped-ion incumbents (IonQ, public Quantinuum at $15.7B) who are already selling and could reach commercial error correction with fewer, better qubits — making PsiQuantum's million-qubit slog obsolete before it finishes. Worst capital-allocation move: committing to two simultaneous datacenter-scale builds before demonstrating a single utility-scale machine — classic deep-tech over-commitment funded by a hot market. What must hold for the $7B mark: that the loss problem is engineering (not physics), that 2027 holds, that the tape stays hot, and that the interim-CEO transition doesn't stall execution. If milestones slip 20–30% (i.e. 2027 → 2029+): the mark is uninvestable — an infinite-burn pre-revenue company in a cooled sector with stale private marks and follow-on rounds at flat-or-down valuations. Single scenario that permanently impairs: the loss/error-correction threshold proves physically intractable at silicon-photonic scale — at which point a decade and $4B buy a patent portfolio and a sunk fab relationship, not a company.
A $2.5B market cap on $682K of FY25 revenue — QUBT is a $1.5B treasury wrapped in a photonics R&D lab, sold as a quantum-computing story; the balance sheet is real, the revenue is not, and a securities-fraud class action over the exact gap between the two is unresolved.
The most credible analog-photonic-compute bet in Europe — a vertically-integrated TFLN coprocessor with a real fab, a first paying cloud customer (IONOS) and an ARM-founder-grade board — but at $80M raised it is a seed-stage David swinging at a $4B-valued field (Lightmatter) and a just-acquired peer (Celestial AI → Marvell, $5.5B); WATCHING for IPO-proximity, not yet tradeable.
A genuinely differentiated all-optical scale-up fabric with a real CEO and a real AMD/ARIA anchor — but a $93.5M micro-cap fighting $1.2B–$4.4B peers and a hyperscaler-backed open-standard tide; the bet is a strategic acqui-buy (à la Marvell/Celestial), not a standalone IPO.