Xanadu Quantum Technologies

Phase A — Understand the business

Lens 1 · Company Overview

Xanadu builds full-stack photonic quantum computers — it uses light (photons) as the computational medium, the "photonics modality," vs. the matter-based modalities (superconducting, trapped-ion, neutral-atom) of most rivals. The pitch: photons run at room temperature, need no cryogenics/laser-cooling, are manufacturable on existing silicon-photonics processes, and can be networked over optical fibre between racks — which the company argues is the only modality that scales to the hundreds-of-thousands-of-qubits regime. Two hardware milestones anchor the story: Borealis (216-qubit, claimed "quantum supremacy" in 2022 — a 2-minute computation vs. ~7M years on Fugaku) and Aurora (2025 — "world's first networked, modular, scalable photonic quantum computer," 4 racks / 35 photonic chips / 13 km of fibre, published in Nature).

The software leg is PennyLane, an open-source, hardware-agnostic quantum-programming SDK used across "more than 120 universities" and on rival hardware (IBM, IonQ, QuEra). It is the company's distribution wedge — get the world's quantum developers building in PennyLane, then funnel mature workloads onto Xanadu's photonic hardware as it scales.

How it makes money — barely, and that's the point. Xanadu is explicitly pre-commercial: "we have not achieved sustained commercial revenue from the sale or deployment of utility-scale quantum computers". FY25 revenue was $4.617M, from two channels: (i) professional services (sponsored research, proof-of-concept, quantum education) and (ii) compute/other (QCaaS access, co-development of algorithms). PennyLane is open-source and generates no recurring subscription revenue today. Revenue is a technology-viability signal, not a business — management says outright it "will not be a reliable indicator of our performance" near-term. The intended future model: QCaaS, dedicated sovereign system sales, a paid enterprise PennyLane, IP licensing (it already supplies PICs to "a major defense contractor" for a US government quantum-sensing program), and adjacent photonics monetization (quantum internet/sensing/comms).

• Customers: governments and Fortune-500 R&D partners — Volkswagen, Toyota, Mitsubishi Chemical, Rolls-Royce, Raytheon/RTX (BBN), Corning, plus US/Canada/EU national labs. Highly concentrated: three customers = 78% of FY25 revenue; two customers = 63% (FY24) and 85% (FY23). The buyers are mostly government agencies on multi-year arrangements.
• Suppliers / fabrication partners: Applied Materials (300mm transition-edge sensors), Tower Semiconductor (silicon-photonics foundry), DISCO (wafer processing), Corning (low-loss fibre), Thorlabs, HyperLight (thin-film lithium niobate).
• Contract structure: project/milestone professional-services + government grants/contributions; no take-or-pay, no meaningful recurring revenue yet.

Lens 2 · Supply Chain

Upstream → Xanadu → end customer, named stakeholders:

Inputs / fabrication (upstream):

• Silicon-photonic PIC foundry: Tower Semiconductor (TSEM) — expanded Feb 2026 to build PICs on Tower's platform. This is a genuine single-point-of-dependence for at-scale chip fab.
• Process / materials / equipment: Applied Materials (AMAT — 300mm superconducting TES process), DISCO (wafer dicing/processing), Corning (GLW — low-loss optical fibre & fibre arrays), Thorlabs (fibre components), HyperLight (thin-film lithium-niobate, <2 dB/m waveguide loss).
• In-house: C$10M Toronto photonic-packaging facility (opened June 2025); cryogenic detector equipment; the planned Project OPTIMISM manufacturing build (heterogeneous integration, PIC packaging, wafer test) — the vertical-integration bet.

The company (midstream): designs photonic chips + GKP-qubit architecture + qLDPC real-time error correction + PennyLane compiler/SDK; assembles modular "quantum racks".

End customers (downstream): US government / DARPA (Quantum Benchmarking Initiative Stage B, up to $15M; AFRL 4-yr PIC partnership; RTX/BBN + UMD INSPIRED program), Government of Canada/Ontario, Singapore A*STAR (MoU), and enterprise R&D (VW, Toyota, Mitsubishi Chemical, Rolls-Royce).

Chokepoints / single-source dependencies:

1. Optical loss is the physical chokepoint — not a vendor, but the binding constraint. The filing concedes only "twelve logical qubits with photon loss not meeting the required levels" today; the independent "1% loss wall" analysis frames sub-1% per-component photon loss as the make-or-break threshold for photonic FTQC.
2. Tower Semiconductor / silicon-photonics fab — a third-party foundry dependency for any volume ramp.
3. Government funding is a supply-chain input in its own right — the SIF loan funded the Project that built the fab capability; Project OPTIMISM (C$390M, in negotiation) would fund the manufacturing scale-up. Cut the grants and the roadmap slows.

Lens 3 · Competitive Advantages (moats)

What's genuinely defensible:

• A differentiated, peer-reviewed architecture. Room-temperature, modular, fibre-networked photonics with GKP qubits + real-time qLDPC error correction is a distinct technical bet, validated in Nature (Borealis 2022, Aurora 2025). Management claims a "10-to-100-fold reduction in error-correction overhead" vs. superconducting/trapped-ion — unverified but architecturally plausible if loss is tamed.
• PennyLane as a distribution moat / standard. 120+ universities, taught in curricula, runs on competitors' hardware — a real ecosystem foothold and the closest thing Xanadu has to a network effect. Caveat: it's open-source and monetizes ~nothing today, and its value partly depends on rivals (IBM/IonQ/QuEra) continuing to support it.
• Vertical integration — the only pure-play claiming to own hardware + the leading open SDK + algorithms end-to-end.
• Sovereign/defense entanglement. DARPA QBI Stage B, AFRL, RTX/BBN, a live US quantum-sensing PIC deployment, Lockheed Martin on the cap table, ex-US-Undersecretary-of-Defense Heidi Shyu on the board — a national-security moat that's hard for a foreign startup to replicate and aligns with the "$2B US sector build-out".

Bargaining power: weak, structurally. The filing is candid — competitors (Amazon, Google, IBM, Intel, Microsoft) "have substantially greater brand recognition, customer relationships and financial... resources" and "may be able to cross-subsidize their quantum offerings". Xanadu is a price-taker to its foundry and a supplicant to government funders. The moat is technological optionality, not commercial power. None of it is monetized — a moat around a pre-revenue lab.

Lens 4 · Segments

Single operating and reportable segment — "the Company operates in a single operating and reportable segment and manages segment performance... based upon consolidated net loss". (segments.csv is empty; this is correct, not a gap.) Revenue splits qualitatively into professional services (the bulk) and compute/other (QCaaS), no quantitative breakout disclosed.

Geography: Canada-centric operations (Toronto HQ, CAD functional currency); revenue skewed to US and Canadian government agencies. Revenue by quarter: Q1'25 $0.70M → Q2 $1.06M → Q3 $0.99M → Q4 $1.875M; FY25 $4.617M (+191% YoY) after FY24 fell 36% to $1.589M (from $2.479M FY23). The trend is lumpy and milestone-driven — the FY24 dip and the Q4'25 spike both trace to a handful of government contract milestones, not a growth curve. Read it as project accounting, not demand.

---

Phase B — Measure performance

Lens 5 · Earnings Result

Most recent print: Q1 2026 (reported May 14, 2026):

• Revenue $2.8M, +4x YoY (from $0.7M Q1'25); net loss $20.6M (vs. $12.2M Q1'25); Adj. EBITDA loss $13.9M (vs. $10.6M); cash $272.5M (from $16.2M at YE25, post-deal).
• No consensus to beat/miss against in any real sense — coverage is two analysts. Revenue is milestone noise; the number that matters is cash $272.5M and burn.

FY2025 (the audited backbone):

• The engine is R&D, accelerating: $55.2M FY25 (+41%), driven by a $12.6M jump in wafer-consumption expense as it scales chip fab. Quarterly net loss ramped Q1'25 $(12.2M) → Q2 $(15.1M) → Q3 $(20.3M) → Q4 $(23.0M) — burn is rising, by design.
• G&A FY25 carried $7.4M of one-off de-SPAC professional fees; clean G&A is ~$8M.
• Margins are not a meaningful frame — gross "margin" exists only on the trivial $4.6M services line (cost of revenue just $0.36M; ~92% gross on a rounding-error base). This is a cash-burn story, full stop.
• Balance sheet flags: cash fell $61.4M in FY25 to $16.2M pre-deal — the company was running low and needed the SPAC. Total assets $70.6M (FY25) vs. $151.0M (FY23) — a shrinking balance sheet pre-recap. Working capital $31.2M. Debt is the government SIF Loan (received C$35.7M; revenue-based, forgivable, repay from Apr 2028, 150% cap) + FedDev Loan ($2.7M).
• Going concern: recurring losses "initially raised substantial doubt"; management concluded the doubt is alleviated by the $302M de-SPAC/PIPE proceeds → "sufficient... for at least twelve months". So: solvent, but on a clock.
• Market reaction (the real Lens-5 event): see Lens 8 — the stock has been a rollercoaster (ATH ~$57 CAD/$36 USD → ~$11 now), and the single biggest move was a ~65% one-day crash in early May 2026 on a resale-registration filing.

Lens 6 · Earnings Calls (sentiment trend)

Only ~2 quarters exist as a public company (FY25 results Apr 9, 2026; Q1'26 May 14, 2026). transcripts/ is empty; this is ``-grounded and thin by necessity. Management's consistent framing across both: "investment phase," "pre-commercial," profitability "takes a back seat," revenue a viability signal not a KPI. Tone is roadmap-and-milestone forward (Aurora, 12 logical qubits, 2029-30 target, government partnerships), not numbers-forward. The recurring phrases: "utility-scale," "fault-tolerant," "full-stack," "room-temperature," "sovereign compute." The thing they stopped saying: any near-term commercial-revenue ramp — the language is carefully conditioning the market for years of losses. No tone shift yet — too few data points. Watch the next 2-3 calls for whether "Project OPTIMISM closed" and "loss-wall progress" enter the script.

Lens 7 · Comps

Pure-play quantum is a valuation-on-narrative cohort — every name trades on government support and 2030s optionality, not fundamentals. Multiples are ``, dated; where I can't source one cleanly I mark it.

ROE/dividend columns are n/a across the cohort — all loss-making, none pay dividends; a 5-yr avg ROE is meaningless for pre-revenue names (Xanadu's "ROE" is a large negative on a shrinking equity base).

Read: Xanadu's ~710x EV/sales on FY25 revenue (≈293x on Q1'26-annualized) is the richest in a cohort that is itself the poster child for "bubble" P/S (IonQ 109x is the cheap one here). On a cash-and-optionality basis it's more defensible: ~$3.0B EV against $272.5M cash means the market assigns ~$2.8B to the technology/roadmap — a venture-style bet wearing a public ticker. The honest peer is PsiQuantum (private), not the trapped-ion/annealing names; Xanadu is best understood as the only public way to own the photonic-FTQC thesis, which is exactly why it commands a scarcity premium.

Lens 8 · Stock-Price Catalysts

Short history (public since Mar 27, 2026) but already violently catalyst-driven:

• Mar 27, 2026 — Nasdaq/TSX debut at the $10.00 PIPE reference. De-SPAC raised ~$302M gross.
• Apr 2026 — melt-up to ATH ~$57 CAD (Apr 16) / ~$36.12 USD close (May 1) — the quantum-momentum trade + tiny float (only ~3.2M ex-SPAC shares + $275M PIPE).
• Early May 2026 — ~65% one-day CRASH when Xanadu filed a 424B3 resale prospectus to register up to ~293.6M Class B shares for resale — the entire insider/VC base became sellable, an overhang shock. This is the single most important price event and the defining risk of the name.
• May 14, 2026 — Q1'26 results (4x revenue, wider loss, $272.5M cash); stabilized into the high-teens, then drifted to ~$11 by late June.
• Ongoing — Project OPTIMISM (C$390M) government-funding negotiation (announced Mar 11) is the live up-catalyst if it closes.

Pattern: XNDU trades on (1) the sector-wide quantum momentum cycle (it moved with IONQ/RGTI/QBTS on the May "quantum trade roars back" day ), (2) supply/float mechanics (lock-up & resale-registration dominate everything else), and (3) government-funding headlines. It does not trade on revenue. Expect lock-up expiries and any secondary/ATM raise to be the dominant price drivers for the next 12 months.

---

Phase C — Judge people & books

Lens 9 · Management

• Christian Weedbrook — Founder, Chair & CEO (49). Ran Old Xanadu since founding in Sept 2016; ex-MIT/UofT postdoc, PhD quantum optics (Univ. Queensland). Track record: built a globally-recognized photonics program from zero — Borealis "supremacy," Aurora, PennyLane adoption, and a top-tier investor base. A scientist-founder, not a commercial operator — the company has never sold a product at scale, and the team's "limited experience operating a public company" is a named risk factor. Skin in the game: very high — 18.2% of Class A → 17.9% of total voting power, and via the dual-class 10:1 structure he + insiders control the company.
• Michael Trzupek — CFO (joined Jan 2026). The most reassuring hire on the page: ex-CFO of Imagination Technologies and Core Scientific (CORZ), plus ~7 years in Microsoft finance (Xbox/Surface) and an Intel start. A real public-company/semis CFO brought in specifically to operate a listed balance sheet — exactly the gap a scientist-founder needs filled.
• COO Rafal Janik (physicist, internal since 2019); CLO Natalie Wilmore (ex-IBM/Skillz/Pagaya); CPO Rebecca Laramée (internal).
• Board signal is strong for the sovereign-compute thesis: Heidi Shyu (former US Undersecretary of Defense for R&E — i.e. the person who ran DoD's tech portfolio), Glenda Dorchak (GlobalFoundries + Cerebras boards — semis/foundry depth), Michelle Reynolds (Reddit Chief Accounting Officer, Audit Chair — credible financial-controls anchor), Eliot Pence (ex-Anduril Head of International — defense GTM), Bill Fradin (SPAC sponsor).
• Capital allocation: the only history is burn into R&D (rational for the stage) funded by VC + forgivable government loans (cheap, non-dilutive-ish capital — genuinely well-played). The de-SPAC was the right move given cash fell to $16.2M. Founder + insiders hold ~98% of Class A and ~51.8% of Class-B economics — alignment is high, but so is entrenchment.
• Red flags (governance): dual-class concentrates control with founder/insiders regardless of economic dilution; foreign-private-issuer + Canadian home-country governance exempts XNDU from several Nasdaq standards and Section 16 short-swing rules; SPAC sponsor earnout (1.1M shares vesting at $12.50/$15.00 hurdles). Standard de-SPAC/founder-control caveats, not fraud signals.

Lens 10 · Forensic Red Flags

As a forensic analyst, on a pre-revenue de-SPAC the risks are dilution, going-concern, capitalization policy and grant accounting — not classic revenue-recognition games (there's almost no revenue to game).

• Revenue recognition: low-risk by size. $4.6M across services (hours/cost input method) + QCaaS (straight-line). Concentration (78% from 3 customers) is the real exposure — a single government contract slipping swings the whole line.
• R&D capitalization: conservative / clean — Xanadu expenses hardware and materials built for research unless they have an "alternative future use". No evidence of flattering the P&L by capitalizing R&D. Good.
• Government grants in "other operating income": FY25 booked $5.0M of CQCP grant and revalued warrants through that line. Watch that grants/credits aren't read as operating quality — they're non-recurring and policy-dependent. The SIF loan is a forgivable, revenue-based instrument carried at amortized cost with non-cash interest — model the repayment (from Apr 2028, capped 150% of principal) but note forgiveness optionality.
• Cash vs. earnings: operating cash burn ($68.0M FY25) exceeds net loss ($70.7M) only slightly — no aggressive accrual divergence; the loss is cash. Clean.
• SBC: modest — $4.16M FY25 (vs. $70.7M net loss); not a non-GAAP-flattering vehicle. Adj. EBITDA add-backs (D&A, SBC, warrant revaluation, $6.9M non-recurring de-SPAC fees) are reasonable.
• Warrant liabilities: $1.6M fair-value change FY25 — the usual de-SPAC warrant mark-to-market noise; immaterial.
• The real forensic risk is the cap table, not the books: the 293.6M-share resale registration is a structural dilution/overhang fact, and the dual-class + unlimited-share authorization let the company issue freely.

Regulatory findings (required):

• SEC: "No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER) as of 2026-06-29. Total SEC findings: 0.". (Expected — the entity has existed publicly for ~3 months.)
• Non-SEC: web search for "Xanadu Quantum Technologies" (FTC/DOJ/FDA/CFPB/consent decree/settlement/fine) enforcement surfaced no material enforcement actions — only routine litigation-risk disclosure and the resale-prospectus selloff coverage, not an action.
• 10-K Item 3 (here, 20-F legal proceedings): the filing discloses ordinary forward-looking litigation risk (incl. potential business-combination-related claims) but no material pending proceeding is itemized.
• Compliance exposure to flag (not findings): XNDU is subject to CFIUS and the Investment Canada Act, plus export controls on quantum tech — a foreign-issuer/dual-use overhang that could constrain US government work or M&A.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

This is not an EPS story — it is a runway-to-catalyst and dilution story. Modeling 3-year EPS for a pre-revenue moonshot would be false precision; the right Lens-11 object (clinical-overlay logic, applied to deep-tech) is does cash reach the value-inflection milestone, and at what dilution.

Runway (the number that matters):

• Cash $272.5M at Q1'26 end.
• Burn: FY25 op-cash burn $68.0M (~$17M/qtr) and rising — Q1'26 net loss $20.6M, Adj-EBITDA loss $13.9M; R&D is accelerating into wafer runs. Forward cash burn ~$20–26M/qtr.
• Runway ≈ 10.5–15 quarters (~2.6–3.8 years) → roughly into 2028–2029 before any new capital.
• Critical mismatch: the value-inflection ("early fault-tolerance building blocks" 2026; error-corrected demos through 2028; 100,000 physical / 500 logical qubits 2029-30) sits at or beyond the end of current runway. Xanadu will almost certainly raise again before it reaches its own headline milestone — via equity (dilutive, into a 293.6M-share overhang), more government money (Project OPTIMISM C$390M — the cheap path), or debt.

Revenue trajectory: services/grant revenue grows lumpily off a tiny base — call it $8–15M FY26, $15–30M FY27 (government-contract-paced), still immaterial to the thesis; real revenue is contingent on FTQC commercialization in the 2029-30+ window.

Base / Bull / Bear (qualitative — no fabricated EPS):

• Base: survives on cash + closes Project OPTIMISM; hits interim loss-wall/error-correction milestones; raises a dilutive round ~2028; trades on roadmap progress. Stock range-bound-to-volatile around scarcity-premium.
• Bull: demonstrable optical-loss progress + a closed C$390M deal + a marquee QCaaS/sovereign contract → re-rates toward IonQ-style scarcity multiple on the photonic thesis.
• Bear: loss-wall doesn't yield on schedule, milestones slip past 2030, government money stalls, lock-up/resale supply floods the float → another leg down toward cash value (~$1/sh of net cash + option value).

Per --watchlist rules, no Brier forecast logged (forecast.ts create skipped). If promoted to a thesis, the scoreable binary is "XNDU demonstrates <1% per-component photon loss / a published error-corrected logical-qubit advance by YE2027," not an EPS line.

Lens 12 · Bull vs Bear

Bull case. Xanadu is the only public pure-play on photonic, room-temperature, networked FTQC — a genuinely differentiated architecture with two Nature-validated milestones and a credible claim to 10–100x lower error-correction overhead if loss is solved. PennyLane is a real ecosystem moat (120+ universities, runs on rivals' hardware). The sovereign-compute alignment is deep and hard to replicate: DARPA Stage B, AFRL, RTX, Lockheed on the cap table, a former US Undersecretary of Defense on the board, and C$390M of Canadian government money in negotiation on top of forgivable SIF financing — a state-backed runway extender most startups can't access. Post-deal it has $272.5M cash and ~3 years to prove the next milestone, a top-tier semis CFO to steward it, and the scarcity premium that comes from being the only way to own this thesis. If photonics is the modality that actually scales, this is a venture-style 5-10x option on a public exchange.

Bear case (permanent-impairment risks). (1) The physics may not cooperate on schedule — 12 logical qubits today, with photon loss explicitly below requirement, vs. a 100,000-physical/500-logical target in 2029-30; the "1% loss wall" is unforgiving and the entire valuation is a bet that fabrication closes it. (2) Dilution is near-certain — runway ends before the milestone, into a 293.6M-share resale overhang that already caused a ~65% one-day crash; every raise is value-transfer from new holders. (3) It's pre-revenue at ~710x EV/sales in a cohort the financial press openly calls a bubble (IonQ/Rigetti/D-Wave P/S 109/836/791x; insiders are net sellers sector-wide). Pre-mortem (18 months out, thesis broken): loss-wall progress stalls, Project OPTIMISM stays "in negotiation," a dilutive 2027 raise prints into the overhang, the quantum-momentum trade rolls over, and XNDU round-trips toward its $272M net-cash floor ($1/sh) — a 2022-style deSPAC de-rating. Contrarian view the market is underpricing: the resale overhang + dual-class control means the float-and-funding mechanics, not the science, will set the price for the next year — and that's a bearish technical fact the "quantum supremacy" narrative obscures. Conversely, the bull contrarian: the government-money depth (SIF + OPTIMISM) makes XNDU far less dilution-dependent than its pure-play peers, which the ~710x headline multiple ignores.

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

Dismantling the bull case:

• Revenue concentration is a loaded gun: 78% from 3 (government) customers. One contract slips and the only "traction" evaporates — and government funding is exactly what a budget cycle or a CFIUS/export-control flag can freeze.
• The moat may be narrower than claimed. PennyLane is open-source and monetizes nothing; its value depends on rivals (IBM/IonQ/QuEra) supporting it — they can stop. The "10-100x overhead advantage" is a management claim, not a shipped result. Aurora is 12 logical qubits with sub-spec loss — a lab demo, not a moat.
• Most dangerous competitor bulls underrate: PsiQuantum. Better-funded, same silicon-photonics/measurement-based bet, industrial-scale roadmap, and private — so it doesn't carry XNDU's public-market dilution overhang while it scales. If PsiQuantum hits a manufacturability milestone first, XNDU's "only photonic FTQC" scarcity premium collapses. Add the hyperscalers (Google/IBM/Microsoft) who can cross-subsidize indefinitely.
• Capital-allocation/governance: dual-class entrenches a scientist-founder with no commercial-scaling track record; FPI status strips Nasdaq governance and Section-16 protections; unlimited share authorization + a 293.6M resale registration = structural dilution machinery. Insider selling is the sector norm.
• What must hold for today's ~$11 / ~$3.3B: that fabrication tames optical loss roughly on the 2029-30 timeline, that government money keeps flowing, that the float overhang gets absorbed without a death-spiral raise, and that the quantum-momentum bid persists. If growth/milestones disappoint by 20-30% (i.e. milestones slip ~2-3 years, the consensus FTQC base case), there is no earnings floor — the stock re-rates toward net cash (~$1/sh), a 70%+ drawdown, which the May ~65% crash already rehearsed. Single permanent-impairment scenario: photonic loss proves a harder wall than fabrication can climb this decade and a better-capitalized rival (PsiQuantum/hyperscaler) reaches utility-scale first — plausible enough to demand a venture-sized position, not a core one.

Lens 14 · Management Questions (ordered by information value)

1. Optical loss is the binding constraint — what is your current per-component photon loss, what's the trajectory to sub-1%, and what specifically in chip design/fab gets you there by when? (the whole thesis)
2. Runway vs. milestone: at current/forecast burn, what date does cash run out, and do you reach the next value-inflection milestone before you must raise?
3. Project OPTIMISM (C$390M): probability and timeline to a definitive, drawable agreement — and what job-creation/cost obligations come attached?
4. The 293.6M-share resale registration: how do you manage that overhang, and what are the lock-up expiry dates and any standstill commitments from OMERS/Bessemer/Radical?
5. Next equity raise: size, timing, structure (ATM vs. marketed vs. strategic) and the dilution holders should expect before 2030?
6. Customer concentration: name the revenue base — what % is US government, and how exposed are you to CFIUS/export-control or a budget freeze?
7. PennyLane monetization: when does the paid enterprise tier ship, what's the target attach/ARR, and what happens if IBM/IonQ deprecate support?
8. PsiQuantum: how do you win against a better-funded, private, same-modality competitor that has no public-market dilution constraint while it scales?
9. 2029-30 target (100k physical / 500 logical): what are the 2-3 hard intermediate gates, and what's your honest confidence interval on the date?
10. Sovereign system sales: any signed/near-term dedicated-system deals with a government or defense buyer, and at what value?
11. Manufacturing: dependence on Tower Semiconductor — is the foundry relationship sole-source, and does OPTIMISM bring fab in-house?
12. SIF loan repayment (from Apr 2028, revenue-based, 150% cap) — model the cash impact, and what triggers the change-of-control acceleration?
13. Capital-allocation discipline: with $272M, what's the spend plan, and what milestone would make you slow burn vs. accelerate?
14. Governance: why retain the dual-class/FPI Nasdaq exemptions, and will you commit to any sunset on the 10:1 voting structure?
15. Adjacent monetization (quantum sensing/internet, PIC licensing to defense): how material can near-term licensing revenue be while the FTQC machine is built?

---