Lightelligence

Phase A — Understand the business

Lens 1 · Company Overview

Lightelligence builds optoelectronic (hybrid optical-electronic) computing and optical-interconnect silicon for AI datacenters. Founded 2017 as an MIT spin-out by Yichen Shen (MIT physics PhD, 2016, advised in the Soljačić/Joannopoulos nanophotonics group; Frank Wilczek connection) with co-founders Marin Soljačić and John Joannopoulos; CTO/co-founder Huaiyu Meng. The company is dual-headquartered in Shanghai and Boston, but the entity that listed is the China operating company, Shanghai Xizhi Technology Co. Ltd. — a structurally important fact (see Lens 13).

The business model has two product lines, and the commercial center of gravity has shifted decisively from the first to the second:

1. Optical computing — the original thesis. PACE (Photonic Arithmetic Computing Engine), announced 2021, is a 64×64 optical matrix-multiplier: a silicon-photonic chip (>12,000 discrete photonic devices) flip-chip-bonded to a CMOS die, 1 GHz system clock, purpose-built to solve Ising-model / max-cut optimization problems — benchmarked at "100x faster than a typical GPU" and "25x faster than Toshiba's FPGA-based simulated-bifurcation machine," with later claims of "800x vs an NVIDIA GPU". Hummingbird (2023) is an optical network-on-chip (oNoC): a 64-core inference ASIC whose cores are wired by a programmable optical layer.
2. Optical interconnect — where the revenue actually is. Branded products include TIANSHU, Gazelle, positioned as CPO (co-packaged optics) / scale-up optical interconnect for AI clusters. The company explicitly frames its technology as "a complement to GPUs rather than a replacement, targeting bottlenecks in data transmission and matrix computation". A March 2026 ZTE partnership produced a "128-card optical node" claiming >90% latency reduction vs traditional electrical switching.

Contract structure / payment terms: not disclosed at the granularity the operating battery wants — n/a — private-prospectus-only, not in public sources. What is disclosed is brutal customer concentration: top-5 clients = 78.9% of revenue. Recurring vs one-off mix unknown.

The number that frames everything: 2025 revenue CNY 106.4M (~US$15M), +77% YoY, against a 2025 net loss of CNY 1.3B (~US$180M), which widened 83% YoY. This is a pre-commercial-scale company that happens to be publicly traded.

Lens 2 · Supply Chain

Mapping upstream inputs → Lightelligence → end customer, with named stakeholders where sourceable:

• Upstream — photonic/CMOS fabrication. Optical-compute/interconnect silicon needs a silicon-photonics-capable foundry plus a CMOS process for the electronics and advanced packaging (flip-chip / 2.5D for the photonic+electronic co-package). Lightelligence has not publicly named its foundry partners in the sources reviewed — n/a — not disclosed. Context: US rival Lightmatter is reported to fab photonics at ~90 nm with 12 nm electronics, indicating photonics nodes are mature/trailing-edge (a strategic advantage for a China name — see export-control point in Lens 13). The company stated its business is "not subject to US export control restrictions", consistent with trailing-edge photonics + domestic supply.
• Inputs — lasers, optical modulators, fiber attach, packaging substrates. Named modulator approaches across the field: microring modulators (MRM) at Ayar/Lightmatter, EAMs at Celestial AI; Lightelligence's compute path is MZI (Mach-Zehnder interferometer) interference-based. Specific component suppliers n/a — not disclosed.
• The company — designs the photonic + electronic dies, does the co-packaging integration, ships modules/boards/nodes.
• Downstream / channel & end customers — ZTE (named system partner, 128-card optical node, Mar 2026); broader customers are AI datacenters, cloud/AI providers, semiconductor leaders in China, the US and Singapore. No individual end-customer is publicly named beyond ZTE — and given top-5 = 78.9%, the unnamed concentration is the single most important unknown in the whole file.
• State-as-stakeholder (China-specific). Shanghai state-owned assets sit in the cap table and around the supply chain: Pudong Venture Capital, Zhangjiang Group, Shanghai State-owned Investment Corporation backed Lightelligence; the sector's survival is explicitly attributed to SOE "patient capital". China opened its first dedicated photonic-computing lab as a JV between Shanghai Jiao Tong University and Lightelligence — a quasi-industrial-policy supply-chain node.

This lens is honest about its gaps: the foundry/modulator/laser vendors that would make it specific are prospectus-only and were not ingested. Named where sourceable (ZTE, SJTU, the SOE syndicate); flagged where not.

Lens 3 · Competitive Advantages (moats)

What's genuinely defensible:

• First-mover + first-listed. Lightelligence is "the world's first AI photon-computing stock" and, per the IPO narrative, the first to claim large-scale deployment of hybrid optical-electronic computing. Within China it is the clear capital and maturity leader — "Lightelligence has attracted most of the funds in the track… the total financing of the remaining four [domestic] companies is even less than a fraction of its Series C". ~CNY 3.6B cumulative funding pre-IPO.
• IP + founder pedigree. MIT nanophotonics lineage; the foundational optical-neural-network work (Shen et al., Nature Photonics 2017) is seminal. Patent estate exists but is not quantified in public sources beyond Shen's ~10 personal US filings — n/a — full estate not sourced.
• State alignment. In China, SOE backing + a national photonics lab is a real moat against domestic capital-starvation and a partial shield from US export controls (light-based compute as a deliberate "workaround" to GPU curbs).

Where the moat is thin:

• Huawei owns the market. Lightelligence is China's #1 third-party optical-interconnect provider at just 1.4% share — Huawei holds 98.4%. The flattering "88.3% share" figure the company markets is a much narrower carve-out — the "independent Scale-up optical interconnect" segment — i.e. 88% of the ~1.6% that Huawei doesn't capture. Both figures are true and the gap between them is the whole bear case for the moat.
• Technical routes haven't converged. Four+ competing photonic approaches across the six Chinese startups alone (MZI / phase-change Crossbar / all-optical / metasurface). Betting on MZI is a bet, not a moat.
• Bargaining power: weak. With top-5 = 78.9% of a tiny revenue base, customers hold the leverage. SenseTime taking equity in rival Light Standard shows downstream AI buyers binding suppliers via the cap table — a sign the suppliers need the customers more than vice-versa.

Lens 4 · Segments

No segments.csv in the research layer (header-only) and the IPO prospectus segment breakout was not ingested → web-only, low granularity.

• By product line: revenue is overwhelmingly optical interconnect, not optical compute. The "China's #1 third-party optical-interconnect provider (1.4%)" framing is the revenue segment; PACE/Hummingbird optical-compute is still pre-revenue showcase. Exact split n/a — not disclosed.
• By geography: China-centric revenue (the listed entity is the China opco), with stated commercial presence in the US and Singapore; split n/a — not disclosed.
• Trend: total revenue CNY 38.2M (2023) → CNY 106.4M (2025), +77% in 2025. Accelerating off a microscopic base. The cause is the interconnect ramp (ZTE-type system wins), not optical compute. Direction up, magnitude trivial relative to the loss and the valuation.

---

Phase B — Measure performance — +private overlay: Funding & valuation trajectory / Cap table / Traction

Lens 5 → Funding & Valuation Trajectory (operating "Earnings Result" lens swapped per +private overlay)

Round history (all ``, pre-IPO figures unaudited):

• Founded 2017; ~US$247M raised over 4 rounds from 21 investors pre-IPO per aggregators; the China-press cumulative figure is ~CNY 3.6B. (The USD-aggregator and CNY-press totals don't perfectly reconcile — different round perimeters and FX timing; I surface both rather than pick one.)
• Series C: ~US$210M / CNY 1.5B (≈CNY 1.265B per 36kr), Sep 2025 — investors incl. China Mobile, CRHC Fund, Baidu Ventures, Matrix Partners China (MPCi), CICC Capital, Vertex Ventures China, China Merchants VC, FreesFund. Series C+: ~CNY 300M, 2025. The investor base is predominantly China-domestic / SOE / strategic — a tell about where the company's center of gravity moved.

The IPO (2026-04-28, HKEX main board, 1879.HK):

• Priced at top of range, HK$183.2/share; raised ~HK$2.4–2.5B net (~US$323M).
• First-day +383.6%, opened ~HK$880, closed HK$886 (~US$113) — the largest first-day gain for an HKEX main-board listing in nearly a decade.
• Debut valuation ~HK$81.5B (~US$10.4B).
• Demand: retail ~5,785x oversubscribed; institutional >50x; 20 cornerstones took ~65–71% of the offer, incl. Alibaba, GIC, Temasek, BlackRock, Lenovo, UBS AM, Baillie Gifford.
• Use of proceeds: ~70% to R&D over five years (optical interconnect + optical computing).

Income statement (the part that matters):

• 2025 revenue CNY 106.4M (~US$15M), +77% YoY.
• 2025 net loss CNY 1.3B (~US$180M), +83% YoY, driven by R&D + employee stock-option grants.
• Cumulative losses 2023–2025 ≈ HK$5.27B. Shareholders' equity reported negative post-historical-losses.

Lens 6 → Founder/Management Signal (operating "Earnings Calls / sentiment" lens swapped — no earnings-call history yet for a 7-week-old listing)

No transcripts (transcripts/ empty) and no post-IPO earnings call has occurred → substitute founder signalling from interviews/IPO commentary [all ``]:

• The defining management claim: founder Yichen Shen predicts optical chips rise from <1% to >30% of intelligent-computing-center chips within five years. This is the bull thesis in the founder's own words — and a falsifiable, very aggressive TAM-capture claim.
• Strategic framing has shifted from "optical replaces the GPU" (the 2021 PACE 100x-Ising narrative) to "optical complements the GPU, attacking the interconnect/data-movement bottleneck". That pivot mirrors the entire sector's repositioning toward CPO/interconnect (and mirrors NVIDIA's own $4B optics bet — Lens 8) and is, in my read, a maturation signal: the team followed the revenue.
• Tone in IPO-era coverage: confident, TAM-forward ("270.4B yuan market by 2031" ), leaning on the "first listed photonic stock" status. No earnings-call sentiment series exists to trend yet — re-run this lens after the first interim result.

Lens 7 → Cap Table & Comps (+private overlay: syndicate quality + peer marks)

Syndicate quality — the crossover/strategic tell: the cornerstone book is unusually blue-chip for a China small-cap — Alibaba, Tencent-adjacent strategics, GIC, Temasek, BlackRock, Baillie Gifford, Lenovo, UBS AM. Sovereign-wealth + global crossover presence is normally an IPO-quality signal; here it reads more as thematic AI-photonics FOMO + a scarce "first pure-play" asset than as fundamental validation (the same investors are buying a ~$15M-revenue company). Pre-IPO syndicate skewed China-domestic / SOE / strategic (China Mobile, CICC, Matrix China, Shanghai SOEs).

Peer comps table — optical-computing/photonics cohort. Multiples are `` with date or n/a. I will not fabricate a multiple.

Read: Lightelligence trades at an implied ~450–640x trailing sales — an order of magnitude beyond any rational interconnect or semicap comp, and richer (on a P/S basis) than its still-private US peers' last marks relative to their (undisclosed but larger-shipping) revenue. The multiple is a pure-optionality / scarcity multiple, not a fundamentals multiple. The mature public optics names (COHR, LITE) — which actually supply NVIDIA — are the realistic "what photonics-at-scale earns" anchors, and they trade at single-to-low-double-digit sales multiples.

Lens (added per +private) → Traction & Unit Economics

• Revenue run-rate: CNY ~106M/yr (2025), +77% YoY, off CNY 38.2M (2023). CAGR ~67% '23→'25 — high but tiny.
• Gross margin / unit economics: n/a — not disclosed in public sources (prospectus-only).
• Logos: ZTE (system partner); top-5 = 78.9% of revenue, otherwise unnamed.
• Burn: ~CNY 1.3B annual loss vs ~CNY 106M revenue → the company spends ~12x its revenue, mostly R&D + SBC. IPO proceeds (~US$323M, ~70% earmarked R&D) fund roughly ~1.5–2 years of burn at the current rate, before counting any operating-loss escalation — runway-to-scale, not runway-to-profit.

Lens 8 · Stock-Price Catalysts

Only ~7 weeks of trading history exist (listed 2026-04-28) — there is no 5-year tape, so this lens reads the available >5% moves and the category-level catalysts that will drive this name, all ``:

• 2026-04-28 — IPO debut: +383.6% in one session (HK$183.2 → ~HK$886), the largest HKEX main-board first-day gain in nearly a decade. What it reveals: the price is set by theme + scarcity + retail mania (5,785x oversubscribed), not fundamentals. A move that large on day one tells you the float is thin and sentiment-driven — exactly the profile that mean-reverts.
• Late-Apr → mid-May 2026 — sharp de-rate: −18.6% YTD, −17.5% in 7 days as of 2026-05-18, with the stock at ~US$72.67 by 04-Jun (well below the ~HK$886 debut close). What it reveals: the market is already repricing the scarcity premium; the dominant near-term driver is sentiment/valuation gravity, not news.
• Category catalysts that move the whole photonics complex (and will move 1879.HK): NVIDIA's $4B optics commitment to Coherent + Lumentum (2026-03-02) and OFC 2026's silicon-photonics/CPO dominance — these validated the category and are a tailwind; conversely any AI-hardware-capex wobble or NVIDIA optics-roadmap change is a headwind.
• Name-specific forward catalysts: the ZTE 128-card optical node (Mar-2026) was a positive product proof-point; the lock-up expiry (SOE/strategic supply) and the first post-IPO interim result are the next scheduled >5%-move candidates (see Position-seed catalysts).

Pattern read: for this name the market reacts to theme/liquidity and valuation gravity first, company news second — until a real earnings cadence exists. This is a momentum/positioning stock today, not a fundamentals-reaction stock.

---

Phase C — Judge people & books

Lens 9 · Management

• Track record: Yichen Shen (co-founder/CEO) is a credentialed scientist-founder — MIT physics PhD, foundational optical-neural-network research, MIT TR35 honoree. He has taken the company from a 2017 lab spin-out to the world's first listed photonic-computing company in nine years — a genuine execution achievement, regardless of the valuation. Marin Soljačić (co-founder, MIT professor, also a WiTricity founder) and John Joannopoulos lend deep scientific credibility; Huaiyu Meng is co-founder/CTO; team reportedly includes NVIDIA and Qualcomm alumni in commercial roles.
• Tenure & skin in the game: founder-led since inception; specific post-IPO insider ownership %s n/a — not disclosed (prospectus-only; insider-transactions.csv absent). The pre-IPO cap table is heavy with China SOEs and strategics, which dilutes pure founder control and aligns the company with state objectives.
• Capital-allocation history: the only allocation story so far is R&D-heavy burn funded by serial private rounds + the IPO — appropriate for deep-tech at this stage, but unproven as value-creating (no ROIC to judge; equity is negative). 70% of IPO proceeds to R&D continues the pattern.
• Red flags (management-level): the 2025 loss was inflated by employee stock-option grants — common pre-IPO, but worth watching for ongoing dilution. The US/China entity structure (listing the Shanghai opco, not the Boston parent) is the biggest governance question — see Lens 10/13.
• Archetype: scientist-founder, not professional operator. Implication: world-class on the physics and the narrative; the open question is whether this team can convert a 1.4% third-party share into durable, profitable, diversified revenue against Huawei and a fragmented startup field — a commercial-operator problem, not a physics problem.

Lens 10 · Forensic Red Flags

Grounded web-only; no audited 10-K, no SEC filings (no CIK — private/foreign), no us-GAAP. All flags /; the HKEX prospectus + future interims are the real forensic source and were not ingested.

• Valuation vs fundamentals — the headline flag. ~US$7–9B market cap on ~US$15M revenue and a ~US$180M loss = ~450–640x sales. Any forensic read starts and ends here: the equity is detached from the income statement.
• Negative shareholders' equity — accumulated losses (~HK$5.27B '23–'25 ) have eaten book equity; the IPO raise refills it, but the company is structurally loss-funded.
• Customer concentration: top-5 = 78.9% of revenue. A single logo loss would be a revenue event. Revenue-recognition quality on a few large, possibly project-based interconnect deals is unverifiable from public sources — a genuine ``-tier unknown, not a clean bill.
• SBC flattering/distorting the P&L: the loss "widened 83%… mainly due to high R&D and stock-option grants" — non-GAAP vs GAAP optics worth scrutinizing once interim statements publish.
• Revenue-figure imprecision across sources (HK$106.37M vs CNY 106.4M quoted near-interchangeably) — minor, but a reminder the public numbers are second-hand summaries of a Chinese-language prospectus, not the audited primary.
• Cash-conversion cycle: Smartkarma flagged an "elongated cash conversion cycle" — consistent with project-based hardware sales to a few large customers; pressures working capital.

Regulatory findings (required sub-section). Read regulatory/regulatory-findings.md (Stage-1 pre-fetch). SEC EDGAR EFTS (LR + AAER): 0 findings — Lightelligence has no CIK; it is not a US SEC filer, so no EDGAR enforcement search is possible. Non-SEC web search ("Lightelligence" (FTC OR DOJ OR FDA OR CFPB OR consent decree OR settlement OR fine OR penalty) enforcement) and the broader news sweep surfaced no material enforcement actions, litigation, or penalties as of 2026-06-17. There is no 10-K Item 3 to quote (no US filing). Macro/regulatory exposure that IS material but is not an "enforcement finding": (a) US-China tech tensions and the possibility of future US restrictions on Chinese photonics — CSIS explicitly frames silicon photonics as "an emerging front in US-China tech competition"; the company asserts it is "not subject to US export control restrictions" today; (b) HKEX/PRC listing-and-disclosure regime risk inherent to a China opco listed in Hong Kong. Summary: No material regulatory or legal enforcement findings — verified via SEC EDGAR EFTS (LR, AAER, 0 hits, no CIK), web search, and the absence of any US filing, as of 2026-06-17. Geopolitical/listing risk is real and is treated in Lens 13, not here.

---

Phase D — Project & stress-test — +private overlay on Lens 11

Lens 11 → IPO-readiness / Path-to-tradeable & rough scenario (IPO already occurred — repurposed to "path-to-investable-thesis")

The +private Lens-11 question ("does it reach a tradeable event?") is already answered — it listed on 2026-04-28. So this lens instead frames the path to an investable fundamental thesis and a rough scenario, all off inputs. No EPS model is built — the company is deeply loss-making and EPS is meaningless here; per the +private/clinical spirit, the scoreable question is revenue-scale-to-valuation, not EPS.

Founder's framing: optical chips <1% → >30% of compute-center silicon in 5 years; company-cited TAM CNY 270.4B by 2031. Take these as the bull ceiling, not a base case.

• Bull path: the interconnect/CPO bottleneck is real and accelerating (NVIDIA's $4B Coherent+Lumentum bet validates the category — Lens 8); China's AI-compute build-out + SOE patronage + export-control insulation give Lightelligence a protected domestic ramp; revenue compounds ~50–80%/yr off the tiny base toward several-hundred-CNY-M, the "88.3% of independent Scale-up" share broadens, and the multiple is partly grown into. Even so, growing into a ~640x multiple requires revenue to ~30–50x while margins turn — a multi-year, execution-perfect path.
• Base path: revenue keeps compounding fast in % but stays small in absolute terms (CNY low-hundreds-M), losses persist, the stock de-rates from scarcity-multiple toward something defensible (it's already −18% YTD, −17.5% in 7 days as of mid-May ). The equity is a volatile thematic trade, not yet a fundamentals investment.
• Bear path: Huawei + domestic-substitute competition caps third-party share; a top-5 customer churns; technical route (MZI) loses to a rival approach; lock-up expiries flood supply (the pre-IPO SOE/strategic holders are not long-term price-insensitive holders); the multiple compresses 50–80% toward listed-optics comps. Given the starting valuation, the asymmetric risk is down.

No Brier forecast logged (breadth-loop rule: skip forecast.ts create; and an EPS-style binary is inappropriate for a sub-scale loss-maker). The natural future-scoreable claim to log on a conviction pass would be a revenue or share-price level by FY-end, e.g. "1879.HK FY2026 revenue ≥ CNY 160M" or "1879.HK trades below its HK$183.2 IPO price within 12 months."

Lens 12 · Bull vs Bear

Bull case. Photonics is the next AI bottleneck after copper, power and memory — and the market has just anointed it: NVIDIA committed $4B across Coherent and Lumentum in March 2026 for optical interconnect, and OFC 2026 was dominated by silicon photonics / CPO / OCS. Lightelligence is the only listed pure-play in a category everyone now agrees matters, with the strongest balance of capital, IP pedigree, state backing and first-mover deployment in China, a real interconnect product winning system designs (ZTE), and a founder targeting a 30x share-of-silicon expansion. In a thematic, liquidity-rich AI tape, a scarce "first photonic stock" can stay expensive for a long time and compound the narrative. The contrarian-bull surprise: if optical compute (PACE-class) finds a real workload (optimization, certain inference matmul) ahead of schedule, the TAM re-rates beyond interconnect.

Bear case (the 2–3 that can permanently impair).

1. It's a 1.4%-share supplier in a market Huawei owns at 98.4% — the "88.3%" is 88% of the scraps. Durable share-capture against a vertically-integrated national champion is unproven.
2. The valuation is the risk. ~450–640x sales, negative equity, ~12x-revenue burn. Multiples this detached mean the stock can be permanently impaired even if the company succeeds modestly — there is no margin of safety at the entry price.
3. Geopolitics + structure. A China opco (Shanghai Xizhi) listed in Hong Kong, in a sector CSIS calls a US-China front; future export/capital-flow restrictions, lock-up-expiry selling by SOE/strategic holders, and HK-listing liquidity/discount dynamics are all live.

Pre-mortem (18 months out, thesis broke): the stock is down 60–80% from its debut. What happened — the scarcity multiple normalized (other photonic names listed or the AI-hardware trade cooled), revenue grew in % but disappointed in absolute terms, a concentrated customer slipped, and lock-up expiries released SOE/strategic supply into a thin float. The technology still works; the equity was simply priced for 2031 in 2026.

Are multiples too high? Unambiguously yes on any fundamental metric. The only frame in which they're defensible is pure-optionality on a category NVIDIA just validated — which is a momentum/theme bet, not a valuation.

Contrarian view (what the market is refusing to see): the market is treating "first listed photonic-computing stock" as if listing = winning. The listed entity is the China business competing under Huawei's shadow, not the MIT-pedigree global platform the brand evokes — and the revenue is interconnect, not the optical-compute moonshot that justifies the multiple. The story and the income statement are pointed at two different companies.

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

Dismantling the bull case:

• Where revenue is concentrated, and the shift risk: top-5 = 78.9%. Lose one Chinese systems/AI customer (or have Huawei undercut it) and the growth story breaks overnight. SenseTime taking equity in rival Light Standard shows customers can and do back alternatives.
• Why the moat is weaker than bulls think: 1.4% third-party share vs Huawei 98.4%; four+ unconverged technical routes among Chinese startups alone — MZI may simply lose. The "world-first listed" status is a financing event, not a technology moat.
• The most dangerous competitor bulls underestimate: not the US names (Lightmatter/Ayar/Celestial — walled off by geography/export lines) but Huawei (incumbent, vertically integrated, 98.4%) and the next-better-funded Chinese route (Light Standard, SenseTime-backed, 128×128 tape-out).
• Worst capital-allocation / structure red flags: negative shareholders' equity; SBC-inflated losses; a China-opco listing structure whose relationship to the Boston parent and US IP is opaque in public sources (n/a — not disclosed); SOE-heavy pre-IPO register whose lock-ups will expire.
• What must hold for today's price: revenue must compound ~50–80%/yr for years, third-party share must broaden well beyond Huawei's leftovers, gross margins must turn positive at scale, geopolitics must not bite, and the scarcity multiple must persist. Every one of those must hold.
• Valuation if growth disappoints 20–30%: at ~640x sales there is no support — a growth wobble plausibly halves-to-quarters the equity; the realistic floor is listed-optics-comp multiples (single-to-low-double-digit sales), i.e. −80%+ downside to the income statement, cushioned only by theme and float scarcity.
• Single scenario that permanently impairs: a US capital-markets / export action against Chinese photonics or a Huawei price/封装 response that caps third-party share — either re-rates the stock to a struggling-small-cap multiple permanently. Plausibility: moderate and rising given the CSIS framing.

Lens 14 · Management Questions (ordered by information value)

1. What is the exact revenue split between optical interconnect and optical computing, and what share of 2025's CNY 106M came from optical compute (PACE/Hummingbird-class) vs interconnect? (Decides whether the multiple's premise — optical compute — has any revenue.)
2. Who are the top-5 customers (78.9% of revenue), and what is the largest single customer's %? (The concentration is the whole risk.)
3. What is the legal and IP relationship between the listed Shanghai Xizhi entity and Lightelligence Inc (Boston) — who owns the foundational MIT-derived patents, and does the listco have global rights or only China?
4. Gross margin by product line today, and the path to positive company gross margin — at what revenue scale?
5. What are the lock-up expiry dates and share counts for SOE and strategic pre-IPO holders? (Supply overhang.)
6. Concretely, how does the founder's "<1% → >30% of compute-center silicon in 5 years" translate into your own revenue plan by year?
7. What is your defensible answer to Huawei's 98.4% share — where exactly do you win third-party sockets, and why can't Huawei foreclose them?
8. Which foundry/packaging partners fabricate your photonic and electronic dies, and how exposed are they to US export controls or domestic-capacity limits?
9. What is current cash, monthly burn, and the runway after the ~US$323M raise — and the revenue level at which burn inflects?
10. Why MZI over phase-change-Crossbar/EAM/metasurface — and what happens to your roadmap if MZI loses the architecture race?
11. How much of the 2025 loss was stock-based comp vs cash R&D, and what is the forward SBC/dilution trajectory?
12. What recurring vs one-off/project revenue mix sits inside the interconnect business, and how does that drive the "elongated cash-conversion cycle" analysts flagged?
13. What is the realistic timeline to a commercially shipping optical-compute product with paying customers (not benchmarks)?
14. How do you think about being a China-domiciled company in a sector that US policymakers (CSIS) name as a strategic front — what is your contingency if restrictions arrive?
15. What capital-allocation framework governs the ~70%-of-proceeds R&D spend — what milestones gate it, and what would make you return capital rather than spend it?

---