Accelink Technologies

Phase A — Understand the business

Company Overview

Accelink Technologies is China's flagship optical-component integrated device manufacturer (IDM) — it designs and makes optoelectronic chips, devices, modules, and subsystems spanning the full stack from indium-phosphide (InP) laser chips up to finished pluggable transceivers. Headquartered in Wuhan, it traces to the Ministry of Posts & Telecommunications' Solid-State Device Institute (est. 1976), was corporatized in the early 2000s, and listed on the Shenzhen exchange in August 2009 — the first Chinese optoelectronic-device supplier to go public. In December 2012 it absorbed Wuhan Telecommunication Devices Co. (WTD), consolidating China's two largest optical-device franchises.

Business model. Three overlapping product families:

1. Transmission — optical transceiver modules (10G→1.6T), optical fiber amplifiers (EDFAs), DWDM, optical passive/smart devices for telecom backbone, metro, and access networks. This is the legacy, China-Telecom/Huawei/ZTE-anchored book.
2. Data & datacom — high-speed datacenter transceivers (100G/400G/800G, 1.6T ramping), the AI-interconnect growth engine.
3. Components / chips — in-house lasers (incl. 25G EMLs, the only domestic company cited producing them), detectors, TOSA/ROSA sub-assemblies, and silicon-photonics integration.

Customers (Lens 2 expands): Huawei, ZTE, the three Chinese carriers, Nokia (described as placing "complete trust" in Accelink), and increasingly Western/Chinese hyperscalers via the datacom book; a 2024 Cisco co-development of a 1.6T OSFP-XD silicon-photonics transceiver signals direct hyperscaler-adjacent engagement.

Contract structure: standard component supply (PO-driven, not take-or-pay); no recurring/subscription revenue. Demand is cyclical and tied to carrier capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. cycles (historically lumpy) now overlaid with the AI datacenter build-out (currently supply-constrained, favorable).

Market position: ranked #4 global optical-component supplier for ~18 consecutive years; ~5.8% global share in 2024 per one source, with a forward claim of ~7.5% by end-2025 from a less-authoritative source.

Supply Chain

Upstream → Accelink → end customer, named stakeholders:

• Upstream inputs (the chokepoints):

• InP substrates — 2-3-4" indium-phosphide wafers. Global monthly supply ~400k units vs. 2026 demand of 700-800k → acute shortage. Substrate suppliers: Sumitomo Electric, AXT, JX Nippon (not Accelink-specific but the industry pool).

• High-speed DSP — the binding external dependency. A Broadcom / Marvell duopoly for PAM4 DSP; Accelink reportedly sources ~60% of high-end DSP from third parties. This is the single most important structural fact about the business: the highest-value silicon in an 800G/1.6T module is bought, not made.

• EML lasers — for the highest-speed lanes, externally constrained. Western leaders Lumentum, Coherent dominate merchant EML; Nvidia's ~$4B March-2026 lock-up of Lumentum + Coherent EML capacity pushes non-priority buyers' lead times beyond 2027. Accelink's in-house 25G EML helps at lower speeds but does not solve the 200G/lane problem.

• DSP-design IP, CMOS SerDes, optics packaging consumables.

• Accelink (the integrator + partial chip-maker): lasers/detectors → TOSA/ROSA → module assembly + test. Vertical integration is real but partial — strong in lower-speed lasers and passives, dependent at the bleeding edge (DSP, high-speed EML).

• Downstream buyers: Huawei, ZTE, Nokia (telecom OEMs); China Telecom / China Mobile / China Unicom (carriers); Cisco (co-dev partner); Chinese cloud / hyperscalers (datacom). Geographic spread: China, US, Canada, Germany, Italy, Japan, Korea, India.

• Geographic de-risking node: Phabritek Sdn Bhd (Penang, Malaysia) — wholly-owned, incorporated 2022, opened Nov 2023, RM100M initial investment, high-end optoelectronic modules/components. This is the export-control hedge — an offshore footprint outside mainland China.

Single-source dependencies / chokepoints: high-speed DSP (Broadcom/Marvell) and merchant EML (Lumentum/Coherent) are the genuine single-points-of-failure. In a shortage regime these gate volume, and Nvidia's lock-up means Chinese second-tier buyers are explicitly back-of-queue.

Competitive Advantages (moats)

What's genuinely defensible:

• IDM breadth — one of the few Chinese players covering chips → modules across both telecom and datacom. Owns 25G EML and TOSA/ROSA capability for silicon photonics — real process know-how that pure assemblers lack.
• State backing + scale — controlled by CICT (China Information & Communications Technology Group), giving privileged access to Chinese carrier demand, policy support, and patient capital (the Sept-2025 CNY 3.5B raise was state-led). Domestic-champion status in a sector Beijing treats as strategic.
• IP estate — >2,100 active IP entries with 250+ filings in 2025 (one source); 18-year top-10 ranking signals durable R&D.
• Telecom relationships — entrenched, multi-decade incumbency with Huawei/ZTE/carriers; Nokia's "complete trust."

Where the moat is thin — the central bear point: The IDM model leaves Accelink stuck between two winning archetypes. The specialists (InnoLight, Eoptolink) focus only on high-speed datacom Ethernet and win the AI wallet share + margins; the chip owners (Broadcom, Marvell, Lumentum, Coherent) own the DSP/EML IP and capture the excess profit. As one analysis puts it bluntly: "over the long term, excess profit stays not in assembly, but in chip IP". Module assembly has "relatively lower barriers to entry," so the moat at the module layer compresses as competition intensifies. Accelink's bargaining power is weak upstream (it needs Broadcom/Lumentum more than they need it, acutely so during the shortage) and moderate downstream (entrenched but in a contestable, multi-vendor market).

Verdict on moat: real but mis-located. The durable moat is in the chips it doesn't fully own; the layer it dominates (assembly + telecom passives) is the lower-value, more-contestable one.

Segments

our figures is header-only → no research-layer segment data. Web sources do not give a clean product/geography revenue split. What is sourceable:

• Core revenue remains optical communication modules & subsystems (telecom transmission historically the largest, datacom the fastest-growing).
• 800G high-speed products reportedly contributed >35% of total revenue by late 2025 — the clearest evidence of the datacom-mix shift and the engine of the +60% TTM growth.
• Geography: majority domestic (China), with export reach across US/EU/Japan/Korea/India.

Trend: decisively accelerating on the datacom side (800G mass shipping, 1.6T sampling), cause = AI datacenter interconnect demand. The legacy telecom book is the ballast — lower growth, lower margin, but the relationship base that funds the datacom push.

n/a — segment-level revenue/EBIT not sourced for a precise product or geographic breakout. This is a real gap a primary-filing read would close.

Phase B — Measure performance

Earnings Result

FY2024 (the cleanest sourced annual print):

• Revenue CNY 8.27B, +36.49% YoY (from CNY 6.06B FY2023).
• Net profit CNY ~656-661M, +~6.8% YoY.
• Gross margin ~22.5% (down from ~24.2% FY2023).
• Operating cash flow −CNY 641M (vs. +CNY 1.05B prior year) — a red flag: revenue +36% but OCF swung sharply negative, signaling a working-capital build (inventory/receivables) to chase the AI ramp. Flagged again in Lens 10.

The defining tension — revenue exploding, profit barely growing. FY2024 revenue +36% but net profit +7% → margins compressed as the business scaled. This is the opposite of the operating leverage InnoLight/Eoptolink show, and it is the quantitative heart of the bear case.

Acceleration into 2025:

• Q3 2025 (qtr ending Sep-30): revenue CNY 3.29B, +45% YoY; TTM revenue CNY 11.43B, +60.66% YoY.
• Q1 2025: gross margin 25.61%, net margin 6.76% — margins recovering modestly but net margin still single-digit.
• Q1 2026: "revenue and net profit surged… strong cash flow" — directionally strong but specific Q1 2026 margin/OCF figures not sourced (n/a).

Guidance: our figures header-only; no formal sourced guidance. Management commentary points to 1.6T mass production in Q2 2026 with 500K units/month 1.6T capacity.

Market reaction: the stock is up ~5x off its 52-week low (CNY 42.26 → 246.80 high) — the tape has priced a dramatic re-rating well ahead of the margin proof.

CONFLICT FLAG (revenue): one source cites FY2025 revenue CNY 8.6B, +15%. This is irreconcilable with the CNY 11.43B TTM/+60% figure from stockanalysis at Q3 2025 — an FY2025 of 8.6B would require a collapse in Q4. I weight the +60% TTM / ~CNY 11-12B FY2025 trajectory as far more credible (consistent across stockanalysis, prospeo, quarterly prints) and treat the Grokipedia 8.6B/+15%/7.5%-share cluster as unreliable. Resolve via the primary CSRC annual report before any position.

Earnings Calls (sentiment trend)

transcripts/ is empty → no research-layer transcripts. A-share issuers do not hold US-style quarterly calls; disclosure is via CSRC annual/interim reports + exchange Q&A. Sourceable management focus, by theme:

• 2024 → 2026 arc: narrative shifted from "telecom recovery + 400G" toward "AI datacenter interconnect, 800G mass shipment, 1.6T leadership, silicon photonics + CPO." The recurring phrases are now 800G mass-ship, 1.6T small-batch/sampling, 3nm DSP, silicon photonics, CPO optical engine, power efficiency.
• What they emphasize: domestic chip self-sufficiency, capacity expansion (Wuhan plant, 1.6T 500K/mo), automotive-grade optics as a new vector.
• Tone: unambiguously bullish/expansionary — consistent with a company in the middle of a demand boom. The risk in a uniformly bullish A-share management narrative during a retail-driven melt-up is that it provides little falsifiable signal; treat as low-information.

throughout — no transcript grounding. Sentiment read is directional, not rigorous.

Comps

Peer table — Chinese optical-module trio + Western chip/module leaders.

5-year average ROE: n/a for all names (not retrievable at confidence from these sources).

CRITICAL CONFLICT FLAG (Accelink net margin): the deepfundamental comp snapshot lists Accelink net margin 28.4% — but this directly contradicts the company's own FY2024 print (CNY 656M / CNY 8.27B = ~7.9%) and Q1 2025 net margin of 6.76%. A 28.4% net margin would imply ~CNY 2.3B+ net profit, which no other source supports. I treat ~7-8% as Accelink's true net margin and the 28.4% figure as almost certainly an aggregator error (possibly gross margin, or a mislabel). This conflict is load-bearing for the thesis — do not use 28.4%.

Read: on the corrected ~8% net margin, Accelink is the lowest-margin of the Chinese trio by a wide gap (InnoLight ~20%, Eoptolink ~33%) yet trades at a comparable-to-premium EV/Sales (~37 vs ~22-23). That is the valuation crux: a generalist with structurally thinner margins priced like — or richer than — the specialists out-earning it.

Stock-Price Catalysts

What moves the stock (>5% moves), pattern over recent years:

• The dominant driver is the AI-optical narrative, not company-specific earnings quality. The 52-week range (CNY 42.26 → 246.80, ~5x) tracks the 2024-2026 global AI-interconnect mania that lifted the entire Chinese optical-module complex (InnoLight, Eoptolink, Accelink, HG Genuine) in lockstep.
• Recent: +27% in one month, +~140% market-cap in the year to Oct-2025.
• Product milestones move it: 800G mass-shipment confirmations, 1.6T sample/mass-production news, CPO optical-engine announcements.
• Capital events: the Sept-2025 CNY 3.5B private placement (CICT-led) — shares reportedly +3% on placement-proceeds news.
• Sector/macro beta: US-China export-control headlines and Nvidia supply-chain news (EML lock-up) swing the whole group.

Pattern read: this is a high-beta thematic vehicle — it reacts to the sector AI story and product-roadmap headlines far more than to its own (mediocre) profitability. That makes it a momentum instrument; the same sensitivity cuts hard on the way down if the AI-capex narrative wobbles.

Phase C — Judge people & books

Management

• Chairman: Huang Zexuan, elected Nov 2025. A fresh-in-seat chair at the top of a parabolic move — limited track record to underwrite at this company in this role.
• Track record / archetype: this is a state-owned-enterprise / professional-manager structure, not a founder-led company. Leadership rotates within the CICT/Wuhan-institute system. Implication: execution is competent and policy-aligned, but capital allocation answers to state priorities (domestic self-sufficiency, strategic capacity) over per-share returns — minority A-share holders are not the primary principal.
• Skin in the game: our figures absent → no insider-ownership data sourced. Dominant holder is CICT (state); mutual-fund/institutional float beneath. Management personal ownership: n/a. This is the SOE governance discount in action.
• Capital-allocation history: aggressively reinvesting — CNY 3.5B private placement (242M new shares, 35 investors incl. CICT) Sept 2025; CNY 300M automotive-grade capex; new Wuhan high-end plant (158,000 m²); Phabritek Malaysia. The pattern is growth-capex + equity issuance, i.e. dilutive expansion to chase volume. With OCF negative in FY2024, the build is being funded by external capital, not internally generated cash — a returns red flag even if strategically rational.
• Red flags: the 242M-share placement is meaningful dilution; negative operating cash flow during a boom; SOE related-party dynamics with CICT (controlling shareholder also the capital provider) — not improper, but a minority-alignment concern.

Forensic Red Flags

Forensic lens, web-only — every figure secondary; primary CSRC filings needed to confirm.

• Cash flow vs. earnings divergence — the headline flag. FY2024 OCF −CNY 641M against +CNY 656M net profit. Earnings positive, operating cash deeply negative → quality-of-earnings concern. Most plausibly a working-capital build (inventory + receivables) to pre-stock for the AI ramp; but it must be confirmed it isn't aggressive revenue recognition or receivables stretch. Receivables/inventory vs. revenue trend: n/a (the exact disaggregation a 10-K-equivalent would give).
• Margin compression while scaling — gross margin fell 24.2%→22.5% FY23→FY24 as revenue grew 36%; recovered toward 25.6% in Q1 2025. Consistent with input-cost (DSP/EML) pressure rather than fraud, but worth monitoring.
• Dilution — 242M new shares (Sept 2025) is real per-share leakage; net-debt/leverage post-raise: n/a.
• SBC / goodwill / intangibles — n/a (WTD merger goodwill likely on the books; not quantified here).
• Segment reporting — thin public disclosure; the 800G/1.6T mix is described qualitatively, not audited in these sources.

Regulatory findings (required sub-section) — from regulatory/regulatory-findings.md (Stage-1 pre-fetch) + web:

• SEC (EDGAR LR + AAER): none possible — Accelink has no CIK and is not an SEC filer. Total SEC findings: 0.
• Item 3 Legal Proceedings: n/a — no 10-KForm 10-KA company’s audited annual report to the US regulator. The most complete thing it publishes. exists (A-share issuer). The equivalent disclosure lives in the CSRC annual report, not fetched here.
• Non-SEC enforcement (web search): the targeted search ("Accelink" FTC/DOJ/FDA/CFPB/consent decree/settlement/fine) surfaced no material enforcement actions against Accelink specifically. The relevant exposure is structural, not adjudicated: US export-control / Entity-List risk for Chinese optical-component makers. As of these sources, Accelink is not confirmed on the US Entity List, but the sector faces escalating BIS controls (the Sept-2025 rule extending controls to ≥50%-owned subsidiaries of listed entities is the kind of mechanism that could ensnare a state-controlled firm) — hence the Phabritek/Malaysia hedge. Treat Entity-List addition as a live tail risk.
• Summary: No material adjudicated regulatory or legal findings — verified via SEC EDGAR EFTS (no CIK → not applicable) and web search as of 2026-06-18. The dominant regulatory risk is prospective US export control, not any existing action.

Phase D — Project & stress-test

Forward Projection

Build-up is — no research-layer financials, no sourced consensus EPS. Anchored on the sourced FY2024 actual + the +60% TTM trajectory. Currency CNY; per-share figures heavily caveated because the 242M-share Sept-2025 placement changed the share count and a precise post-raise share count is n/a (so EPS estimates carry wide error).

Base inputs: FY2024 revenue CNY 8.27B, net profit ~CNY 656M (~7.9% net). TTM (Q3-25) revenue CNY 11.43B/+60%.

• Bear (margin stays thin, AI capex cools): FY2025 rev ~CNY 11.5B, net margin ~7% → net profit ~CNY 0.8B. FY2026 rev +15% to ~CNY 13.2B, margin flat 7% → ~CNY 0.92B. FY2027 rev +5% → ~CNY 0.97B. Margins never inflect; growth decelerates as 1.6T commoditizes.
• Base (datacom mix lifts margin modestly): FY2025 rev ~CNY 11.8B, net margin ~8% → ~CNY 0.94B. FY2026 rev +30% to ~CNY 15.3B, margin 9% (800G/1.6T scale) → ~CNY 1.38B. FY2027 rev +20% to ~CNY 18.4B, margin 9.5% → ~CNY 1.75B. Mix-shift to high-speed lifts margin toward low-teens but never to specialist levels.
• Bull (margin re-rates toward specialists + share gains): FY2025 ~CNY 12B @ 9% → ~CNY 1.08B. FY2026 +45% to ~CNY 17.4B @ 12% → ~CNY 2.09B. FY2027 +30% to ~CNY 22.6B @ 14% → ~CNY 3.16B. Requires in-house DSP/EML progress to capture the value it currently buys.

Reality check vs. price: at ~CNY 51B market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth., even the bull FY2027 ~CNY 3.16B net profit ⇒ ~16x P/E three years out; the base FY2027 ~CNY 1.75B ⇒ ~29x P/E 3 years out. On the ~50x trailing P/E today, the multiple only makes sense if you underwrite the bull margin re-rate — i.e. you must bet Accelink becomes a chip-owner, not stay an assembler. That is precisely the bet the bear case says is structurally hard.

Brier forecast — NOT logged. Per --watchlist unattended rules, I do not run our model create. (If promoted to a position, the scoreable base call would be: "002281.SZ FY2026 net profit ≥ CNY 1.35B," p≈0.5, resolves ~2027-04-30.)

Bull vs Bear

Bull case. Accelink is a genuine top-4 global optical IDM with the broadest Chinese stack — chips, modules, telecom + datacom — at the dead center of the largest interconnect demand wave in history. 800G is mass-shipping (>35% of revenue), 1.6T mass production lands Q2 2026 at 500K units/month, and the EML/InP shortage is a seller's market that rewards anyone with capacity. State backing (CICT) funds aggressive capacity at below-market cost of capital, the Malaysia footprint hedges export controls, and the IDM model means margin upside if in-house DSP/silicon-photonics/CPO progress lets it claw value back from Broadcom/Lumentum. Secular tailwind (AI datacenter capex) + domestic-champion policy support + a credible 1.6T/CPO roadmap = a multi-year volume story with optionality on margin.

Bear case (permanent-impairment risks).

1. Value capture is structurally elsewhere. The excess profit sits in DSP (Broadcom/Marvell) and EML (Lumentum/Coherent) IP that Accelink buys. Module assembly has low entry barriers; as 1.6T commoditizes and second-tier capacity (incl. Accelink's own) floods in, module pricing and margins compress. Accelink's ~8% net margin vs. specialists' 20-33% is not a temporary gap — it's the structural signature of the wrong layer.
2. Stuck-in-the-middle. Neither the focused specialist (InnoLight/Eoptolink win the Nvidia datacom wallet + margins) nor a pure chip owner. The history of the optical industry is unkind to generalists between those poles.
3. Cash-flow / dilution quality. Negative FY2024 operating cash flow during a boom, funded by a 242M-share dilutive raise — growth bought with paper, not earned with cash.
4. Expectations baked in: ~50x trailing P/E and ~37x EV/Sales after a 5x run, on single-digit margins, with the sell-side average 12-month target (~CNY 101) implying ~55% downside vs. ~CNY 225 spot. The price already discounts a margin re-rate the business has never delivered.

Pre-mortem (18 months out, thesis broke): AI-capex growth decelerated / Chinese hyperscaler optical orders normalized; 1.6T arrived in volume from five vendors at once and ASPs/margins fell faster than units rose; Nvidia's EML lock-up kept Accelink supply-gated out of the highest-margin Western datacom sockets; the stock de-rated from ~50x toward a margin-appropriate ~20x as the retail-driven A-share momentum unwound — a >50% drawdown with the business still "fine" operationally. Or: a US Entity-List addition impaired its export and Western-customer access overnight.

Are multiples too high? On the corrected ~8% net margin: yes, demanding. ~50x P/E is a specialist multiple on a generalist margin structure. Justifiable only on the bull margin-re-rate path.

Contrarian view (what the market refuses to see): the market is pricing all three Chinese optical names as one AI basket, ignoring that Accelink earns ~8% where InnoLight/Eoptolink earn 20-33% — the tape rewards the narrative (top-4 IDM, 1.6T, CPO) and overlooks that Accelink sits structurally at the low-margin layer. The kicker: the consensus target (−55%) already disagrees with the price — the disconnect between the momentum bid and the analyst base case is itself the signal.

Devil's Advocate (short-seller)

Dismantling the bull case.

• Where it structurally breaks: Accelink makes money assembling modules whose two most valuable parts (DSP, high-speed EML) it must buy from supply-constrained Western duopolists who are explicitly prioritizing its competitors' end-customer (Nvidia). In the upcycle that caps its volume; in the downcycle, with no IP moat at the module layer, price competition torches the ~8% margin toward break-even.
• Revenue concentration / shift: heavy domestic-China + telecom-carrier dependence; the datacom growth that's re-rating the stock is the part most exposed to (a) Chinese-hyperscaler order normalization and (b) US export controls cutting off Western datacom sockets.
• Why the moat is weaker than bulls think: the "IDM" badge masks that the in-house chips are lower-speed (25G EML); at 200G/lane it's as dependent as any assembler. The moat is in telecom passives — a slow, low-margin, contestable business.
• Most dangerous competitor bulls underestimate: not InnoLight (already winning) but the chip owners themselves — Broadcom/Marvell/Lumentum/Coherent — who capture the profit regardless of who assembles, and the prospect of Nvidia/hyperscalers designing modules around merchant silicon that routes around Chinese assemblers entirely (CPO disintermediates the pluggable assembler).
• Worst capital-allocation moves: serial dilution (242M shares) + growth capex funded by negative-OCF external capital; SOE governance where the controlling shareholder (CICT) is also the capital provider — minority holders second in line.
• What must hold for today's price: continued AI-capex acceleration and a margin re-rate from ~8% toward the teens and no export-control hit and the A-share momentum bid persisting. Several independent legs, each fragile.
• If growth disappoints 20-30%: on a ~50x P/E single-digit-margin name, a growth miss compresses both E and the multiple — a 50-70% drawdown is easily modeled (toward the ~CNY 101 analyst target or below).
• Single scenario that permanently impairs: US Entity-List designation cutting access to Western DSP/EML inputs and Western customers simultaneously — plausibility: moderate and rising given the sector's strategic profile and the Sept-2025 subsidiary-controls rule.

Management Questions (ordered by information value)

1. What is the in-house vs. merchant content of your 800G and 1.6T modules by value — specifically, what share of the DSP and high-speed EML is sourced from Broadcom/Marvell/Lumentum/Coherent, and what is your concrete roadmap and timeline to internalize it?
2. FY2024 operating cash flow was negative CNY ~641M against positive net profit — walk through the working-capital drivers, and when does OCF turn durably positive?
3. Your net margin (~8%) is a fraction of InnoLight's (~20%) and Eoptolink's (~33%). What specifically gets you to the mid-teens, and on what timeline — or is single-digit the steady state of the IDM model?
4. How exposed are you to US export controls / Entity-List risk, and exactly what does the Phabritek (Malaysia) footprint insulate — what share of high-end capacity can you run fully outside mainland China?
5. After the 242M-share placement, what is the share count, and what is your framework for future equity issuance vs. self-funding growth?
6. How much of your datacom revenue is Chinese hyperscalers vs. Western customers, and how do you win Western AI datacom sockets when Nvidia has locked up your suppliers' EML for its preferred vendors?
7. What is your 1.6T ASP and gross-margin assumption as five-plus vendors ramp into 2026-2027 — how much price erosion is in your plan?
8. On CPO: does co-packaged optics disintermediate you as a pluggable-module assembler, or do you capture the optical-engine value — and how?
9. What is the CICT relationship operationally — to what degree are capacity, capital, and customer decisions set by state priority vs. per-share return?
10. What is your silicon-photonics in-house capability at 200G/lane, and does it reduce EML dependence at the bleeding edge or only at lower speeds?
11. What are your 5-year ROE/ROIC figures, and how do you judge the returns on the Wuhan plant + automotive capex?
12. What is the telecom vs. datacom revenue and margin split, and where is each heading?
13. What is your inventory and receivables position relative to revenue, and how do you manage the risk of an AI-capex air-pocket leaving you over-stocked?
14. What share of your IP estate is bleeding-edge (DSP/SiPh/CPO) vs. legacy passives/telecom?
15. Who is the competitor or technology shift you fear most, and why?