Coherent Corp. (NYSE: COHR) — The Best-Integrated Player in the Cycle, Priced as if There Is No Cycle

An independent equity research note Report date: 2026-06-10 Company: Coherent Corp. (formerly II-VI Incorporated) · NYSE: COHR · CIK 0000820318 Sector: Information Technology — Electronic Equipment, Instruments & Components (photonics / optical networking / lasers) Fiscal year end: June 30 · HQ: Saxonburg, Pennsylvania · ~30,200 employees

Price reference: ~$356/share (2026-06-09); ~195.5M shares out; market cap ~$70B; net debt ~$1.6B; enterprise value ~$71B. All financial figures reconcile to SEC filings (10-K FY2025, 10-Q Q3 FY2026) and the Q3 FY2026 earnings call unless noted. Management commentary is treated as hypothesis, validated against filings and external evidence.

⚡ Claude’s Take

This block is the author’s own independent opinion and general information only — it is not investment advice. The analysis that follows is deliberately position-free and carries no price target anywhere except inside this block.

Verdict: HOLD the business, AVOID the entry. Great house on a frothy street — accumulate only on a real correction (roughly the $210–$255 zone, ~30–34x FY27 non-GAAP EPS / ~7–8x forward sales). Not a short. Conviction: medium.

Coherent is a genuinely good business that the market has, for once, correctly identified — and then bid to a 99th-percentile valuation against its own ten-year history. The thing the bulls own is real: a true cost-and-scale advantage in indium phosphide (the binding chokepoint of the entire AI-optics supply chain), uniquely broad co-packaged-optics content, an NVIDIA partnership that put $2B of equity on the balance sheet and a supply agreement on the books through the end of the decade, accelerating revenue (+27% pro forma YoY and guiding faster into FY27), gross margin climbing 39.6% → a >42% target, and a balance sheet de-levered from 2.1x to 0.5x in a year. On forward sales (~10x) it is literally the cheapest way to own the AI-optical thesis in its cohort — half of Lumentum’s multiple. That is the bull case in one paragraph, and it is not silly.

The reason I will not chase it here is that every one of those strengths is a cyclically-extended condition, not a structural constant, and the price already underwrites their permanence. The scarcity rents Coherent collects on indium phosphide exist because supply is short — and Coherent itself, Lumentum, Greensboro, and a wave of Chinese EML startups are all quadrupling capacity into the same demand boom: the textbook Marathon late-capital-cycle setup that has ended every prior optical up-cycle (JDSU’s ~95% drawdown is the genre’s tombstone). The module/transceiver layer Coherent wraps around its chips is commoditizing and Chinese-led. CPO, the headline growth vector, partly cannibalizes the pluggable-transceiver franchise that is Coherent’s profit engine today. Customer power is drifting toward a buyer base of three or four hyperscalers plus NVIDIA. And the incentive plan pays management on revenue, EBITDA and relative TSR — not one dollar of pay is tied to return on the capital being shoveled into fabs. None of this breaks the thesis; all of it argues that you are paid to wait for the cycle, not to pay up for it. The single thing that flips me bullish: durable evidence that the gross-margin/return profile holds through an InP capacity normalization (i.e., the rents are structural, not scarcity). The single thing that flips me bearish: the first hyperscaler capex air-pocket or a 1.6T price reset, which a $70B equity priced for permanent acceleration is not discounting. Tag: the best-integrated player in the cycle, priced as if there is no cycle.

1. Executive Summary

Coherent Corp. is a vertically-integrated photonics company — the merged entity of legacy II-VI Incorporated and the legacy Coherent Inc. (acquired July 2022 for ~$7B, the company taking the Coherent name in September 2022). It designs and manufactures the full optical stack from compound-semiconductor materials (indium phosphide, gallium arsenide, silicon carbide) through lasers, modulators and detectors, to finished optical transceivers, optical-circuit switches (OCS), co-packaged optics (CPO) and industrial laser systems. Two reportable segments now organize the business: Datacenter & Communications (~75% of revenue), the AI-driven growth engine, and Industrial (~25%), a higher-margin, partly-recurring legacy base currently being pruned.

The investment question is not whether this is a good business — it is whether a good, cyclical business is worth a near-record multiple. Revenue grew from $4.71B (FY2024 trough) to $5.81B (FY2025, +23%) and is annualizing toward ~$7B in FY2026, with Q3 FY2026 revenue a record $1.81B (+21% GAAP / +27% pro forma YoY). GAAP profitability has inflected violently — net income went from –$156M (FY2024) and +$49M (FY2025) to +$564M in just the first nine months of FY2026 — as merger-amortization drag fades and operating leverage arrives. Non-GAAP gross margin reached 39.6% (heading to a stated >42% target) and non-GAAP EPS grew 55% YoY. The balance sheet, a genuine prior worry (the ~$7B Coherent deal was financed with debt plus $2.15B of mandatory preferred), is transformed: net leverage fell to 0.5x after the Series B preferred converted to common, NVIDIA invested $2.0B of fresh equity, and the company paid down ~$493M of debt in nine months.

The core competitive asset is vertical integration in indium phosphide (InP) — the material chokepoint for every high-speed optical link, since even silicon-photonics transceivers require an InP laser. Coherent’s first-to-volume 6-inch InP platform yields >4x the devices per wafer at less than half the per-device cost of the industry-standard 3-inch, and is the engine of both its capacity expansion and its margin climb. This is a real Greenwald-style cost + economies-of-scale advantage, augmented by an intangible/process moat (proprietary materials such as Thermadite, in-house garnet for isolators, two decades of InP/VCSEL know-how). It is the reason NVIDIA chose Coherent for co-packaged optics.

But the advantage is durable yet narrow, and it sits inside a structurally bifurcated industry. The component/chip tier Coherent dominates is concentrated and capacity-constrained today; the module-assembly tier it also occupies is commoditizing and Chinese-led (InnoLight and Eoptolink took ~60% of NVIDIA’s incremental 800G volume). The scarcity rents driving today’s pricing and margin are a cyclical feature that the entire industry — Coherent included — is racing to compete away by roughly quadrupling InP capacity over two years. The honest characterization: a best-in-class operator with a real but cyclically-extended moat, riding a single secular thesis (optics displacing copper in AI data centers) that is genuine but unproven through a downturn, at a valuation (P/S in the 99th percentile of its own history) that prices the up-cycle as permanent. No recommendation or price target appears below this summary; the analysis frames valuation as embedded expectations and scenarios.

2. Business Overview

2.1 What Coherent does

Coherent is a photonics company whose defining structural characteristic is vertical integration: it owns the value chain from compound-semiconductor materials to finished optical systems. Few competitors span this range; most occupy one or two layers. The in-house stack, as detailed at the May-2025 Investor Day and the March-2026 OFC briefing, comprises:

• Compound-semiconductor materials and devices: indium phosphide (InP) substrates and epitaxy; InP electro-absorption modulated lasers (EMLs) and directly-modulated lasers to 400G/lane; high-power InP continuous-wave (CW) lasers (100mW today, 400mW for CPO); high-speed photodiodes; gallium-arsenide VCSELs (Coherent is among the largest VCSEL makers globally, having shipped >1 billion for datacom); and silicon-photonics PICs (designed in-house, fabricated externally).
• Passive and enabling optics (under-appreciated): optical isolators (Coherent grows its own garnet crystal), mux/demux filters, microlens arrays, polarization-maintaining fiber, fiber-attach units, and thermoelectric coolers — i.e. nearly every ingredient of a CPO module is sourced internally.
• Finished products: 800G and 1.6T pluggable transceivers; ZR/ZR+ coherent transceivers for data-center interconnect (DCI); transport components and systems (EDFAs, ROADMs, pump lasers, multi-rail amplifiers); liquid-crystal-based Optical Circuit Switches; emerging CPO/NPO modules; excimer/solid-state/CO₂ industrial lasers; and engineered materials (thermoelectrics, silicon carbide, advanced ceramics).

2.2 How it makes money — segments and mix

Within Datacenter & Communications, the data center sub-business (transceivers, OCS, CPO) grew +37% YoY and +13% sequentially in Q3 FY2026; the communications sub-business (DCI/ZR+, transport, telecom) grew +60% YoY and +16% sequentially. The Industrial segment spans semiconductor-capital-equipment lasers, display/precision manufacturing, and engineered materials; it declined modestly in Q3 but bookings (notably semi-cap) are improving.

This two-segment structure replaced the prior three-segment Networking / Materials / Lasers reporting and reflects a deliberate reorientation toward the datacenter franchise, accompanied by active portfolio pruning (detailed below): the Aerospace & Defense unit was sold (Q1 FY26, ~$400M, +$115M gain), the Munich materials-processing tool division was divested (closed Jan 2026, at a loss), and a ~25% stake in the silicon-carbide business was sold to DENSO and Mitsubishi Electric in 2023 for $1.0B (Coherent retains ~75%).

2.3 Recurring vs. cyclical revenue

This is not a recurring-revenue business. The large majority of revenue is non-recurring hardware sold into hyperscaler and network-equipment-maker capital budgets, which is inherently capex-cyclical. The genuinely recurring slice — service and replacement parts on the installed base of industrial lasers — is roughly 25% of the Industrial segment, i.e. only ~6% of total company revenue. There is no subscription or SaaS revenue. The recently-signed long-term agreements (LTAs) with NVIDIA and others introduce multi-year demand visibility — including customer upfront capacity contributions and minimum-volume commitments — which is materially better than spot selling, but is still capex-cycle-exposed rather than annuity-like. Fact: management states orders now extend into calendar 2028 and customer LTAs reach the end of the decade.

Verdict (Business model): A capital-intensive, cyclical hardware/components business that has integrated deeply enough to capture the scarce, high-value part of the optical value chain (InP devices), but which remains exposed to the AI-capex cycle and carries no annuity revenue. The integration makes it structurally superior to a pure module assembler; it does not make it an asset-light compounder. Capex rose to $290M in Q3 FY2026 alone (from $112M a year earlier) and is guided higher — a direct measure of how much capital this growth consumes.

3. Industry Dynamics

3.1 The secular driver — AI converting the data center from electrical to optical

The structural tailwind is real and physics-based: as AI clusters scale, the volume of data moved between accelerators, racks and buildings is exploding, and electrical interconnect cannot keep up on reach or energy efficiency. Optics is displacing copper progressively — scale-out (between racks/rows) and scale-across (between buildings, i.e. DCI) are already essentially all-optical; the next frontier is scale-up (in-rack, GPU-to-GPU), which is ~100% copper today and is the target of co-packaged optics. Management sizes the Datacenter & Communications opportunity at roughly $44B by 2030 (~$32B data center + ~$12B communications). The transceiver speed cadence is both compressing (800G→1.6T→3.2T arriving faster than prior nodes) and overlapping (800G still growing as 1.6T ramps), which favors a broad-portfolio technology leader that can serve multiple data rates simultaneously.

3.2 The binding constraint — indium phosphide

Indium phosphide is the industry’s chokepoint. It is the material from which high-speed EMLs and CW lasers are made, and crucially even a silicon-photonics transceiver requires an InP CW laser as its light source. InP supply has been the pacing item for the whole industry for several quarters. This single fact explains much of the current profit pool: scarce InP capacity earns scarcity rents, which flow disproportionately to the few players with scale InP fabs. Coherent’s response — and the industry’s — is a furious capacity build, discussed below.

3.3 The bifurcation — two industries in one value chain

The optical value chain splits into two structurally different industries, a distinction central to handicapping Coherent:

1. Module / transceiver assembly = structurally bad. Barriers are low; the activity is labor-, scale- and price-driven; and it is increasingly dominated by Chinese manufacturers. InnoLight and Eoptolink together took roughly 60% of NVIDIA’s incremental 800G transceiver orders for the 2025 cycle; Chinese optical-module vendors reportedly occupy seven of the industry’s top ten seats. This tier commoditizes, and its historical pattern is brutal price erosion (the canonical JDSU 2000–2002 collapse of ~95% is the cautionary tale of the genre).
2. InP device / laser-chip components = structurally good (for now). Concentrated, capacity-constrained, high-margin — this is where the profit pool actually sits. The revealing question the peer (Lumentum) work poses: Chinese modules own the seats, so why are the chip companies making the money? Because the scarce, hard-to-make InP devices inside those modules are where the rents accrue.

Coherent’s vertical integration lets it straddle both tiers — collecting the component rents and the module revenue, and capturing the inter-segment margin internally. That straddle is its single most important structural advantage over a pure assembler (Fabrinet) or a sub-scale module vendor.

3.4 The capital cycle — a Marathon late-cycle warning

Applying the Marathon/Chancellor capital-cycle lens (supply-side analysis: high returns attract capital, which competes those returns away), the optical-component industry is flashing a textbook late-cycle oversupply precondition. Everyone is adding InP and transceiver capacity into a demand boom that is itself being funded by hyperscaler and NVIDIA pre-commitments:

• Coherent is doubling InP capacity by end-CY2026 (now expected a quarter early) and doubling it again by end-CY2027 — a 4x build in two years, almost entirely on 6-inch wafers.
• Lumentum is expanding San Jose InP and bringing a fifth fab (Greensboro) online by 2028.
• Chinese EML startups (e.g. Zetta, Changguang Huaxin) are climbing the speed curve.
• Hyperscalers are actively funding and qualifying second sources to de-risk supply.

When the binding constraint (InP) is relieved across the whole industry roughly simultaneously, the “pricing optimization” tailwind Coherent currently enjoys reverses. Scarcity rents are real today but are a cyclically-extended condition, not a structural property of the industry.

3.5 Regulation and geopolitics

The sector is not rate- or reimbursement-regulated, but it carries material geopolitical supply risk: InP substrate feedstock is heavily China-sourced (an analyst at OFC cited ~70%; management claims five-plus diversified suppliers and three-to-five-year agreements). Export controls — in either direction — could disrupt either feedstock or end-market access. China is simultaneously a key supply input, a key competitor (modules), and a meaningful end market.

Verdict (Industry): Structurally bifurcated — bad at the module-assembly tier, good at the InP-component tier — with a Marathon late-cycle oversupply risk building across both, and live geopolitical supply exposure. Coherent sits in the better half by virtue of integration, but the entire complex is one capacity glut or demand air-pocket away from the optical industry’s historical boom-bust pattern. This is an attractive place to be mid-cycle and a dangerous one to over-pay for late-cycle.

4. Competitive Position

4.1 The competitive map

4.2 Naming the moat (Greenwald taxonomy)

Coherent’s moat is a cost advantage plus economies of scale in indium phosphide and the surrounding photonic-device stack — not network effects, not switching costs, not brand. Two mechanisms:

• Proprietary process technology: first-to-volume 6-inch InP (>4x devices/wafer at <half the per-device cost, with yields already exceeding the 3-inch lines), differential EMLs, in-house garnet for isolators, two decades of InP/VCSEL manufacturing know-how, and in-house OCS control software. Management frames the materials edge (Thermadite, thermoelectrics) as “a special recipe, difficult to reverse-engineer” — a classic intangible/process moat.
• Breadth-as-integration: for CPO, Coherent is the only supplier that can bring the entire module portfolio internally — laser, external-laser-source module, isolator, thermoelectric cooler, fiber-attach unit, microlens array, PM fiber, PIC and VCSEL. Customers avoid stitching together a fragile multi-vendor supply chain. This is the strongest defensible claim, and is explicitly why NVIDIA selected Coherent.

4.3 Pressure-testing the moat — be skeptical

A moat is only real if a financial outcome would deteriorate without it. Here it passes partially:

• Where it IS a moat: the InP cost/scale advantage shows up directly in gross-margin expansion (39.6% in Q3 FY26, +105bps YoY, driven primarily by Datacenter & Communications via 6-inch yield and pricing). If Coherent lost InP self-supply, it would have to pay the scarcity rents it currently collects — a measurable economic loss. The CPO breadth advantage is reflected in winning the NVIDIA franchise that pure-play laser or module vendors did not.
• Where it is NOT a moat (or is eroding): the module/transceiver-assembly layer has no durable advantage — Chinese players own it, it commoditizes, and Coherent’s margin there depends on the chip it stuffs inside, not the assembly. The InP scarcity rent is cyclical, not structural — it exists because supply is short, and the entire industry (Coherent included) is racing to relieve it. Switching costs are weak: these are specified components sold to sophisticated buyers who pre-allocate orders and deliberately fund second sources. Customer concentration (NVIDIA plus a handful of hyperscalers) means pricing power drifts toward the buyer over time. And the record backlog management cites is, like all optical backlogs in an up-cycle, partly inflated by double-ordering — “not indicative,” in the industry’s own historical phrase.

Verdict (Competitive position): A durable but NARROW advantage. Coherent has a genuine Greenwald cost/scale-plus-intangible moat in InP and a real integration edge in CPO; these protect the high-margin component tier and won the NVIDIA deal. But that moat is wrapped around a commoditizing module business and rests on a scarcity condition the industry is collectively dismantling. This is not a wide-moat compounder. It is the best-positioned player in a structurally cyclical, capacity-racing industry — “the best house on a frothy street.”

5. Growth History and Forward Opportunities

5.1 History

Across two decades the combined company has compounded revenue at roughly 20%/year, heavily aided by M&A (the II-VI/Coherent merger, the earlier Finisar and Oclaro acquisitions in the lineage). The relevant recent record: revenue troughed at $4.71B in FY2024 (a telecom/industrial down-cycle plus merger disruption), recovered to $5.81B in FY2025 (+23%), and is accelerating — pro-forma YoY growth rose from ~+22% in Q2 FY2026 to +27% in Q3 FY2026, with management explicitly guiding FY2027 growth to exceed FY2026’s (an unusual acceleration claim from a company already at a $7B run-rate). Backlog is at record levels with orders extending into calendar 2028.

5.2 The quarterly trajectory

The sequential growth rate is itself accelerating (from +2% to +7% per quarter), which is the empirical backbone of management’s acceleration narrative and is unusual for a hardware company of this size.

5.3 Forward growth vectors and a quality read

5.4 The cannibalization question (the central growth nuance)

Management frames CPO as pure SAM expansion (copper→optical in scale-up). That is true for scale-up. But scale-out CPO displaces the pluggable transceivers Coherent already sells — its highest-volume franchise. Coherent’s own CTO concedes pluggables retain flexibility/serviceability advantages CPO sacrifices and that the two “will coexist.” The net effect on Coherent specifically is unproven: it could be additive (more total optical content per system, and Coherent supplies both form factors), or it could compress a profitable, high-volume installed franchise as systems consolidate optics into the package. This is the most important growth subtlety the bull case under-weights.

5.5 Durable secular or cyclical spike?

The honest read: mostly a cyclical AI-capex spike with a genuine secular core. The secular core is optics inexorably displacing copper (physics) plus Coherent’s InP/CPO integration. The cyclical overlay is hyperscaler capex levels, double-ordering risk, and InP scarcity pricing that mean-reverts as capacity quadruples. Growth is high-quality in character (margin-accretive, demand-pull, profit-pool-concentrated) but not yet demonstrated as durable through a down-cycle — and the optical industry’s history says the down-cycle always comes.

Verdict (Growth): High-quality, multi-vector, and accelerating — but single-thesis (AI capex) at root, with a live self-cannibalization question and untested durability through cycle. A spectacular up-cycle resting on a foundation that has never been stress-tested at this scale.

6. Financial Quality

6.1 Revenue and margin trajectory

Revenue (FY June): FY2021 ~$3.1B (legacy II-VI) → FY2022 $3.32B → FY2023 $5.16B (first full combined year) → FY2024 $4.71B (down-cycle) → FY2025 $5.81B (+23%) → FY2026 annualizing ~$7.05B at the guide midpoint. GAAP gross margin: FY2025 was $2,043M on $5,810M = 35.2%; non-GAAP gross margin has climbed in 7 of the last 8 quarters to 39.6% in Q3 FY2026 (+~570bps from the FY2025 exit trough toward the Q4 guide), against a stated >42% target from the May-2025 Investor Day. The margin bridge is explicit and credible: (1) 6-inch InP yield and cost (half the per-device cost), (2) pricing optimization in a tight market, (3) mix toward Datacenter & Communications. The first is structural and durable; the second is cyclical and will reverse with the cycle.

6.2 Profitability inflection — and a quality-of-earnings caveat

GAAP net income inflected hard — from –$156M (FY2024) and +$49M (FY2025) to +$564M in the first nine months of FY2026 alone. Two quality caveats temper the headline: (1) Q1 FY2026 net income included a ~$115M pre-tax gain on the Aerospace & Defense divestiture (one-time); and (2) GAAP results have been heavily depressed for years by merger intangible amortization (against ~$2.96B of net intangibles) and restructuring — so the GAAP figures understated underlying earnings on the way down, and the convergence of GAAP toward non-GAAP now is partly the natural fading of that amortization, not solely operating improvement. Non-GAAP EPS — the metric to track — grew 55% YoY in Q3 FY2026 to $1.41, with the Q4 guide $1.52–1.72 implying ~$5.4–5.5 for FY2026. The direction is unambiguously positive and operating leverage is real; the magnitude of the GAAP inflection is flattered by a one-time gain and an amortization tailwind.

6.3 Returns on capital — the weak spot

This is where the “good business” claim is most vulnerable. Reported TTM ROE is ~4.7% and ROA ~3.1% (third-party, on a GAAP basis depressed by amortization and a now-much-larger equity base). Even normalizing for amortization, returns on the gross capital employed are mediocre because the balance sheet still carries $4.40B of goodwill and $2.96B of net intangibles from the Coherent deal, against which the business must earn. On a forward non-GAAP basis returns are improving rapidly (rising margins on a growing revenue base), but Coherent is not a high-ROIC compounder today — it is a capital-intensive manufacturer earning a cyclically-improving return on a goodwill-laden balance sheet. Critically, the incentive system contains no ROIC or return metric, so management is not being steered toward capital efficiency precisely as it deploys the most capital in its history.

6.4 Cash flow and capital intensity

Operating cash flow has been steady at ~$550–635M annually, but capex is surging — $547M in 9M FY2026 (+77% YoY), with a further ~$202M sitting in accounts payable (a leading indicator of an even higher run-rate), directed at the 6-inch InP build. Free cash flow is therefore thin-to-negative in the heavy-investment phase: the business is converting its cash generation, its divestiture proceeds, and its equity issuance into fab capacity. This is defensible if the demand is durable and the returns materialize; it is the classic capital-cycle trap if they do not.

6.5 Balance sheet

Transformed over nine months. Total debt fell from $3,687M (Jun-25) to $3,194M (Mar-26); cash rose from $909M to $1,593M on the balance sheet (management cites ~$3.0B including short-term investments after the NVIDIA inflow). Net leverage fell to 0.5x (from 2.1x a year earlier). Debt is well-termed: Term A ($1,141M, SOFR+1.50%, due 2030), Term B ($1,080M, SOFR+1.75%, due 2029), 5.000% Senior Notes ($990M, due 2029), with an undrawn $700M revolver. The mandatory preferred overhang is gone (converted). The balance sheet is no longer a risk; the dilution that fixed it is the cost.

Verdict (Financial quality): Improving fast, but the economics are cyclical and capital-intensive, not compounding. Margins are expanding on a real (6-inch InP) structural driver plus a cyclical (pricing) one; profitability is inflecting; the balance sheet is repaired. But returns on capital are mediocre against a goodwill-heavy base, free cash flow is consumed by a capacity build, and the GAAP earnings inflection is partly flattered by a one-time gain and fading amortization. Economics do improve with scale here — but from a low base and into a cyclical headwind that has not yet arrived.

7. Capital Allocation

7.1 The financing arc

Coherent’s recent capital-allocation story is dominated by the deleveraging and recapitalization of a balance sheet that was stretched to fund the 2022 Coherent acquisition. That ~$7B deal was financed with a $4.0B senior secured facility, $990M of 5% Senior Notes, $1.4B of Series B-2 mandatory preferred, and an earlier $2.3B mandatory convertible — leaving goodwill of $4.40B and net intangibles of ~$2.96B still on the books. Three events in FY2026 transformed the capital structure:

1. Series B preferred conversion (Dec-2025 quarter). The combined $2.15B-face, 5%, $85-strike Series B preferred (held by Bain Capital) converted into 30.12M common shares, moving $2.51B from mezzanine to common equity after Bain waived future dividends. This removed a costly, dilutive overhang — but the dilution was already economically embedded; conversion simply realized it on-balance-sheet.
2. NVIDIA $2.0B equity investment (Mar-2026). A private placement of 7.79M shares at $256.80, alongside the CPO supply collaboration. This is genuine new cash and a strategic validation — but issued to Coherent’s own largest prospective customer, at a price already well into the up-cycle.
3. Debt paydown. ~$493M of debt retired in nine months, taking net leverage to 0.5x.

The cost of the repair: shares issued rose ~24% in nine months (from 171.9M to 212.3M), the bulk from the preferred conversion plus the NVIDIA placement plus ongoing SBC. The deleveraging that looks prudent today was substantially funded by dilution at a high valuation.

7.2 Capex, R&D, and shareholder returns

Capital priorities are unambiguous and explicitly stated: (1) capacity capex (6-inch InP) and (2) debt paydown — not shareholder returns. There is no common dividend (none since 2012) and no buyback; the company is a net issuer. R&D intensity is steady at ~10% of revenue. Capex is accelerating into the InP ramp. Customer LTAs increasingly carry upfront capacity contributions from customers (“skin in the game”) plus minimum-volume commitments — a genuinely favorable feature that de-risks the capacity bet, though the amounts are disclosed only as contract liabilities, not broken out.

7.3 M&A and divestitures

The pattern is rational portfolio focus: the value-accretive A&D sale (+$115M gain), the value-destructive but sensible cleanup of the Munich division (sold at a loss), and the 2023 SiC stake sale to DENSO/Mitsubishi ($1.0B for ~25%, bringing in strategic capital and partners while retaining control). This is focusing, not empire-building. (The “Bystronic M&A call” of Oct-2025 surfaced in the transcript feed but appears nowhere in the SEC filing corpus through the Mar-2026 10-Q — an open item, not a disclosed Coherent transaction.)

7.4 Incentives — the red flag

Executive compensation is the clearest governance concern. CEO Jim Anderson (who joined in June 2024 from Lattice Semiconductor, replacing Vincent Mattera) received a ~$100.9M inducement equity grant in FY2024 and modest cash thereafter (FY2025 total $3.7M, with no FY2025 equity). The plan design:

• Annual cash incentive: Revenue (50%) + Adjusted EBITDA (50%); FY2025 paid out at 170% of target on record revenue and EBITDA.
• Long-term incentive (PSUs): 100% relative TSR vs the S&P Composite 1500 Electronic Equipment index; FY2023–2025 PSUs paid 108%.
• What is missing: no ROIC, no EPS, no gross-margin or return-on-capital metric anywhere. Pay is geared to volume growth and market beta.

This is a Marathon/Greenwald red flag in plain sight: management is paid to grow revenue and EBITDA and ride the optical beta, with no governor on the returns earned on the capital being deployed — precisely as it quadruples capacity and issues equity. Say-on-pay recovered to ~96.6% support in 2025 after a weak prior-year vote (driven by Mattera’s >3x cash severance and thin TSR disclosure), and the company added a 3x cash-severance cap and better disclosure — governance is improving at the margin, but the core metric gap remains.

7.5 Insider behavior

The Form 4 record shows zero open-market purchases by any insider across FY2025–2026 (no conviction buying) and only routine, mostly 10b5-1-planned selling into strength (CFO Luther, CTO Eng, several directors). CEO Anderson’s only transaction was tax-withholding on vesting RSUs (code F). Neutral-to-mildly-negative — unsurprising after a ~4–5x run, but offering no bullish signal.

Verdict (Capital allocation): Mixed — disciplined balance-sheet defense wrapped around a textbook late-cycle capacity offense, funded by heavy dilution and steered by an incentive plan with no return metric. The defense is genuinely competent (leverage 2.1x→0.5x, overhang cleared, debt termed out, portfolio focused). The offense is a high-conviction, pro-cyclical bet — quadrupling InP capacity, partly with equity issued at a boom valuation to its own customer — that the current scarcity rents are durable. Management is allocating capital aggressively and skillfully in the moment; whether it is allocating wisely across the cycle is the open question, and the comp plan does nothing to discipline it.

8. Changes and Headwinds — Last Two Years

Strategic and structural changes (thesis-strengthening or neutral):

• Leadership refresh: Jim Anderson became CEO (Jun-2024), Sherri Luther CFO (Oct-2024), with a near-total C-suite refresh importing semiconductor operating talent (Lattice pedigree) into a historically materials-house culture. Intent is execution-positive; durability unproven.
• Re-segmentation to Datacenter & Communications / Industrial, reorienting the company around the AI franchise.
• Portfolio pruning: A&D sold (+$115M gain, Sep-2025), Munich division divested (Jan-2026, at a loss), SiC stake sold (2023).
• The NVIDIA partnership (Mar-2026): $2B equity + multibillion CPO supply agreement through end-of-decade — the single most thesis-strengthening event, validating CPO breadth.
• Recapitalization: Series B preferred conversion, ~$493M debt paydown, leverage to 0.5x.
• 6-inch InP ramp ahead of schedule (capacity-doubling pulled a quarter earlier), with first 6-inch-based transceivers shipped in Q3 FY2026.

Headwinds and risks (thesis-weakening or watch-items):

• Dilution: ~24% more shares in nine months.
• Industrial softness: the segment declined modestly YoY, only now showing booking improvement.
• Capex surge consuming free cash flow.
• Valuation re-rating: the stock moved from ~$74 to a ~$413 high and ~$356 now — a ~4–5x move that has pulled forward years of the thesis into the multiple.
• Intensifying competition and the capacity race, and the live CPO-cannibalization question.

Verdict: The operational and strategic changes are decisively thesis-strengthening (NVIDIA, margins, balance sheet, focus). The principal new headwind is not operational but valuation and dilution — the business got better and the stock got much more expensive, faster.

9. Risk Analysis (Risk Matrix)

The dominant risk cluster is the cycle: demand reversal, capacity glut, pricing reset and valuation de-rating are correlated and would arrive together. A catastrophic permanent loss is unlikely (real assets, real technology, repaired balance sheet, NVIDIA anchor), but a large drawdown from current levels on any cyclical disappointment is a live, even probable, medium-term scenario.

10. Valuation Discussion (Embedded Expectations)

No price target or recommendation in this section — only the expectations embedded in the current price and scenario analysis.

10.1 Where the multiple sits

At ~$356/share, ~195.5M shares, market cap ~$70B and EV ~$71B:

• EV/Revenue (TTM ~$6.6B): ~10.8x. On the FY2026 run-rate (~$7.05B): ~10x. On a forward FY2027 estimate (~$8.7B if growth accelerates as guided): ~8x.
• P/E: trailing GAAP ~118x (depressed by amortization); FY2026 non-GAAP (~$5.45 EPS) ~65x; FY2027 consensus non-GAAP (~$8.10) ~44x.
• vs. own history: AZI valuation-index percentiles — P/S 99.1, P/B 98.5, P/E 81, composite 92.8 vs ~10-year history. Coherent has essentially never been this expensive on sales or book.

10.2 Peer comparison

*LITE EV/EBITDA distorted by a depressed TTM denominator.

The pivotal observation: on P/S, Coherent (~10.8x) is the cheapest in the AI-optical cohort — roughly half Lumentum’s ~26x and well below the fabless designers (Marvell/Broadcom ~24–26x). It clusters with Ciena (~11x) as an optical hardware maker rather than with the chip designers. This is the bull’s relative-value argument, and it is legitimate: Coherent is the lowest-multiple way to own the InP/CPO thesis with the broadest portfolio and the most scale. But the “cheap vs. peers” framing only holds against a cohort that is itself at a 99th-percentile valuation, and Coherent’s discount partly reflects its structurally lower gross margin (~40% vs Lumentum’s ~44%, vs fabless 50–60%) and heavier capex. The discount is deserved, not purely opportunity; the question is whether margin convergence toward the >42% target narrows it.

10.3 What must the market be underwriting?

To justify ~$70B EV at ~44x FY2027 non-GAAP EPS, the embedded expectations are approximately:

• Sustained 20%+ revenue growth for multiple years (FY2026 ~+21%, FY2027 faster, then continued compounding) — i.e. the AI-optics build does not hit an air-pocket and the speed-cadence (1.6T→3.2T) keeps ASPs and volumes rising.
• Gross margin reaching and holding >42% — i.e. the 6-inch InP structural benefit more than offsets the cyclical pricing reversal as capacity floods in.
• CPO and OCS scaling into multi-billion revenue without the pluggable cannibalization compressing the core.
• Returns on the capex eventually showing up — the capacity build earning a return rather than becoming stranded late-cycle cost.

In short, the price underwrites that today’s scarcity rents are structural and that the cycle has been repealed. The market is plausibly right about the secular direction (optics displacing copper) and the company’s relative positioning; it is most exposed on the durability and pricing assumptions — that margins and growth hold as the entire industry’s 4x InP capacity build arrives.

10.4 Scenario analysis (illustrative, non-GAAP)

• Bear: Hyperscaler capex pauses or InP capacity glut resets pricing in CY2027. Revenue growth decelerates to high-single-digits, gross margin stalls ~38–39%, FY2027 EPS ~$5–6. A cyclical name on disappointing growth re-rates to ~20–25x → an equity value well below current levels (a 40–55% drawdown is conceivable). The optical industry’s history (JDSU, the 2001 and 2019 telecom busts) says this is not a tail scenario.
• Base: Growth moderates from +27% toward ~15–20%, gross margin reaches ~42%, FY2027 EPS ~$8, the multiple normalizes toward ~30–35x as the market accepts the cyclicality → roughly the current price to modestly higher, i.e. the stock has already discounted a successful base case. Total return is carried by earnings growth against multiple compression.
• Bull: CPO and OCS inflect faster than expected, NVIDIA volume scales, gross margin pushes past 42% toward the mid-40s, the InP scarcity proves stickier than the bears think, and FY2027 EPS reaches ~$9–10 with the multiple holding ~45x → meaningful further upside (the stock continues to compound with earnings). This is the momentum-thesis continuation.

Verdict (Valuation): The price embeds a successful base-to-bull outcome and prices the cycle as if it were repealed. Coherent is the cheapest entry into the AI-optical thesis within its cohort, but the cohort is at a 99th-percentile valuation and Coherent’s own multiple is near its historical extreme. There is little valuation cushion for the cyclical disappointment the industry’s capacity build makes increasingly likely. The risk/reward is unattractive here; it becomes attractive on a cyclical pullback that the demand visibility and NVIDIA anchor make survivable.

11. Variant Perception

Consensus view: Coherent is a top-two AI-optical/photonics franchise with a real InP scarcity moat, an NVIDIA-validated CPO opportunity, accelerating growth, expanding margins and a repaired balance sheet — and it is “cheap” relative to optical/AI peers. Sell-side rating is strongly positive (16 buy/strong-buy, 3 hold, 0 sell).

Strongest bull case: The optics-displaces-copper transition is a multi-year, physics-driven secular shift, not a fad; Coherent’s vertical InP integration is a genuine, widening cost/scale moat that structurally lowers cost as 6-inch scales; the NVIDIA partnership de-risks CPO and provides end-of-decade visibility; OCS, CPO, multi-rail and thermal are multiple additive growth vectors layering on top of an accelerating core; margins are climbing toward >42% with the InP benefit “more ahead of us than behind”; and at ~10x forward sales it is the lowest-multiple, broadest-portfolio way to own the entire theme. Growth is accelerating into FY2027 — rare and valuable.

Strongest bear case: This is a cyclical, capital-intensive component maker at a 99th-percentile (own-history) valuation, riding a single AI-capex thesis. The scarcity rents driving margins are being competed away by an industry-wide 4x InP capacity build (Marathon late-cycle); the module tier commoditizes (Chinese-led); CPO cannibalizes the pluggable franchise; customer/partner power concentrates in NVIDIA and a few hyperscalers; the balance-sheet repair was bought with ~24% dilution; returns on capital are mediocre against $7.4B of goodwill+intangibles; and the comp plan rewards growth with no return governor. The optical industry’s history of ~90%+ drawdowns is not ancient — and a $70B equity priced for permanent acceleration is not discounting a cycle that always returns.

The 3–5 assumptions that matter most:

1. Durability of demand — does AI-optics capex sustain 20%+ growth, or hit an air-pocket?
2. Margin permanence — does the 6-inch InP structural benefit hold gross margin >42% through the industry capacity glut and pricing reset?
3. CPO net effect — additive content or cannibalization of the pluggable profit pool?
4. Pricing power direction — does it stay with Coherent (scarcity) or drift to hyperscaler/NVIDIA buyers (concentration)?
5. Capital-cycle timing — when does the industry’s 4x InP build flip scarcity to surplus?

What would falsify each side: The bull is falsified by a hyperscaler capex cut, a 1.6T price reset, or gross margin stalling below 42% as capacity arrives. The bear is falsified by gross margin reaching the mid-40s and holding through an InP capacity normalization (proving the rents structural), CPO scaling without core compression, and growth durably exceeding 20% for multiple years.

12. Fact vs. Interpretation Table

13. Open Questions

1. Does CPO net-cannibalize the pluggable franchise, and what is the net content/margin effect on Coherent specifically?
2. How much of the gross-margin gain is structural (6-inch InP) vs. cyclical (pricing)? Management attributes it to both; the split determines margin durability through a glut.
3. When does the industry’s 4x InP capacity build flip scarcity to surplus — CY2027? Later? This is the single most important cyclical timing question.
4. What are the actual LTA terms (volume commitments, pricing, customer capex contributions)? Disclosed only as aggregate contract liabilities.
5. How concentrated is revenue in NVIDIA + top hyperscalers? Not quantified in filings; the concentration is the key pricing-power risk.
6. What was the “Bystronic” transaction referenced in the Oct-2025 transcript feed but absent from the SEC corpus?
7. Will the comp committee add a return-on-capital metric as capex scales? Governance is improving but the metric gap persists.
8. Geopolitical InP feedstock exposure — true China dependence and mitigation.

14. What Must Be True

For the BULL case to be right:

1. AI-data-center optical capex sustains 20%+ revenue growth for multiple years (no air-pocket).
2. Gross margin reaches and holds >42% — the 6-inch InP structural benefit out-runs the cyclical pricing reset as industry capacity quadruples.
3. CPO/OCS/multi-rail scale into multi-billion revenue additively, without compressing the pluggable core.
4. Pricing power stays with Coherent (scarcity/integration) rather than drifting to hyperscaler/NVIDIA buyers.

Falsification test: A hyperscaler capex cut, a 1.6T transceiver price reset, or non-GAAP gross margin stalling below ~40% for two-plus quarters as new InP capacity arrives — any one falsifies the bull and should trigger a re-rating.

For the BEAR case to be right:

1. The industry’s 4x InP capacity build creates a glut that resets pricing and compresses margin within ~6–8 quarters.
2. Module commoditization and customer concentration pull pricing power toward buyers.
3. CPO cannibalizes more pluggable profit than it adds.
4. The current valuation de-rates toward the cohort/own-history mean as the cyclicality is re-recognized.

Falsification test: Non-GAAP gross margin pushing into the mid-40s and holding through a visible InP capacity normalization (proving the rents structural, not scarcity), plus CPO scaling without core compression, plus growth durably above 20% — this combination falsifies the bear and justifies the premium.

15. Source Appendix

Primary sources (full citations in the standalone Source Appendix, Appendix B of the combined report):

• SEC filings (EDGAR, CIK 0000820318): 10-K FY2025 (filed 2025-08-15), 10-Q Q3 FY2026 (filed 2026-05-06) and prior 10-Qs, DEF 14A (filed 2025-10-02), 8-Ks (NVIDIA investment 2026-03-02; Series B conversion 2025-11-21; say-on-pay 2025-11-17), Form 4 insider filings. Financial time-series via SEC XBRL company-concept API.
• Earnings & event transcripts (Coherent IR / AZI feed): Q3 FY2026 call (2026-05-06), Q2 FY2026 (2026-02-04), Q1 FY2026 (2025-11-05), Q4 FY2025 (2025-08-13), Analyst & Investor Day (2025-05-28), OFC/CPO special call (2026-03-17).
• Quantitative aggregators (reconciled to filings): AZI fundamentals snapshot & valuation-index (2026-06-09); yfinance comps (2026-06-10). Note: AZI income/cash-flow statement arrays are unreliable for COHR post-merger and were not used; statements are taken from EDGAR/10-K/10-Q.
• Peer/industry comparison (public filings & market data): Lumentum (LITE), Marvell (MRVL), Broadcom (AVGO), NVIDIA (NVDA) for competitive and capital-cycle framing.

Management commentary has been treated as hypothesis throughout and validated against filings, financial outcomes, and external/peer evidence. Facts, interpretations, assumptions and open questions are labeled accordingly.

APPENDIX A — Standard Diligence Questionnaire

Coherent Corp. (NYSE: COHR) — Standard Diligence Questionnaire

Answers grounded in SEC filings, company disclosures, and the Q3 FY2026 earnings call. Fact / Interpretation / Assumption labels applied where material. Greenwald (Competition Demystified) and Marathon (Capital Returns) lenses applied where they add insight.

General

What thoughtful questions have other investors asked about this company? From the Q3 FY2026 call and conference questions: (1) Why isn’t gross-margin fall-through bigger given 6-inch InP, exited low-margin businesses, and mix shift? (Answer: early in the 6-inch ramp; benefit “more ahead of us.”) (2) Why doesn’t doubling InP capacity double revenue? (Answer: 2–3-month latency from device production to transceiver shipment.) (3) The perceived gap vs. a primary competitor (Lumentum) on OCS/CPO forecasts. (4) Did 1.6T transceiver revenue exceed $100M in the March quarter? (Not broken out.) (5) EML vs. SiPho mix and margin implications (roughly equal margin; both need InP). The sophisticated investor debate centers on margin durability, the InP capacity-glut timing, and CPO cannibalization — the same axes as our variant-perception section.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: Early-to-mid in an up-cycle, accelerating. Revenue is at an all-time high and still accelerating sequentially (+7% QoQ in Q3 FY26), and management guides FY27 growth above FY26. Margins are climbing, not peaking. But the industry is plausibly approaching a capital-cycle peak as InP capacity quadruples — so company earnings are rising while the cyclical risk builds beneath them.

Driven by external environment or internal actions? Both. External: AI-data-center capex and InP scarcity (pricing). Internal: the 6-inch InP ramp (structural cost), portfolio pruning, SG&A leverage, ERP consolidation. The structural/internal share is what determines durability.

How stable are revenues? Fact: Not stable through cycle — FY2024 revenue fell to $4.71B from $5.16B before recovering to $5.81B (FY25) and ~$7.05B (FY26e). This is a cyclical hardware business; the FY2024 dip is the recent proof.

Outlook for products/services? Strong near-term: record backlog, orders into CY2028, LTAs to end-of-decade, multiple new vectors (OCS, CPO, multi-rail, thermal). Longer-term clouded by capacity-glut and commoditization risk.

How big will this market be? Fact (mgmt): Datacenter & Communications ~$44B by 2030; CPO >$15B incremental; OCS >$4B; multi-rail >$2B. Growing, global, AI-driven. Treat TAMs as directional management framing, not independently verified.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Interpretation: More — the InP/transceiver capacity race (Coherent, Lumentum, Greensboro, Chinese EML startups) plus hyperscaler second-sourcing and Broadcom/Marvell vertical encroachment all intensify competition into the boom.

How profitable is the business (ROIC, ROE)? Fact: Modest on a reported basis — TTM ROE ~4.7%, ROA ~3.1% (GAAP, depressed by merger amortization and a much-larger equity base). Interpretation: Returns on gross capital are mediocre against $4.40B goodwill + $2.96B intangibles, though improving rapidly on a forward non-GAAP basis. Not a high-ROIC compounder today. The correct sector framing is a capital-intensive manufacturer with cyclically-improving returns, not an asset-light compounder.

How profitable is the industry — competitors, barriers? Bifurcated: the InP-component tier is high-margin and concentrated (high barriers: capital, process know-how, materials); the module-assembly tier is low-margin and commoditizing (low barriers, Chinese-led). Coherent straddles both.

Can the business be easily understood? Moderately. The franchise logic (vertical InP integration → optical content for AI) is clear; the technical detail (EML vs. SiPho vs. VCSEL, CPO vs. pluggable, 6-inch yields) is specialized and requires domain literacy to handicap.

Can it be undermined by foreign low-cost labor? Fact: Partly — yes at the module-assembly tier (Chinese InnoLight/Eoptolink already lead). No (so far) at the InP-device tier, which is process/capital-gated rather than labor-gated. Coherent’s defense is to keep value concentrated in the chip/component layer.

Do brands matter? No. These are specified components/systems sold to sophisticated buyers on technology, supply assurance and price — not brand.

Nature of competition? Technology leadership (data-rate roadmap), capacity/supply assurance, vertical integration breadth, and price. Increasingly, relationship/LTA competition (NVIDIA, hyperscalers).

Customers’ switching costs? Interpretation: Weak-to-moderate. Qualification cycles create some friction, but buyers deliberately multi-source and pre-allocate. Not a switching-cost moat.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Interpretation: The proprietary InP/VCSEL process know-how and materials recipes (Thermadite, in-house garnet) are internally-developed intangibles not capitalized — genuine economic assets understated on the balance sheet. Conversely, the ~$7.4B goodwill+intangibles from the Coherent deal are over-stated relative to current returns.

Off-balance-sheet liabilities? No unusual items identified; operating leases and pension are standard. LTAs carry forward purchase/capacity commitments (and customer contributions as contract liabilities).

How conservative is the accounting? Interpretation: Reasonable. GAAP has been conservative-by-circumstance (heavy amortization/restructuring depressed reported earnings). Watch the non-GAAP add-backs (SBC ~$160M/yr is real dilution) and the one-time A&D gain in Q1 FY26 net income.

How CapEx-hungry is the business? Fact: Very, currently — capex $547M in 9M FY26 (+77% YoY), guided higher, with ~$202M more in payables. This is a heavy-capex phase funding the InP build; free cash flow is thin-to-negative.

Capital Allocation & Management

How much FCF, and how is it used? Fact: Operating cash flow ~$550–635M/yr, but largely consumed by capex right now. Priorities: (1) capacity capex, (2) debt paydown. No dividend, no buyback. Customer upfront capex contributions partially fund the build.

Significant acquisitions recently? No recent acquisitions; the story is divestitures (A&D, Munich) and the 2022 Coherent mega-deal still being digested. (The Oct-2025 “Bystronic” transcript reference is unconfirmed in filings — open question.)

Buying back shares? No — net issuer (Series B conversion +30.1M, NVIDIA +7.79M, SBC). ~24% more shares in 9 months.

Issuing large amounts of stock to insiders? SBC ~$160M/yr; CEO received a ~$100.9M inducement grant in FY2024. Material but not unusual for the sector/transition.

Compensation policy / incentive metrics? Fact: Annual cash = Revenue (50%) + Adjusted EBITDA (50%) — paid 170% in FY25; LTI = 100% relative TSR. No ROIC, EPS or margin metric. Interpretation: Rewards volume/beta, not capital efficiency — a red flag as capex scales. Say-on-pay ~96.6% (2025), recovered after a weak prior year.

Motivations of management? Interpretation: New regime (Anderson ex-Lattice, since Jun-2024) importing semiconductor operating discipline; clearly oriented toward growth and margin expansion. Incentive alignment to returns is weak; to growth is strong.

Valuation & Market Data

ADR / MLP / K-1? No — ordinary NYSE common, US domestic filer (10-K/10-Q). No K-1.

Dividend policy? None since 2012; zero current yield. Capital goes to capex and debt.

How profitable is the business? Non-GAAP operating margin ~20%, rising; GAAP modest but inflecting. Improving, cyclical.

Is net income diverging from cash from operations? Fact: In 9M FY26, GAAP net income ($564M) is boosted by the one-time A&D gain; operating cash flow is solid but capex absorbs it. Watch FCF, not net income, through the build phase.

Risks & Downside

What would cause the stock to decline? A hyperscaler capex air-pocket; an InP capacity glut / 1.6T price reset; gross-margin stall below ~40%; valuation de-rating from 99th-percentile levels; NVIDIA/hyperscaler concentration turning to pricing pressure; CPO cannibalization; further dilution.

Risk of catastrophic loss? Interpretation: Low — real assets, real technology, repaired balance sheet (0.5x leverage), NVIDIA anchor. A large drawdown (40%+) on cyclical disappointment is plausible; a permanent total loss is not, absent a fraud/impairment shock.

Chance of total loss? Very low.

Recent News & Events

Has the business environment changed recently? Yes — decisively better operationally: the NVIDIA $2B investment + CPO supply deal (Mar-2026), accelerating revenue, margin expansion, and balance-sheet repair to 0.5x leverage. The valuation environment also changed: a ~4–5x stock move pulled years of thesis into the price.

Significant acquisitions / divestitures? Divestitures: A&D (Sep-2025, +$115M gain), Munich division (Jan-2026, at a loss). 2023 SiC stake sale ($1.0B/25% to DENSO/Mitsubishi).

Change in accounting policies? Re-segmentation to Datacenter & Communications / Industrial; Series B preferred reclassified to common on conversion. No adverse accounting-quality signal identified.

Recent changes — markets, facilities, management? New 6-inch InP facilities (Sherman TX, Sweden, Zurich coming); near-total C-suite refresh (CEO Anderson Jun-2024, CFO Luther Oct-2024); reorientation around AI-datacenter optics.

APPENDIX B — Source Appendix

Coherent Corp. (NYSE: COHR) — Source Appendix

Source register for the Coherent Corp. research memo (report date 2026-06-10). Primary sources (SEC filings, company transcripts, regulatory documents) are listed first; public market data and peer comparisons follow. Every non-obvious fact in the memo traces to one of these. Management commentary is treated as hypothesis throughout and validated against filings, financial outcomes, and external/peer evidence.

A. SEC filings (EDGAR, CIK 0000820318)

All filings obtained from SEC EDGAR (CIK 0000820318). Financial time-series cross-checked against the SEC XBRL company-concept API (data.sec.gov).

B. Company earnings & event transcripts

Company earnings and event transcripts (company IR / public transcript services). Management framing is treated as hypothesis and validated against filings.

C. Quantitative aggregators (convenience data, reconciled to filings)

D. Peer / industry comparison (public filings & market data)

Used for the competitive map, profit-pool analysis, and capital-cycle framing; independent primary research was performed for COHR itself.

• Lumentum (LITE) — closest direct peer; merchant-EML scarcity economics, InnoLight/Eoptolink module share, Greensboro capacity build, capital-cycle framing.
• Marvell (MRVL) — optical DSP / custom-silicon / light-engine encroachment.
• Broadcom (AVGO) — vertically-integrated CPO incumbent (Tomahawk co-packaged switches), internal EMLs.
• NVIDIA (NVDA) — partner/customer/shareholder; CPO architecture and demand context.

Peers cited for framing and cross-check only; all COHR conclusions rest on the public primary sources above.