Coherent Corp. (COHR) - Deep Dive Research Report

Prepared: May 10, 2026 | Concalls used: Q4 FY25 (Aug 13, 2025), Q1 FY26 (Nov 5, 2025), Q2 FY26 (Feb 4, 2026), Q3 FY26 (May 6, 2026)

1. What the Company Does

Coherent Corp. makes the light that keeps the internet alive. More precisely, it designs and manufactures the photonic components - lasers, transceivers, modulators, and optical switches - that convert electrical signals into light pulses and route those pulses across the fiber networks inside data centers and between cities. Without products like Coherent's, the data traveling between the GPU clusters powering ChatGPT or Google Gemini would have nowhere to go.

The company traces its roots to 1971, when it was founded as II-VI Incorporated in Saxonburg, Pennsylvania. The name referred to Groups II and VI of the periodic table - the chemical families that form compound semiconductors like zinc selenide and cadmium telluride, which II-VI used to make infrared optical materials for early laser systems. For five decades, II-VI was a patient, acquisitive compounder in specialty materials and photonics, growing by buying and integrating businesses that most industrial conglomerates found too niche to bother with.

The company's character fundamentally changed with two transformative acquisitions. In 2019, II-VI acquired Finisar, the California-based transceiver giant, for $3.2 billion. Finisar brought a high-volume manufacturing heritage in optical transceivers - specifically the indium phosphide (InP) laser platform and VCSEL technology used in Apple Face ID sensors - as well as enormous production facilities in Sherman, Texas and Sunnyvale, California. This deal repositioned II-VI from a specialty materials company into a full-stack photonics manufacturer capable of competing in the data center interconnect market.

The second transformation came in July 2022, when II-VI outbid Lumentum in a bruising auction to acquire Coherent, Inc. - the original laser company founded in 1966 that had commercialized some of the world's first industrial and scientific lasers - for $7.01 billion. That number was 23% higher than Lumentum's initial offer, signaling how badly II-VI's leadership wanted the industrial laser business and coherent optical technology. After that deal closed, II-VI rebranded the combined entity as Coherent Corp. and changed its NYSE ticker from IIVI to COHR.

The resulting company is a sprawling photonics platform with three historically distinct businesses: a high-growth data center networking business built on Finisar's transceiver heritage and InP manufacturing; an industrial laser business inherited from the original Coherent, Inc.; and an engineered materials business (most notably silicon carbide substrates) that II-VI built up over decades. Under CEO Jim Anderson, who joined in March 2023 from Lattice Semiconductor, the company has been aggressively shedding the non-core pieces of this portfolio - selling the aerospace and defense business, divesting the Munich materials processing division, and exiting 33 manufacturing sites - while simultaneously doubling and tripling down on the data center photonics opportunity.

The core value proposition for data center customers is simple but technically demanding: as GPU clusters scale from hundreds to hundreds of thousands of chips, the electrical wires that used to connect servers stop working. Copper cables can carry 400 gigabits per second over very short distances; beyond two meters, signal integrity degrades. At 800G and beyond, optical fiber is the only physics-viable solution. Coherent makes the components that convert electrical signals to light (lasers), route that light (switches), and convert it back to electrical signals (photodetectors and transceivers). The company's edge is that it does this vertically - it grows its own InP crystals, fabricates its own laser chips on wafers it controls, and assembles them into finished transceivers and switches in its own factories. That vertical integration is not a buzzword; it is the specific thing that lets Coherent control its own capacity, price competitively against assemblers who must buy chips on the open market, and protect yields as processes improve.

A concrete example of how this plays out: when a hyperscaler like Google expands an AI training cluster, it needs to connect thousands of NVIDIA H100 or Blackwell GPU servers. Each server needs roughly 16 optical transceiver ports running at 800 gigabits per second. At that density, demand for transceivers in a single data center expansion can consume hundreds of thousands of units. Coherent receives purchase orders from either the hyperscaler directly or from a networking OEM like Cisco, builds the transceiver modules at its Sherman, Texas facility using InP laser chips it fabricated internally on 6-inch wafers, tests them at 40 gigabytes per second lane speed, and ships them. The customer plugs them into switch ports and server NICs; the data center comes alive optically. Switching costs are real but moderate at the transceiver level - the form factor (QSFP-DD or OSFP) and protocol are standardized, so customers can source from multiple vendors. The real lock-in is at the component supply level: companies that control InP laser chip production control the supply chain for the entire transceiver industry.

2. Business Segments

Coherent reorganized its reporting structure in fiscal 2026, consolidating from three segments (Networking, Materials, Lasers) used in its FY2025 10-K into two operational segments: Datacenter and Communications (roughly 75% of revenue) and Industrial (roughly 25%). Understanding both the old and new structures matters, because the old structure reveals the provenance and competitive dynamics of each business unit.

2.1 Datacenter and Communications (formerly Networking)

This is the engine. The Datacenter and Communications segment provides optical components, modules, and systems for AI infrastructure, hyperscale data centers, and global telecommunications networks. It encompasses optical transceivers (the pluggable modules inserted into network switches and servers), the indium phosphide laser chips that go inside them, optical amplifiers, coherent optical modules for long-haul telecom, and the Optical Circuit Switch (OCS).

The core capability here is InP semiconductor manufacturing. Indium phosphide is a III-V compound semiconductor - not silicon. It has a direct bandgap, meaning it converts electrical current to light with high efficiency, which is why it is the substrate of choice for high-speed lasers. Growing InP crystals, slicing them into wafers, and fabricating laser chips from those wafers is a decade-plus manufacturing discipline. Coherent inherited this capability from the Finisar acquisition, which itself had decades of InP experience going back to the Finisar founders who left AT&T Bell Labs in 1988.

Coherent operates InP wafer fabrication at Sherman, Texas (the original Finisar facility) and Yarfala, Sweden (acquired with the Oclaro purchase in the Finisar era). In August 2025, it activated the world's first 6-inch InP production line at Sherman - a milestone because all prior production globally used 3-inch or 4-inch wafers. A 6-inch wafer has four times the surface area of a 3-inch wafer, meaning more than four times as many chips per wafer run. Coherent's management stated that 6-inch yields - meaning the fraction of chips that pass quality testing - were already exceeding their existing 3-inch line yields as of Q1 FY26, an unusually fast learning curve for a new wafer size transition in compound semiconductors. The cost implication is dramatic: more than four times the chips per wafer at less than half the per-chip cost, according to management.

Within the segment, the OCS (Optical Circuit Switch) is the emerging product that management has spent the most time explaining. Traditional data center networks use electrical switches - expensive, power-hungry silicon boxes - to route traffic between servers. An OCS keeps the data signal in the optical domain through the switch, eliminating the optical-to-electrical-to-optical (OEO) conversion at each switching hop. Coherent's OCS uses liquid crystal technology (no moving parts, unlike MEMS-based alternatives) to redirect light beams. It ships in 64x64 port configurations (small cluster applications) and 320x320 port configurations (large-scale AI clusters where the savings on electrical switch elimination are enormous). By Q3 FY26, Coherent had shipped to more than 10 customers and was generating growing revenue and backlog. Management expanded the OCS total addressable market from $2 billion (cited in Q1 FY26) to more than $4 billion (cited in Q3 FY26), reflecting broader-than-expected use cases including scale-out, scale-up, scale-across, and spine switching architectures.

The telecom portion of this segment provides coherent optical modules (coherent here refers to the optical modulation technique, not the company name) for long-haul and metro fiber networks. This business is recovering from a multi-year inventory correction that depressed telecom capex spending in 2023-2024. By Q2 FY26, communications revenue was growing 60% year-over-year, indicating the inventory digestion cycle has fully reversed.

This segment's revenue grew approximately 49% in FY2025 (when reported as Networking) and continued accelerating into FY2026. It is the company's growth engine, margin improvement driver, and strategic priority.

2.2 Industrial (formerly Materials and Lasers)

The Industrial segment is a combination of two historically separate businesses: the engineered materials business (primarily silicon carbide substrates, but also infrared optics, wide bandgap semiconductors, and other specialty materials) and the industrial laser systems business inherited from the original Coherent, Inc.

The silicon carbide (SiC) business deserves detailed treatment because it has an unusual ownership structure. SiC substrates and epitaxial wafers are the foundational material for power semiconductors used in electric vehicles - specifically in the inverters and onboard chargers that convert battery power to motor drive. Coherent makes both 150mm and 200mm SiC substrates. In December 2023, Japanese automotive giants DENSO and Mitsubishi Electric each invested $500 million in Coherent's SiC subsidiary (total: $1 billion), receiving 12.5% stakes each. Coherent retained 75% ownership. Alongside the equity stakes, both companies signed long-term supply agreements for SiC substrates and epitaxial wafers.

This structure - strategic investors who are also anchor customers, with equity and supply agreements bundled together - creates unusual lock-in. DENSO and Mitsubishi Electric need SiC wafers for the power semiconductors going into Toyota, Lexus, and Mitsubishi Electric's industrial systems. They effectively pre-financed Coherent's capacity expansion and locked themselves in as customers. The SiC business suffered in FY2025 as EV demand growth slowed, particularly in China, and excess inventory accumulated across the supply chain. By Q4 FY25, management stated that silicon carbide demand had stabilized and was not expected to be a headwind in FY2026.

The industrial laser business provides high-power laser systems primarily for display manufacturing (OLED panel cutting, annealing, and scribing equipment), semiconductor fabrication, medical applications, and general industrial manufacturing. These are not commodity lasers; they are precision systems requiring deep application engineering and often customer qualification periods. The end markets are episodic - demand spikes when a Samsung or LG builds a new OLED factory, then stabilizes. In FY2025, this business grew 3% year-over-year, driven by OLED fab expansion in Asia.

The Industrial segment as a whole is the value-preservation story: it generates cash, funds the data center investment, and provides optionality if SiC or display markets re-accelerate. Management has been continuously pruning the least profitable parts - selling the Munich materials processing division in Q2 FY26 (which had $25M per quarter in revenue at below-average margins), exiting 33 manufacturing sites total.

Segment Comparison

3. Products and Business Detail

Optical Transceivers

The transceiver is a pluggable module - roughly the size of a thick USB drive - that slides into a port on a network switch or server NIC. It converts electrical signals from the switch ASIC into light pulses that travel down a fiber, and converts incoming light back to electrical signals. Coherent's transceiver portfolio spans:

400G Transceivers (QSFP-DD form factor): Four lanes at 100 gigabits per second each. Still the backbone of many production data center networks. Coherent sells both silicon photonics and InP-based variants. This is a mature, high-volume product.

800G Transceivers (QSFP-DD and OSFP form factors): The current workhorse of AI cluster interconnect. Eight lanes at 100G per second each. Coherent is one of a handful of companies shipping 800G at volume. This product drove the FY2025 revenue surge (+49% in Networking).

1.6T Transceivers (OSFP form factor): Eight lanes at 200 gigabits per second each. This requires either silicon photonics, EML (electro-absorption modulated lasers), or VCSEL technology capable of 200G per lane. Coherent began shipping EML-based and silicon photonics-based 1.6T in late FY2025. VCSEL-based 1.6T (using Coherent's proprietary 200G VCSELs manufactured in Sherman) was slated for ramp in H2 calendar 2026. The 1.6T transition is the single largest near-term revenue driver.

EML (Electro-absorption Modulated Lasers): EMLs are laser chips where a built-in modulator varies light intensity at extremely high speed. Coherent makes a differential-EML (D-EML) that doubles signal amplitude while reducing crosstalk - a design advantage for 1.6T and future 3.2T systems. EMLs are also sold as components to other transceiver makers who lack their own laser fabrication capability.

VCSEL (Vertical Cavity Surface Emitting Lasers): VCSELs emit light perpendicular to the wafer surface, enabling high-density arrays. Coherent inherited VCSEL manufacturing from the Finisar acquisition, specifically the 3DS (3-dimensional stacking) VCSEL platform originally developed for Apple's Face ID system. The company signed a multi-year agreement with Apple (announced Q4 FY25) for VCSELs for a new generation of iPhone and iPad sensors, with revenue expected in H2 calendar 2026.

CW (Continuous Wave) Lasers: CW lasers emit a constant, unmodulated beam of light. In data center transceivers, a CW laser provides the light source that an external modulator then switches on and off. For co-packaged optics (CPO), high-power CW lasers manufactured on Coherent's 6-inch InP platform are the critical component. The NVIDIA supply agreement specifically covers these high-power CW lasers.

Optical Circuit Switch (OCS): As described in Section 2.1, these use liquid crystal technology to route optical signals without converting to electrical form. The 64x64 configuration is designed for smaller cluster applications; the 320x320 is designed for large-scale AI hypercluster interconnect where the savings from eliminating electronic switches are most compelling. Coherent claims its liquid crystal approach has reliability advantages over MEMS (micro-electromechanical systems) mirror-based switches because there are no moving parts. By Q3 FY26, Coherent was generating meaningful revenue from OCS, with backlog weighted toward the larger 320x320 systems.

Co-Packaged Optics (CPO): The next frontier. Rather than pluggable modules that sit outside the switch ASIC, CPO integrates the optical I/O directly onto the same package as the switch chip or GPU. This reduces electrical path length from the ASIC to the laser from centimeters to millimeters, saving power and improving signal integrity at very high speeds. Coherent supplies the photonic components - CW lasers, external laser source modules, fiber array units - that go into CPO assemblies. The NVIDIA partnership, announced March 2026, covers CPO components for NVIDIA's next-generation GPU interconnect architectures. Scale-out CPO revenue is expected in H2 calendar 2026; scale-up CPO (for intra-cluster GPU-to-GPU interconnect, the highest-value application) in H2 2027.

Coherent Optical Modules for Telecom: Used in wavelength division multiplexing (WDM) systems for long-haul fiber networks and metro networks. These use coherent detection (phase and amplitude modulation) to transmit hundreds of gigabits per second over a single wavelength across thousands of kilometers. This is a different, more complex product than data center transceivers and is sold to telecom equipment vendors like Cisco, Nokia, and Ciena.

Silicon Carbide Products

Coherent produces SiC substrates in both standard (undoped) and n-type (nitrogen-doped) varieties, and also produces epitaxial wafers where a thin SiC layer is grown on top of the substrate with controlled electrical characteristics. Wafer sizes are 150mm (six inch) and 200mm (eight inch). The 200mm format is critical for reducing cost per die in SiC power semiconductor production - larger wafers mean more chips per run. Coherent began shipping 200mm SiC epi-wafers (350 micron and 500 micron thickness) in 2025. The customers for these wafers include DENSO and Mitsubishi Electric under long-term supply agreements, plus other power semiconductor device makers.

Industrial Lasers

From the original Coherent, Inc. heritage: precision laser systems for OLED manufacturing (including excimer lasers used for LTPS - low-temperature polysilicon - processing in OLED panel fabs), laser-based annealing and scribing equipment, medical laser systems (ophthalmic and dermatology applications), and scientific research lasers. These products require deep application knowledge - knowing not just the photon physics but the exact process requirements of the customer's manufacturing line.

Manufacturing Footprint

Coherent operates more than 20 manufacturing sites in the United States across 13 states, reflecting both organic growth and the accumulation of acquired facilities. Key facilities:

• Sherman, Texas: The former Finisar flagship. Home of the world's first 6-inch InP production line. VCSELs, EMLs, CW lasers, and SiC production all operate here.
• Yarfala, Sweden: Second InP fabrication site. Activated for 6-inch InP production in September 2025. Provides geographic and geopolitical redundancy.
• Sunnyvale, California: R&D and some transceiver assembly.
• Multiple European sites: Being consolidated as part of the ongoing site-exit program (33 sites exited over six quarters as of Q2 FY26).

The dual-site (Texas + Sweden) InP manufacturing capability is a deliberate risk-management decision. If one site has a yield issue or a facility problem, production can partially shift. From a geopolitical standpoint, European manufacturing provides supply chain diversification for customers who are concerned about concentration in a single country.

4. Customers

Who Buys and Why

AI Hyperscalers (Google, Meta, Microsoft, Amazon): These are the volume purchasers of optical transceivers and the emerging buyers of OCS systems. They do not buy directly from Coherent in most cases; they buy from networking OEMs like Cisco, Arista, and Nokia or from their own supply chain operations. But they are the end demand signal, and their data center expansion plans drive multi-year visibility into transceiver demand. Several hyperscalers have provided Coherent with customer-specific long-term forecasts extending to 2028, which is extraordinary visibility for a hardware company.

NVIDIA: As of the March 2026 partnership announcement, NVIDIA is now both an equity holder ($2 billion investment) and a contracted buyer of CPO components. The relationship covers high-power CW lasers, external laser source modules, and fiber array units for NVIDIA's next-generation GPU interconnect architectures. The supply agreement runs through the end of the decade and includes minimum purchase commitments. This is the highest-profile customer relationship Coherent has disclosed.

Telecom Equipment Vendors (Cisco, Nokia, Ciena, Huawei - where permitted): These companies integrate Coherent's coherent optical modules and transceiver components into routers, switches, and WDM line systems for telecom carriers. Sales cycles are long (12-24 months from design-in to volume production), qualification requirements are strict (telco-grade reliability demands 15+ year product lifespans), and switching costs are high once a module is designed into a line card.

Apple: Multi-year VCSEL supply agreement for consumer electronics applications (Face ID and other sensing functions). This is a concentrated, high-volume, relatively captive relationship - Apple designs its sensors around specific VCSEL specifications and qualifies a supplier through a lengthy testing process. Revenue is expected in H2 calendar 2026 from the next-generation product.

DENSO and Mitsubishi Electric: Anchor customers for SiC substrates, having invested $1 billion in the SiC subsidiary in exchange for supply agreements. They are locked into Coherent as a strategic supplier, not just a commodity vendor, because the equity relationship aligns incentives on both sides.

Semi-Capital Equipment Makers: Buyers of Coherent's OLED and display manufacturing laser systems. This is a project-based market; orders come in when a panel maker decides to build or upgrade a fab. AMAT, TEL, and other equipment makers integrate Coherent laser modules into their systems, or Coherent sells directly to the panel fab.

Switching Costs

At the transceiver module level, switching costs are moderate. Transceivers are pluggable and follow industry-standard form factors (QSFP-DD, OSFP) and protocols (IEEE 802.3). A hyperscaler buying 800G transceivers from Coherent can, in principle, switch to Lumentum or Innolight at the next procurement cycle. What actually creates stickiness:

• Qualification testing: Data center operators run extensive thermal, power, and protocol testing before deploying a new vendor's transceivers at scale. The qualification cost (engineering time, potential downtime risk) is real.
• Long-term agreements with financial commitments: The LTAs Coherent has been signing include upfront customer investment, minimum demand guarantees, and supply commitments. Once a customer has pre-paid for capacity and committed to minimums, switching is contractually expensive.
• InP component supply: For customers who buy Coherent's EML or CW laser chips as components (rather than finished transceivers), switching requires qualifying a new chip supplier, which in compound semiconductors can take 12-18 months.
• OCS integration: Deploying OCS requires redesigning the data center network architecture and qualifying the transceiver lineup that works with the switch. Once an OCS is deployed, the customer is committed to the technology for the data center's operating life.

Concentration

Customer concentration is a real risk. The company does not name customers publicly in most cases, but the pattern is clear: a handful of hyperscalers drive the majority of the data center transceiver revenue. The book-to-bill ratio in the data center segment exceeded 4x in Q2 FY26, meaning Coherent was booking more than four orders for every one it shipped - a signal of both extraordinary demand and limited ability to scale supply fast enough. If any single hyperscaler paused or reduced its AI infrastructure buildout, the effect on Coherent's revenue would be material.

5. Competitive Landscape

Optical Transceivers and InP Components

Innolight (ZhongJi InnoLight): The most important competitor to understand. Founded in 2008 in China with US venture capital, Innolight has grown into arguably the highest-volume transceiver assembler in the world for the data center market. Its strategy is to buy components (including lasers and chips) from suppliers including Coherent and assemble them at lower cost. Innolight does not have its own InP laser fabrication, which means it is both a customer of Coherent's component business and a competitor in the module business. The geopolitical dynamic matters here: US export control policy could restrict Innolight's access to US-origin InP chips, which would simultaneously hurt Coherent's component sales and benefit Coherent's module sales.

Lumentum: US-listed, San Jose-based. Strong in EMLs, VCSELs, and ROADM (reconfigurable optical add-drop multiplexer) components for telecom. Lumentum is the one competitor that management implicitly acknowledges in the context of 1.6T transceivers - it was described in third-party analysis as "the only supplier currently shipping 200G-per-lane EMLs at volume" as of early 2026. This is the specific point of near-term competition: if Lumentum has a head start on 200G EML yield, it may capture early 1.6T market share. Coherent has its own 200G EML and D-EML but the volume ramp timing relative to Lumentum is not publicly disclosed. Notably, Lumentum was the losing bidder when II-VI acquired Coherent Inc. in 2022 - there is competitive history between these management teams.

Broadcom: Has an optical interconnect business, primarily silicon photonics-based, and also sells complete ASIC + optics integrated systems. Broadcom is more a systems competitor in specific architectures than a direct transceiver-for-transceiver competitor.

Accelink: Chinese, state-connected, broad portfolio of optical components and transceivers. Serves both domestic Chinese hyperscalers and some global customers. Less technologically differentiated than Coherent or Lumentum at the high end.

Applied Optoelectronics (AAOI): Much smaller US competitor focused on data center and cable/telco transceivers. Primarily a module assembler, not vertically integrated.

Fabrinet: Not a competitor - Fabrinet is a contract manufacturer that assembles optical components for Lumentum and others. It is relevant as a proxy for the gross margin that is available to an assembler versus a vertically integrated manufacturer like Coherent.

Silicon Carbide

Wolfspeed: The US's largest dedicated SiC wafer and device company. Notably, Wolfspeed declared bankruptcy in 2025, creating significant uncertainty about its supply continuity. This should be a tailwind for Coherent's SiC business as automotive Tier 1s seek to qualify alternative suppliers.

ROHM (Japan): Strong in SiC power device manufacturing; less focused on substrate supply. Not a direct substrate competitor.

STMicroelectronics: European, primarily a SiC device maker; sources substrates from multiple suppliers.

Onsemi: US, primarily SiC devices, not substrates.

Coherent's SiC position is defensible primarily through the DENSO and Mitsubishi anchor customer structure and the 200mm wafer capability. The $1 billion strategic investment is not just capital - it is a public signal to the automotive supply chain that Coherent is a qualified, serious, and financially stable substrate supplier.

Industrial Lasers

TRUMPF (Germany, private): The world's largest industrial laser company. Dominates the sheet metal cutting and welding laser market where Coherent has limited presence. Coherent's industrial laser business is largely non-overlapping with TRUMPF's volume market.

IPG Photonics: US-listed fiber laser company. Does not compete in the specific OLED manufacturing or excimer laser niches where Coherent is strongest.

Coherent's industrial laser moat is largely in applications (OLED display manufacturing, medical precision lasers, scientific instruments) that require deep customer co-development and are small enough not to attract TRUMPF-scale competition.

Barriers to Entry

For the InP laser chip business: extremely high. You need crystal growth equipment (molecular beam epitaxy or metal-organic chemical vapor deposition reactors), wafer fabrication cleanrooms, decades of process know-how, and customer qualifications. The capital cost to replicate Coherent's Sherman facility from scratch would be measured in billions. China has been trying to build domestic InP capability for years; the yields are not yet competitive.

For transceiver module assembly: much lower. Innolight built a high-volume transceiver business in roughly a decade by buying components from others. The assembly process is automatable and the form factors are standardized.

For OCS: moderate. Liquid crystal technology is known but engineering it for data center reliability, latency, and port counts at 320x320 scale requires substantial IP and production experience. Coherent has a multi-year head start.

6. Industry

What Drives Demand

The demand for high-speed optical transceivers is fundamentally a function of data center bandwidth expansion. AI training and inference require moving enormous volumes of data between GPUs at speeds that copper cannot physically support beyond a few meters. Every GPU cluster expansion is also an optical transceiver expansion. NVIDIA's H100 and Blackwell GPU clusters use 400G and 800G optical ports; the next generation will require 1.6T and beyond.

At a higher level, traffic in hyperscale data centers in North America has grown at 30%+ per year, driven by AI inference workloads (language models, image generation, code assistance) that are inherently bandwidth-intensive because they require distributing computation across thousands of chips.

The telecom portion of the demand equation is different: it is driven by 5G infrastructure buildout, submarine cable expansions, and the continuous upgrade of long-haul fiber networks from 100G to 400G to 800G wavelengths.

Industry Size and Growth

The global optical transceiver market was approximately $14.7 billion in 2025 according to Fortune Business Insights. The AI-specific subsegment was approximately $16.5 billion in 2025 and is projected to reach $26 billion in 2026 - a 57% year-over-year expansion (TrendForce, April 2026). The broader market is projected to grow at approximately 17% CAGR to $46 billion by 2034.

TrendForce identified component shortages - specifically EML chips and CW lasers - as the primary bottleneck on supply, not demand. McKinsey projected that 800G transceiver production would fall 40-60% short of demand through 2027, with 1.6T shortfalls persisting through 2029. This context explains why Coherent's book-to-bill ratio exceeded 4x in the data center segment in Q2 FY26: they are supply-constrained, not demand-constrained.

For SiC, the automotive power semiconductor market is driven by EV penetration, which has slowed growth from initial projections. The 200mm wafer transition is the key near-term milestone that will determine which SiC substrate suppliers thrive.

Global Supply Chain Position

Coherent sits near the top of the data center optical supply chain. Its InP wafer fabrication is upstream of everyone except the raw material suppliers. The chain looks approximately like: raw indium and phosphorus compounds → InP crystal growth (Coherent, and a handful of specialized crystal growers) → wafer fabrication on InP (Coherent, Lumentum, a few others) → chip fabrication into EMLs/CW lasers/VCSELs (Coherent, Lumentum, Mitsubishi, Sumitomo) → transceiver module assembly (Coherent, Innolight, Fabrinet customers) → network switch or server NIC (Cisco, Arista, NVIDIA, Broadcom).

Regulation and Geopolitics

US export controls on semiconductor technology have become a central feature of the competitive landscape. The US Department of Commerce has progressively tightened restrictions on what optical and photonic components can be shipped to China. InP chips, as compound semiconductors with dual-use potential, are subject to export license requirements for certain end uses. This creates both risk (if a large portion of Coherent's component customer base is in China or assembles for Chinese end markets) and opportunity (if Chinese competitors are cut off from US InP technology, the addressable market for US-manufactured transceivers grows).

Coherent's geographic manufacturing diversification - specifically the Sherman, Texas plus Yarfala, Sweden dual-site structure - is partly a response to this geopolitical reality. US-manufactured optical components carry implicit supply chain security value for US hyperscalers who face congressional scrutiny over reliance on Chinese-made components.

Cyclicality

The optical transceiver market is cyclical, though the cycles have been compressed by the scale of AI capital expenditure. The 2022-2023 period saw a severe inventory correction as telecom capex collapsed post-COVID overbuild and hyperscalers worked down excess 100G/400G inventory. Coherent's Networking revenue declined significantly in that period. The 2024-2025-2026 period has been the reverse: extraordinary demand pulling on supply across every speed tier simultaneously. The question is whether the current expansion reflects a multi-year structural shift (AI capex remains at elevated levels through the decade) or a cyclical overshoot that will eventually correct.

Management's answer, supported by customer-provided multi-year forecasts and long-term agreements with financial commitments, is that this is structural. The visibility they cite - specific customer forecasts extending to 2028, LTAs with minimum demand guarantees, the NVIDIA equity partnership - is qualitatively different from prior demand signals.

7. Growth Triggers

All triggers sourced directly from the four earnings calls. Quotes included where particularly specific.

From Q4 FY25 Concall (August 13, 2025)

• World's first 6-inch InP production line commissioned at Sherman, Texas in August 2025. Management stated this was "a critical milestone in our capacity expansion roadmap" and that the 6-inch platform produces more than four times as many chips per wafer run at less than half the per-chip cost.

• Apple multi-year VCSEL agreement signed for new-generation Face ID and sensing VCSELs for iPhones and iPads, manufactured at Sherman. Initial revenue expected H2 calendar 2026.

• 1.6T transceiver ramp: Initial 1.6T shipments had begun by Q4 FY25. Management expected the product to ramp through fiscal year-end and into FY2026, initially EML-based.

• OCS initial revenue: Management described customer engagement in OCS as "very strong," with initial revenue flowing and backlog building. (Q4 FY25 concall, August 13, 2025)

• Industrial stabilization: Silicon carbide demand described as "stabilized" after the 2024-2025 correction, with no further headwind expected in FY2026. Industrial laser growth was expected from OLED fab expansion orders. (Q4 FY25 concall, August 13, 2025)

From Q1 FY26 Concall (November 5, 2025)

• 6-inch InP production began at both Sherman AND Yarfala in September 2025. Management disclosed that "initial six-inch yields are actually higher than our current three-inch indium phosphide yields" - an unusually fast yield ramp that implies the learning from Sherman transferred to Sweden within weeks. (Q1 FY26 concall, November 5, 2025)

• InP internal capacity expected to double within 12 months from Q1 FY26 levels. This was the specific commitment made to investors. (Q1 FY26 concall, November 5, 2025)

• OCS TAM stated at $2 billion; seven customers shipped (64x64 and 320x320 configurations), with sequential revenue and backlog growth. (Q1 FY26 concall, November 5, 2025)

• CPO: Large purchase order secured from "a market-leading AI data center customer" for high-power CW lasers made on 6-inch InP. Initial CPO revenue expected late calendar 2026, "more significant contribution next calendar year and beyond." (Q1 FY26 concall, November 5, 2025)

"The bookings are at a record level and the visibility of the business is the best it's ever been. We have large customers providing visibility three years out." - CEO Jim Anderson, Q1 FY26 concall, November 5, 2025

• VCSEL 1.6T: Silicon photonics and EML 1.6T ramping immediately; VCSEL-based 1.6T expected mid-calendar 2026. (Q1 FY26 concall, November 5, 2025)

From Q2 FY26 Concall (February 4, 2026)

• InP wafer starts quadrupled from September to December quarter. By Q2 FY26, Coherent had already achieved 80% of its target capacity doubling. (Q2 FY26 concall, February 4, 2026)

"Six-inch wafers produce more than four times as many chips at less than half the cost" compared to three-inch - CFO Sherri Luther, Q2 FY26 concall, February 4, 2026

• Book-to-bill ratio exceeded 4x in the data center segment. Management stated "bookings extending into calendar 2027" and "some customer forecasts reaching into 2028." No comparable company has disclosed such a forward demand signal. (Q2 FY26 concall, February 4, 2026)

• New LTAs signed with upfront customer investment, supply commitments, and minimum demand guarantees. Management described these as "significant in size" but did not disclose customers. (Q2 FY26 concall, February 4, 2026)

• OCS backlog increase with 10+ customer engagements. Majority of backlog weighted toward larger 320x320 systems. "Revenue expected to ramp throughout 2026 and beyond." (Q2 FY26 concall, February 4, 2026)

• Industrial recovery: Sequential growth expected starting Q4 FY26 (June 2026 quarter), primarily from increased semi-cap equipment orders. (Q2 FY26 concall, February 4, 2026)

From Q3 FY26 Concall (May 6, 2026)

• NVIDIA $2 billion equity investment and multiyear CPO supply agreement announced (March 2026). Agreement covers high-power CW lasers, external laser source modules, and fiber array units through end of decade. Minimum purchase commitments included.

"The NVIDIA partnership...is one of the most strategically significant agreements in our company's history." - CEO Jim Anderson, Q3 FY26 concall, May 6, 2026

• InP capacity doubling achieved one quarter ahead of schedule (by Q4 FY26, not Q1 FY27 as originally guided). Plan to more than double capacity again by end of calendar 2027. (Q3 FY26 concall, May 6, 2026)

• Scale-out CPO revenue expected H2 calendar 2026; scale-up CPO in H2 calendar 2027. (Q3 FY26 concall, May 6, 2026)

• OCS TAM expanded to $4 billion+ (doubled from $2B cited in Q1 FY26) due to expanding use cases. More than 10 shipping customers as of Q3 FY26. (Q3 FY26 concall, May 6, 2026)

• Record backlog with orders extending into calendar 2028. Management stated the company is "bringing on substantially more capacity over the coming quarters." (Q3 FY26 concall, May 6, 2026)

• FY2027 growth expected to exceed FY2026 growth rates - an explicit multi-year acceleration statement from management. (Q3 FY26 concall, May 6, 2026)

8. Key Risks

1. Hyperscaler Capex Concentration and Volatility

Mechanism: Coherent's data center revenue is driven almost entirely by capital spending decisions at a small number of hyperscalers (Google, Meta, Microsoft, Amazon) and NVIDIA (for GPU infrastructure). These companies collectively set their annual capex budgets and can revise them materially in any quarter. If one large hyperscaler decides to pause, delay, or reprioritize its AI infrastructure spend - due to macroeconomic pressure, a product cycle delay, or a shift in AI architecture (for example, if inference efficiency improves dramatically and reduces per-query compute requirements) - Coherent's revenue could decline sharply and quickly. The company's book-to-bill ratio above 4x and orders extending to 2028 provide visibility, but visibility can change if customers invoke cancellation clauses or reschedule deliveries. The 2022-2023 telecom inventory correction is the relevant precedent: what seemed like durable demand collapsed over six months when customers discovered they had over-ordered.

Calibration: High-probability moderate drag risk in a normalization scenario; low-probability catastrophic risk if AI capex abruptly reverses. Given the stated LTAs with financial commitments and the NVIDIA equity partnership (which aligns NVIDIA's financial interests with Coherent's success), the risk is partially mitigated but cannot be eliminated.

2. Execution Risk on InP Capacity Ramp

Mechanism: Coherent has committed to more than tripling its InP manufacturing capacity over roughly two years. This requires running 6-inch InP production at dual sites (Texas and Sweden) at sustained high yields, hiring and training compound semiconductor process engineers (a scarce skill set), integrating new equipment, and qualifying the resulting chips with customers whose specifications do not change. Any yield degradation at the 6-inch line (contamination, crystal growth defects, process drift) would not just reduce output - it would generate a revenue shortfall precisely when customers are expecting supply they have pre-committed to. Compound semiconductor manufacturing is notably harder to scale than silicon CMOS; there is no equivalent to the semiconductor equipment ecosystem's mature playbook for InP 6-inch wafers, because Coherent is literally doing this first.

Management's words at Q3 FY26: "We are bringing on substantially more capacity over the coming quarters." The flip side of this confidence is that if the ramp stumbles, there are no alternative InP suppliers at the scale required.

Calibration: Medium probability, potentially severe impact on both revenue and customer relationships if it materializes.

3. Chinese Competition and Geopolitical Policy Uncertainty

Mechanism: Chinese transceiver assemblers like Innolight have demonstrated the ability to compete at price in the 400G market and are moving aggressively into 800G. If they develop credible domestic InP laser chip capability (currently their key gap), they would no longer need to buy components from Coherent and could undercut Coherent's module prices significantly. The US government's export control policies could cut both ways: tighter controls on InP components would hurt Coherent's component sales to Chinese assemblers (which are currently a revenue stream), while helping Coherent's module sales to US hyperscalers who prefer US-supply-chain assurance. Policy could also become more permissive, negating the supply chain security premium.

Calibration: Long-term structural risk on the competitive dimension; near-term policy uncertainty on export controls creates real quarter-to-quarter unpredictability for the component business.

4. Technology Substitution - CPO Disrupting Pluggable Transceiver Revenue

Mechanism: If co-packaged optics (CPO) adoption accelerates ahead of schedule - particularly for scale-up interconnect in GPU clusters - it could cannibalize the pluggable 800G and 1.6T transceiver market that is currently Coherent's largest revenue line. Coherent is positioning for this by developing CPO components for NVIDIA, but the timing mismatch is real: pluggable transceiver revenue is the business today; CPO revenue is the business in 2027 and beyond. If the transition happens faster than expected, Coherent would need to ramp a new product line while its existing revenue streams are under pressure from architecture change.

Calibration: Low probability in the next 24 months (the current consensus industry view is that pluggable transceivers dominate through at least 2027), but a genuine structural shift beyond that horizon.

5. Leverage and Capital Allocation Risk

Mechanism: While Coherent has dramatically reduced its debt leverage ratio (from 2.4x to 0.5x over the past year), the company has accumulated this debt through a $7+ billion acquisition (Coherent, Inc.) that has not yet reached its full earnings potential. The NVIDIA equity investment has helped cash and leverage, but it also represents equity dilution. Simultaneously, the company is spending $290 million per quarter in capex (Q3 FY26) to fund the InP capacity expansion. If revenue growth slows while capex commitments remain high, free cash flow could come under pressure and force choices between investing in capacity, paying down remaining debt, or meeting investor expectations.

Calibration: Currently low risk given the 0.5x leverage and $3 billion cash position, but a revenue deceleration scenario could make this more acute.

6. Silicon Carbide Cyclicality and EV Demand Risk

Mechanism: SiC substrate demand is linked to EV production volumes and power semiconductor content per vehicle. The 2024-2025 correction was driven by slower-than-expected EV adoption growth (particularly in Europe and China) and excess inventory in the supply chain. If this pattern repeats - EV demand disappointment or structural overcapacity in SiC substrates - the Industrial segment could become a drag on overall margins. The DENSO/Mitsubishi anchor customer structure provides some floor, but it does not protect against a broad market correction.

9. Walk the Talk

Concall dates used: Q4 FY25 (August 13, 2025), Q1 FY26 (November 5, 2025), Q2 FY26 (February 4, 2026), Q3 FY26 (May 6, 2026).

The Q3 FY26 call took place May 6, 2026 - four days before this report. Coverage through the current quarter is complete.

From Q4 FY25: Setting Up the Narrative

On August 13, 2025, CEO Jim Anderson and CFO Sherri Luther guided for Q1 FY26 revenue of $1.46 billion to $1.60 billion and non-GAAP EPS of $0.93 to $1.13. They described the world's first 6-inch InP production line beginning in August 2025, the Apple VCSEL agreement, and the A&D divestiture closing in Q1 FY26. They were constructive on data center demand continuing and specifically said silicon carbide demand had stabilized.

The Q1 FY26 outcome: revenue of $1.58 billion (within range, near the top) and EPS of $1.16 (above the top of the guided range). The A&D divestiture closed on schedule and the proceeds were used for a $400 million debt paydown, exactly as guided. Silicon carbide demand did stabilize without material headwinds. The 6-inch line at Sherman began production in August, exactly as announced.

Score: Guidance range hit, EPS beat, strategic actions delivered on schedule.

From Q1 FY26: Accelerating Confidence

On November 5, 2025, management guided Q2 FY26 revenue of $1.56 billion to $1.70 billion and EPS of $1.10 to $1.30. They made a specific commitment: InP internal capacity would double within 12 months from Q1 FY26. They disclosed that 6-inch InP production had begun at both Sherman AND Yarfala in September, and notably that initial 6-inch yields were already exceeding 3-inch yields - a remarkable statement that set a high bar for validation.

The Q2 FY26 outcome: revenue of $1.69 billion (within range, near the top) and EPS of $1.29 (within range, near the top). By Q2 FY26, wafer starts had quadrupled from September to December, and management stated that 80% of the capacity doubling had already been achieved by February 2026 - four months into the 12-month commitment. The yield claim from Q1 was validated in Q2 with the statement that 6-inch yields exceeded 3-inch across all device types at both sites.

"Achieving 80% of our capacity doubling target by the end of Q2 puts us well ahead of the pace needed to complete the doubling by the end of Q3." - CFO Sherri Luther, Q2 FY26 concall

Score: Guidance range hit, strategic execution ahead of stated plan. No dropped commitments.

From Q2 FY26: Raising the Bar

On February 4, 2026, management guided Q3 FY26 revenue of $1.70 billion to $1.84 billion and EPS of $1.28 to $1.48. They disclosed a "significant" CPO purchase order from a market-leading AI customer and stated that OCS backlog had increased with over 10 customer engagements. They guided for InP capacity doubling to complete within the fiscal year (by Q3 or Q4 FY26). They also indicated that the Industrial segment would begin sequential recovery in Q4 FY26 (June quarter).

The Q3 FY26 outcome: revenue of $1.81 billion (within range, near top) and EPS of $1.41 (within range). More pointedly, management stated at the Q3 call that the capacity doubling was completed one quarter ahead of plan - in Q3, not Q4 as guided. The NVIDIA partnership (announced March 2026) was a positive surprise not guided. OCS reached 10+ shipping customers. The CPO revenue timeline (late 2026) remained unchanged.

Score: Guidance hit, capacity milestone delivered early, unexpected positive (NVIDIA) disclosed. No commitments missed.

Q3 FY26 → Forward Guidance

Management now guides Q4 FY26 revenue of $1.91 billion to $2.05 billion and EPS of $1.52 to $1.72. They stated explicitly that FY2027 growth would exceed FY2026 rates. These are aggressive guidance numbers, representing 7-13% sequential growth from Q3's record quarter. The basis for confidence is the NVIDIA supply agreement, accelerating InP capacity, and record backlog extending to 2028.

Verdict

Jim Anderson's Coherent is a management team that guides conservatively and executes ahead of pace. Across four sequential quarters, every financial guidance range was hit (and EPS beat at least once). Every strategic commitment (A&D sale close, site exits, 6-inch InP activation, InP capacity doubling) was delivered on or ahead of schedule. The 6-inch yield claim from Q1 FY26 - probably the most aggressive single statement made - was fully validated by Q2 FY26. The NVIDIA equity partnership was a genuine positive surprise, not a planted expectation.

The pattern is classic conservative-to-positive guidance: management consistently guides with a range they can clear, then performs at the high end or above. For an investor using guidance as a baseline, this management team has demonstrated they can be taken at their word. There are no dropped commitments visible in the four-quarter record.

10. Shareholder Friendliness Index

Dividends: Coherent does not pay a dividend. The company suspended its historical dividend when the capital requirements of the Finisar and Coherent acquisitions made sustaining it imprudent. There has been no statement of intent to reinstate a dividend within the visible planning horizon. Capital is being directed toward debt reduction and capacity investment.

Buybacks and Share Count: Coherent has not executed share buybacks in the covered period. The company has been a net user of capital, not a returner: it paid down approximately $437 million in debt in FY2025 and an additional $400 million in Q1 FY26 (proceeds from the A&D divestiture). The NVIDIA $2 billion equity investment in Q3 FY26, while materially strengthening the balance sheet, represents equity issuance (new shares to NVIDIA) rather than a buyback. The diluted share count has been growing modestly from option and RSU grants to employees, and the NVIDIA investment added further equity dilution. At the same time, leverage has collapsed from over 2x to 0.5x in one year, which is a form of de-risking the balance sheet that benefits shareholders even if it is not capital return.

Verdict: Hoards Capital - but for the right reasons. The company is in a capital investment cycle (over $290 million in quarterly capex), deleveraging from a debt-heavy acquisition history, and funding a once-in-a-decade capacity expansion. The NVIDIA equity investment, while dilutive, materially de-risked the balance sheet. Capital return is not on the near-term roadmap.

11. Insider Activities

Source: SEC Form 4 filings via EDGAR and OpenInsider (openinsider.com/COHR). Last 12 months: May 2025 - May 2026.

Recent Transactions (Most Recent First)

Buys - Reading the Signal

No open-market insider purchases were identified in the 12-month period. No director or executive paid cash out of pocket to buy Coherent shares on the open market. The absence of insider buying is notable given the significant appreciation in the stock price (from roughly $90 in August 2025 to over $350 in April 2026) - management has participated through option gains but has not put new personal capital into the company.

Sells - Working Out the Why

Bain Capital - $1.075 billion secondary (November 7, 2025): Bain became a major shareholder through the II-VI acquisition of Coherent Inc. (Bain was a PE sponsor of the original Coherent). A $1 billion-plus secondary block sale by a PE firm three years post-acquisition is a routine exit, not a negative signal about the company's prospects. PE funds have return timelines and LP redemption obligations that are unrelated to business outlook. This is not a red flag.

Giovanni Barbarossa (EVP General Management) - ~$18 million in two days (October-November 2025): Barbarossa exercised legacy stock options at prices of $35.25, $49.90, and $36.56 per share and immediately sold the resulting shares at roughly $140-$160. This is a textbook cashless exercise: the options had years of accumulated gain and were approaching expiration or the executive was diversifying. The sale prices were far above the exercise prices. This is dilutive in the sense of new shares entering the market, but it is not a bet against the company.

CFO Sherri Luther: Two sales totaling 6,000 shares under a 10b5-1 plan filed in November 2025. A CFO with over 70,000 shares remaining who is selling 6,000 shares per quarter under a pre-planned program is engaging in ordinary financial planning, not expressing a negative view.

Director Howard Xia: Pattern of small, incremental sales across three dates totaling 7,240 shares. Consistent with diversification behavior at director compensation levels.

Net Assessment

Insiders are uniformly net sellers in the 12-month period, but the character of the selling is almost entirely explainable by routine financial behavior: PE sponsor exit, cashless option exercises, scheduled 10b5-1 programs. The Bain Capital block sale was expected by anyone who tracks PE holding periods. The absence of open-market purchases is a mild negative signal - no one is putting fresh personal capital on the line - but it is not unusual in a period where the stock has tripled from its lows and management is compensated heavily through options already. Neutral signal overall.

12. Scenarios

Bull Case

AI infrastructure spending does not follow a normal capex cycle. It turns out to be a decade-long buildout comparable to the build-out of mobile networks in the 2000s, because inference demand from AI applications compounds as more models deploy and user adoption grows. In this world, every successive generation of GPU cluster requires more transceivers at higher speeds: 800G transitions to 1.6T, 1.6T to 3.2T, and so on, each requiring more InP laser chips.

Coherent's 6-inch InP platform turns out to be a genuine moat. No competitor manages to replicate the yield economics within the critical 2026-2028 window when supply is so constrained that customers are signing multi-year LTAs with minimum demand guarantees just to secure allocation. Coherent triples InP capacity as announced, becomes the dominant supplier of the critical component inside most of the world's AI data center transceivers, and captures the economics of being the Intel of InP lasers.

The OCS market inflects faster than expected as hyperscalers redesign next-generation AI clusters specifically around optical circuit switching, eliminating billions of dollars of electronic switch hardware and power cost. Coherent's 320x320 OCS becomes the standard fabric for scale-up GPU clusters. The addressable market expands past management's $4 billion estimate.

CPO arrives on schedule in 2027 for scale-up applications. NVIDIA's endorsement - an equity stake plus a decade-long supply agreement - makes Coherent the default CW laser supplier for CPO and validates the technology transition for other GPU makers and hyperscalers. The combination of transceiver revenue, OCS revenue, and CPO component revenue makes Coherent a multi-vector beneficiary of every layer of the AI optical infrastructure stack.

The Apple VCSEL relationship deepens as Apple expands sensing applications beyond Face ID (spatial computing, health monitoring), creating a growing, captive consumer electronics revenue stream that partially offsets any cyclicality in data center demand.

Base Case

Management delivers roughly what it has guided. InP capacity more than doubles by end of 2027, CPO revenue ramps in H2 2026 and becomes material in 2027, OCS grows steadily from a small base, and the telecom recovery continues through 2026. The 1.6T transceiver ramp in 2026 follows 800G as the workhorse product, providing the sequential growth engine management has described.

The Industrial segment remains a modest drag relative to the data center business: SiC demand is stable but not spectacular as EV adoption normalizes, and industrial laser revenue grows modestly with OLED fab expansion. The divestitures of A&D and the Munich division improve margins without disrupting the core business.

Competition from Innolight and other Chinese assemblers intensifies in the 800G market, compressing module margins slightly. But Coherent's vertical integration means it still earns full value on the InP components embedded in competitor modules, partially offsetting the module-level price pressure. The NVIDIA partnership provides demand visibility that insulates capacity planning decisions from normal demand uncertainty.

The company ends FY2027 as a significantly larger, more profitable, and more focused business than it was in FY2025, with leverage near zero and cash to redeploy toward either capacity growth or shareholder return.

Bear Case

The hyperscaler capex cycle does correct. Not catastrophically, but materially. One or two major hyperscalers announce they are pausing their AI data center expansion programs for 12-18 months, citing either a slowdown in AI application revenue growth or a shift to a new compute architecture (for example, if a dramatically more efficient inference chip reduces per-query GPU requirements). Coherent's backlog, which is concentrated in a small number of large customers, declines rapidly as customers invoke force majeure or reschedule provisions in their LTAs.

At the same time, the InP capacity expansion - funded by heavy capex commitments and partially by the NVIDIA equity - results in a supply glut relative to revised demand. Coherent has built capacity for a demand level that does not materialize on the expected timeline. Margins compress as the company fills capacity at lower prices.

The OCS market turns out to be smaller and slower than management expected. A competing optical switching technology (silicon photonics-based or MEMS-based at scale) achieves better economics, or hyperscalers decide that the network redesign cost exceeds the switch hardware savings.

Meanwhile, China develops domestic InP laser chip capability at adequate yields by 2028, reducing Innolight's reliance on Coherent components and accelerating price pressure in the assembled transceiver market. The SiC business faces continued EV-cycle weakness and Wolfspeed, despite its bankruptcy, finds a buyer who aggressively prices to win market share back.

In this scenario, Coherent is still a real business - the photonic infrastructure need is genuine and the company has real technology. But the multiple expansion from 2024-2026 reverses, growth rates normalize to mid-single-digit or low double-digit, and the NVIDIA equity stake looks like a high-price strategic bet made at peak cycle.

13. Further Reading

Searches performed for SemiAnalysis, Stratechery, and MBI Deep Dives for Coherent Corp. coverage.

• Co Packaged Optics (CPO) - Scaling with Light for the Next Wave of Interconnect - SemiAnalysis [paid] - Deep technical analysis of CPO as the central technology for next-generation AI cluster interconnect; Coherent is a named supplier for CPO laser components.

• Marvell's DSP Dilemma? Networking's Tectonic Shift Led By Broadcom, Nvidia, Arista Networks, Microsoft, Meta, Macom, and more - SemiAnalysis, 2023 [paid] - Older article on optical networking architecture shifts; context on the DSP vs. linear drive debate that shapes transceiver design choices relevant to Coherent's product roadmap.

No coverage found from Stratechery or MBI Deep Dives at the time this report was generated.

Sources:

• Coherent (COHR) Q3 2026 Earnings Transcript - The Motley Fool
• Coherent (COHR) Q2 2026 Earnings Call Transcript - The Motley Fool
• Coherent (COHR) Q1 2026 Earnings Call Transcript - The Motley Fool
• Coherent Q4 FY2025 Earnings Transcript - Benzinga
• Coherent Corp. 2025 Annual Report
• NVIDIA and Coherent Strategic Partnership Announcement - NVIDIA Newsroom
• Coherent Corp. - Wikipedia
• Global AI Optical Transceiver Market - TrendForce April 2026
• Coherent's Silicon Carbide Business Receives $1 Billion - Coherent IR
• COHR Insider Transactions - SecForm4
• Optical Circuit Switch - Coherent Product Page
• SemiAnalysis CPO Article