$COHR KEY READ-THROUGHS FROM COHERENT CORP Q3 FY26 EARNINGS CALL
Coherent’s Q3 FY26 call was a high-signal event for AI optical infrastructure, datacenter interconnect, compound semiconductor capacity, CPO architecture, OCS adoption, hyperscaler capex behavior, and thermal/power bottlenecks. The core message was that AI-related optical demand is accelerating faster than industry supply, with record bookings, record backlog, orders extending into calendar 2028, customer LTAs extending to the end of the decade, and capacity expansion becoming the primary constraint. The call was structurally positive for AI optics, CPO, coherent transport, InP components, and advanced datacenter power/thermal infrastructure, while also raising negative implications for copper-centric interconnect exposure, non-integrated transceiver suppliers, and free-cash-flow conversion across the optical supply chain. The most important read-through is that the AI infrastructure cycle is moving deeper into enabling bottlenecks beyond GPUs: optical bandwidth, InP capacity, OCS, CPO, DCI, cooling, and power efficiency are becoming strategic control points. 
AI ACCELERATOR AND CPO ARCHITECTURE
NVIDIA OPTICAL SUPPLY LOCK-IN VALIDATES CPO AS A STRATEGIC AI ARCHITECTURE BOTTLENECK (READ-THROUGH 1)
Affected company: NVIDIA Corporation (NVDA: US)
Directional impact and magnitude: Positive, medium near-term; high long-duration strategic impact. The direct financial impact on NVIDIA is diluted by NVIDIA’s scale, but the strategic implication is highly significant because optics is increasingly central to the scalability of future AI datacenter architectures.
Supporting call evidence: Coherent stated that its NVIDIA partnership includes “NVIDIA’s $2 billion equity investment in Coherent and a multi-year supply agreement extending through the end-of-the decade.” Management added that the agreement covers “multiple CPO-related products, including our high-power CW laser” and that CPO expands Coherent’s role in AI datacenter architectures, “particularly in the scale-up portion of the network where optics is expected to increasingly complement and over-time displace copper.” In Q&A, management said NVIDIA is “clearly probably our lead customer on CPO,” while also noting that multiple other customers are engaged.
Transmission mechanism: NVIDIA benefits from securing advanced optical supply for future AI scale-out and scale-up architectures. The supply agreement and equity investment reduce a potential bottleneck in high-power CW lasers, external laser source modules, fiber attach units, and other CPO-related optical assemblies. The mechanism is not near-term EPS accretion; it is ecosystem control, supply assurance, and accelerated migration from copper-constrained architectures toward optics-enabled architectures. This supports NVIDIA’s ability to scale larger AI clusters with lower power, higher bandwidth, and improved network reach.
Near-term trading catalyst versus long-duration shift: Near term, the read-through supports confidence in NVIDIA’s AI infrastructure roadmap and benefits optical suppliers levered to NVIDIA’s architecture. Long term, it reinforces a structural transition in which NVIDIA’s platform strategy increasingly extends beyond GPUs into networking, optics, and physical infrastructure. The call supports the view that NVIDIA is preemptively securing upstream capacity in critical bottleneck categories.
Key caveat: The call does not quantify NVIDIA-specific purchase volumes, pricing, revenue timing by product, or CPO attach rates. The read-through is strongest strategically and weaker as a near-term NVIDIA earnings driver.
BANDWIDTH ESCALATION SUPPORTS AI NETWORKING SILICON, CUSTOM ASIC, AND DSP DEMAND (READ-THROUGH 2)
Affected companies: Broadcom Inc. (AVGO: US), Marvell Technology, Inc. (MRVL: US)
Directional impact and magnitude: Positive, medium-to-high. The signal is strongest for long-duration AI networking revenue and high-speed connectivity content. Near-term impact is positive but indirect because the call is from an optical supplier, not a silicon vendor.
Supporting call evidence: Coherent said 800G revenue is expected to grow YoY in calendar 2026, while “1.6T transceivers ramp rapidly through the balance of this calendar year and into next year.” Management also said 1.6T is “ramping at an incredibly rapid pace” and that 800G will “probably grow again next calendar year.” Coherent also highlighted a path to 3.2T using 400G-per-lane silicon photonics and 400G differential EMLs.
Transmission mechanism: Sustained 800G growth alongside rapid 1.6T adoption implies expanding demand for high-speed switch silicon, DSPs, SerDes, retimers, PHYs, custom AI networking ASICs, and optical-electrical interface silicon. Broadcom and Marvell benefit from the same bandwidth scaling vector that is driving Coherent’s optical demand. The positive read-through is not only unit growth; it is mix upgrade toward higher-speed, higher-complexity, higher-content networking silicon.
Near-term trading catalyst versus long-duration shift: Near term, the read-through supports revenue durability for AI networking and optical DSP exposed franchises. Long term, 3.2T, CPO, and near-package optics imply a more complex connectivity architecture that can expand silicon content even if some discrete module economics shift. The call is supportive of a multi-year high-speed networking upgrade cycle rather than a single 800G-to-1.6T replacement cycle.
Key caveat: Coherent did not identify Broadcom or Marvell as customers or suppliers. The read-through is architecture- and content-driven, not relationship-specific.
HYPERSCALER AND AI DATACENTER OPERATORS
HYPERSCALERS ARE LOCKING OPTICAL CAPACITY, EXTENDING AI CAPEX INTENSITY AND SUPPLY CHAIN COMMITMENT (READ-THROUGH 3)
Affected companies: Microsoft Corporation (MSFT: US), Alphabet Inc. (GOOGL: US), https://t.co/SpqvHNUxpK, Inc. (AMZN: US), Meta Platforms, Inc. (META: US), Oracle Corporation (ORCL: US)
Directional impact and magnitude: Mixed. Positive for deployment assurance and long-term AI infrastructure scalability; negative for capex intensity and free-cash-flow optics. Magnitude is low for near-term mega-cap EPS but medium for the broader AI capex narrative.
Supporting call evidence: Coherent stated that customer LTAs extend “to the end-of-the decade,” that orders now reach “into calendar 2028,” and that LTAs generally include “an upfront investment from the customer to help with the capex,” a supply commitment from Coherent, and “some sort of demand minimal — at least minimal demand commitment from the customer.” Management said these LTAs include both “hyperscalers as well as other system customers.”
Transmission mechanism: Hyperscalers are moving from short-cycle procurement to strategic supply reservation in optical networking bottlenecks. Upfront customer investments and minimum demand commitments imply that optical supply is now sufficiently scarce to require capital participation from customers. This supports confidence that AI datacenter deployments remain capacity constrained rather than demand constrained. It also suggests that hyperscaler capex intensity and supply chain prepayments may remain elevated for multiple years.
Near-term trading catalyst versus long-duration shift: Near term, the read-through is positive for AI supply chain stocks because hyperscaler demand visibility is improving and suppliers are receiving customer-backed capacity commitments. For hyperscalers, the near-term implication is mixed because stronger infrastructure deployment supports AI product roadmaps but reinforces ongoing capex and cash flow pressure. Long term, the LTAs indicate that the physical AI infrastructure supply chain is becoming strategically locked, with committed capacity extending into calendar 2028 and beyond.
Key caveat: Coherent did not disclose the identity, size, or terms of the hyperscaler LTAs. The read-through is high-conviction at the sector level, but not attributable to any individual hyperscaler from the transcript alone.
OPTICAL COMPONENTS, INDIUM PHOSPHIDE, AND LASERS
INDIUM PHOSPHIDE SCARCITY SUPPORTS PRICING, UTILIZATION, AND STRATEGIC VALUE FOR LASER AND COMPOUND SEMICONDUCTOR SUPPLIERS (READ-THROUGH 4)
Affected companies: Lumentum Holdings Inc. (LITE: US), Sumitomo Electric Industries, Ltd. (5802: Japan), Furukawa Electric Co., Ltd. (5801: Japan), IQE plc (IQE: UK)
Directional impact and magnitude: Positive, medium near-term; medium-to-high long-duration for InP-exposed suppliers. The magnitude is highest for companies with meaningful exposure to EMLs, CW lasers, photodiodes, InP epitaxy, or high-speed optical components.
Supporting call evidence: Coherent described InP as “a constraint for the industry” and said “capacity expansion remains one of our highest priorities.” Management stated that pricing dynamics are “very healthy” because of the “supply versus demand” imbalance. In response to a question about external laser price increases, management said Coherent has been successful either “passing along those external component price — higher prices or offsetting that with our own internal production.”
Transmission mechanism: Industry-wide InP tightness supports higher utilization, healthier pricing, stronger backlog visibility, and strategic value for merchant InP laser and photonic component suppliers. The demand environment is broad across 800G, 1.6T, CPO, DCI, scale-across, and OCS, all of which require high-performance optical components. Suppliers with differentiated InP capability should benefit from scarcity rents and better customer commitment structures.
Near-term trading catalyst versus long-duration shift: Near term, the read-through is positive for merchant laser suppliers because Coherent’s commentary confirms pricing strength and component scarcity. Long term, the read-through is positive for the strategic value of InP assets, but less straightforward for suppliers that compete against vertically integrated customers.
Key caveat: Coherent’s aggressive internal capacity expansion is a long-duration offset. The company expects to double internal InP output by the end of calendar 2026 and more than double it again by the end of calendar 2027. If large optical vendors internalize more InP capacity, merchant suppliers may benefit from industry demand but face share pressure at vertically integrated customers.
VERTICAL INTEGRATION RAISES THE COMPETITIVE BAR FOR NON-INTEGRATED TRANSCEIVER AND MODULE SUPPLIERS (READ-THROUGH 5)
Affected companies: Applied Optoelectronics, Inc. (AAOI: US), Fabrinet (FN: US), Zhongji Innolight Co., Ltd. (300308: China), Eoptolink Technology Inc., Ltd. (300502: China), Accelink Technologies Co., Ltd. (002281: China)
Directional impact and magnitude: Mixed near-term; negative medium-to-long duration for margin structure and competitive positioning of less vertically integrated suppliers. Magnitude is medium overall and potentially high for companies more dependent on externally sourced EMLs, CW lasers, or constrained optical inputs.
Supporting call evidence: Coherent said most components that go into its transceivers are internally sourced, which “buffers us from any increases” in externally sourced component prices. Management stated that 6-inch InP provides “more than four times as many devices at less than half the cost” versus 3-inch and that Coherent allocates InP capacity to “whatever drives the most — the highest-margin dollars.” The company also shipped its first transceivers containing components from 6-inch lines during Q3.
Transmission mechanism: Vertically integrated suppliers with internal InP capacity can control availability, cost, yield learning, and allocation across the highest-margin applications. Non-integrated module suppliers may face input cost inflation, component allocation constraints, or reduced gross margin flexibility when EMLs, CW lasers, and other optical inputs are scarce. Even if end-demand is robust, value capture may migrate upstream toward suppliers that control critical compound semiconductor capacity.
Near-term trading catalyst versus long-duration shift: Near term, demand for 800G and 1.6T transceivers remains positive for module suppliers and contract manufacturers. Long term, Coherent’s 6-inch InP ramp is a competitive escalation. Cost-down and supply control could pressure less-integrated vendors, especially if customers prioritize suppliers with secured InP capacity and multi-year supply commitments.
Key caveat: Fabrinet is more of a manufacturing partner and contract manufacturing exposure than a pure non-integrated transceiver brand. For Fabrinet, the signal is mixed: optical volume growth is positive, but vertical integration by large customers can limit the share of economics available to outsourced manufacturing.
1.6T AND 800G OVERLAP SUPPORTS A LARGER-THAN-EXPECTED OPTICAL TRANSCEIVER REVENUE STACK (READ-THROUGH 6)
Affected companies: Lumentum Holdings Inc. (LITE: US), Fabrinet (FN: US), Applied Optoelectronics, Inc. (AAOI: US), Zhongji Innolight Co., Ltd. (300308: China), Eoptolink Technology Inc., Ltd. (300502: China), Marvell Technology, Inc. (MRVL: US)
Directional impact and magnitude: Positive, medium-to-high for optical transceiver and optical component exposure. The read-through is especially important because it reduces the risk that 1.6T merely cannibalizes 800G.
Supporting call evidence: Management said, “We expect 800 gig to grow this year. It will probably grow again next calendar year. And then on-top of that, 1.6T is ramping at an incredibly rapid pace.” Coherent also said the current-quarter sequential growth includes a “good portion” from the 1.6T ramp and that 1.6T should be beneficial to gross margin because new data rates generally carry better gross margins early in the lifecycle.
Transmission mechanism: If 800G continues growing while 1.6T ramps, the market is not simply replacing one revenue pool with another. Instead, the total optical transceiver revenue stack expands as multiple speed tiers grow simultaneously. This supports higher units, ASP mix, optical DSP content, EML/CW laser demand, and module manufacturing volumes. The positive mechanism is both volume and mix.
Near-term trading catalyst versus long-duration shift: Near term, the read-through supports positive sentiment for suppliers exposed to 800G and 1.6T ramps. Long term, it increases confidence that AI cluster bandwidth requirements can sustain multiple overlapping transceiver cycles, reducing the risk of abrupt speed-generation transitions causing revenue cliffs.
Key caveat: Coherent declined to disclose whether 1.6T revenue exceeded $100 million in Q3 and does not provide data-rate-specific revenue. Product-level opacity remains a sector issue.
OPTICAL TRANSPORT, DCI, AND AI NETWORKING SYSTEMS
DCI AND SCALE-ACROSS ACCELERATION IS POSITIVE FOR COHERENT OPTICAL TRANSPORT AND AI NETWORKING INFRASTRUCTURE (READ-THROUGH 7)
Affected companies: Ciena Corporation (CIEN: US), Arista Networks, Inc. (ANET: US), Cisco Systems, Inc. (CSCO: US), Nokia Oyj (NOKIA: Finland)
Directional impact and magnitude: Positive, medium. The read-through is highest for coherent optical transport and DCI-exposed vendors, and more moderate for diversified networking vendors.
Supporting call evidence: Coherent’s communications revenue increased 16% sequentially and 60% YoY, driven by “data center interconnect, scale across and traditional telecom applications.” Management said the fastest growth is in “scale across” and DCI, with demand across pump lasers, ZR/ZR+ transceivers, line cards, amplifiers, and full systems. Coherent expects strong sequential growth again in the current quarter.
Transmission mechanism: AI workloads are becoming distributed across multiple datacenter locations, creating demand for higher-capacity DCI, coherent optics, ZR/ZR+ modules, transport platforms, and network systems. This benefits vendors supplying coherent optical transport, datacenter interconnect, and AI network infrastructure. The call indicates that the AI bandwidth cycle is not confined to intra-datacenter switch-to-server links; it is also expanding between datacenters.
Near-term trading catalyst versus long-duration shift: Near term, Coherent’s 60% YoY communications growth is a positive datapoint for DCI demand and may support sentiment around optical transport exposure. Long term, scale-across architectures imply a durable buildout of inter-datacenter bandwidth as AI workloads become geographically distributed.
Key caveat: Coherent is also competing higher in the systems layer. The positive read-through is strongest for total market demand, but the competitive benefit across Ciena, Cisco, Nokia, and Arista depends on individual exposure to AI DCI, coherent optics, customer relationships, and product architecture.
