$MS $LITE $COHR $CIEN EXECUTIVE SUMMARY
Morgan Stanley’s “Into the Spotlight: Optical Market Opportunities” frames the optical stack as entering a second phase of AI-driven scaling in which bandwidth growth and physical constraints in electrical interconnects force architectural change, expanding total addressable market (TAM) while simultaneously reshaping profit pools.  The report’s base case sizes optical markets at approximately $30bn in 2025 and $65bn+ by 2028 (approximately 30% CAGR), with an additional approximately $23bn of incremental opportunities tied to new optical technologies (copper-to-fiber in scale-up, co-packaged optics in multiple forms, optical circuit switching, and passive optical out-of-band management), taking the longer-term industry TAM toward approximately $90bn. Central uncertainty is not the direction of travel toward more optics, but the adoption timing of disruptive architectures and the durability of pricing/margins in a market that historically experienced 10-15% annual price declines yet has recently exhibited price stability or increases amid component shortages. The equity market has already heavily discounted sustained tightness and rapid adoption, with the report noting optical-exposed equities up approximately 300%+ over the last 12 months and trading at approximately 30-40x FY28 estimates, implying limited tolerance for disappointment on either demand (AI capex) or pricing (laser/transceiver deflation). The report highlights Corning (GLW) as relatively more attractive on longer-duration exposure to a copper-to-fiber transition that is not fully embedded in near-term estimates, and flags Lumentum (LITE) as most exposed to a mean-reversion outcome in margins and pricing given the magnitude of profitability expansion implied by valuation.
MARKET SIZING AND STRUCTURAL DEMAND DRIVERS
The report’s core framing is a shift in optical demand away from traditional telco/enterprise cycles and toward cloud AI data centers with materially higher bandwidth intensity, faster upgrade cadence, and more aggressive scaling targets. The cloud data center share of the overall optical market is estimated to have increased from 28% in 2021 to 63% in 2025 and is projected to reach 78% by 2030, implying that optical demand, pricing power, and volatility will increasingly be driven by hyperscaler and “neo-cloud” procurement behavior rather than carrier cycles. The current architecture of AI clusters has remained viable through speed upgrades and interim solutions, including NVLink-driven scaling and extensions of copper reach through active electrical cable (AEC) approaches. The report argues that the next leg of scaling will stress multiple “hard” constraints: copper transmission length at higher speeds, lane/laser speed ceilings, front-panel density/backplane limits, and I/O constraints as compute packaging becomes the bottleneck. These constraints are characterized as converging in the 2027-2030 window, with certain categories starting to be tested at scale in 2027-2028 but broad adoption extending across 5-10 years.
A second structural driver is power. Power is positioned as both a constraint within a site (driving densification and forcing higher efficiency interconnects) and as a catalyst for distributed data center buildouts (driving more inter-site traffic). The report notes that historically only approximately 5% of data left a data center, but suggests that could rise to 20-25% in coming years as power and space constraints limit the ability to scale training/inference clusters in a single location. This creates a “data multiplier” that increases the scale-across/DCI requirement beyond what would have been implied by compute growth alone, with cited industry viewpoints ranging from 3-6x higher DCI requirements for inference (Ciena) to approximately 10x (Meta) and approximately 14x incremental bandwidth requirements when stretching AI clusters across sites/regions (Cisco), depending on workload and topology assumptions.
BASE CASE TAM: WHERE THE $65bn+ AND $90bn COME FROM
The report’s 2028 base case is anchored by “traditional” optical profit pools, especially optical transceivers in scale-out networks. The optical transceiver market is sized at approximately $50.3bn in 2028 and is described as the single largest sub-market within the broader optical stack. Additional “traditional” categories include scale-across/DCI ZR pluggables (approximately $2.8bn in 2028), long-haul optics (approximately $4.0bn), and managed optical fiber network (MOFN) services/equipment (approximately $2.3bn), alongside fiber inside the data center and systems. Against this baseline, the report layers in approximately $23bn of “new” optical opportunities by 2028 and beyond, segmented into 5 markets with approximate 2028 TAMs as follows: copper-to-fiber in scale-up (approximately $6.5bn), co-packaged optics (CPO) in scale-up (approximately $2.9bn), optical circuit switching (OCS) in scale-out (approximately $5.7bn), CPO in scale-out (approximately $5.0bn, explicitly described as largely “laser value” and substitutive versus transceivers), and passive optical out-of-band management (approximately $3.0bn).
A key nuance embedded in the report is that “incremental TAM” is not uniformly additive to incumbent revenue pools. In particular, scale-out CPO is modeled as replacing traditional transceiver modules, shifting bill of materials (BOM) and economics onto the switch and optical engines rather than pluggable transceivers. This creates a tension: the industry TAM can expand while specific segments (notably merchant transceiver revenue) face headwinds if CPO penetration accelerates. The report’s base case assumes approximately 15% CPO penetration in scale-out by 2028, informed by third-party estimates, and indicates that higher penetration would increase architectural disruption and revenue reallocation, even if it does not reduce underlying “bits shipped.”
TECHNOLOGY TRANSITIONS: MECHANISMS, ECONOMICS, AND ADOPTION FRICTIONS
SCALE-UP: COPPER-TO-FIBER TRANSITION
Scale-up links are characterized as short-distance connections (often <1-2m) spanning chip-to-chip, board-to-board, backplane, and direct attach cable (DAC) connectivity where copper has historically dominated due to lower upfront cost, higher reliability, and lower power at short reach. The report argues that copper’s headroom is now primarily constrained by signal integrity at higher speeds, reducing maximum reach and forcing more equalization and active electronics. It cites a current reach of approximately 5m with AEC at today’s speeds and suggests that effective reach could be reduced materially at next-generation bandwidth (described as potentially halving at “Rubin” generation), implying that a move to optics becomes less optional beyond that point.
The adoption barrier is not only capex cost but also operational risk and power. The report cites Microsoft research suggesting that power consumption and failure rates could be up to 100x worse with optics versus copper in certain contexts, making reliability/power the dominant overhang rather than purchase price alone. As a result, the transition is positioned as likely but gradual, with multiple “bridging” technologies (active optical cables, linear optics, novel emitters such as MicroLEDs, hollow-core fiber) potentially extending copper’s life or reducing optical penalties. Market sizing is described as challenging due to limited public data, but the report triangulates toward a copper data-transmission content of approximately $1.5bn today growing toward approximately $5bn by 2030 based on copper demand share and growth assumptions, and notes fiber cabling is at least 2x the cost of copper (with cited examples as high as 6x). It also highlights the non-linear impact of densification: Corning has publicly discussed substantially higher fiber density requirements as GPU counts per node/rack increase, implying that unit volumes (fiber count per switch/rack) can rise faster than simple bandwidth scaling.
From an investment standpoint, this transition is structurally favorable to fiber and connectivity vendors, but its monetization path is unusually sensitive to reliability engineering and operational acceptance. The “100x worse” failure/power framing implies that a pure “copper hits a wall therefore fiber replaces it” narrative understates the probability of extended transition periods with hybrid architectures. That, in turn, raises the likelihood that near-term winners are those supplying incremental “bridge” solutions (AEC/AOC, improved connectors, higher-density cabling) while longer-duration upside accrues to fiber vendors if and when a decisive architectural pivot occurs.
SCALE-UP: CO-PACKAGED OPTICS AND OPTICAL I/O
CPO in scale-up is presented as both an enabler and an accelerant of the copper-to-fiber transition, ultimately addressing the I/O bottleneck as packaging advances toward 2.5D/3D integration and optical I/O. The core value proposition is to reduce electrical trace lengths by integrating optical engines more directly into the package or onto the board near the switch/processing silicon, lowering power and latency while supporting higher bandwidth density.
Adoption friction is high and is explicitly framed as multi-dimensional:
1.Reliability and serviceability: placing optics closer to the ASIC increases “cost of failure,” reduces hot-swappability, and can increase operational downtime if failures require power-down and replacement of larger subsystems rather than modular pluggables.
2.Manufacturing cost and yield: CPO is described as carrying approximately 8-10x higher ASPs than traditional pluggables, with low yields driven by packaging complexity, thermal constraints, signal integrity challenges, and compatibility across heterogeneous components.
3.Ecosystem and standardization: CPO is characterized as less standardized and more concentrated, particularly around Nvidia and Broadcom timelines and design choices, implying greater vendor concentration and greater risk that profits shift from optical component vendors toward the dominant switch/accelerator platforms.
4.Supply-chain novelty: success requires an integrated ecosystem spanning optical components, DSP, ASICs, and advanced packaging/test, which is described as unprecedented at scale relative to mature pluggables ecosystems.
The report also points to heightened strategic activity in optical I/O and CPO-related ecosystems, including Marvell’s announced acquisition of Celestial AI to accelerate scale-up connectivity for next-generation data centers and optical I/O.  It also references Ciena’s acquisition of Nubis Communications for $270m to expand inside-the-data-center strategy, consistent with systems vendors moving upstream into CPO/NPO modules and advanced connectivity.  These transactions validate that platform and systems vendors are increasingly treating advanced photonics as a core capability rather than a procurement line item, a dynamic that can structurally compress the long-term bargaining power of merchant component suppliers if integration deepens.
SCALE-OUT: OPTICAL CIRCUIT SWITCHING
OCS is framed as a method to reduce latency and power by keeping traffic in the optical domain and avoiding repeated optical-electrical-optical (OEO) conversions. Operationally, it resembles a circuit switchboard: paths are configured and maintained until reconfigured, which can be optimal for long-lived, predictable traffic patterns but less suitable for highly dynamic flows. The report emphasizes that OCS is more likely to be deployed as an overlay to improve reliability, pooling, and reconfiguration in AI clusters rather than as a wholesale replacement of packet switching spines, due to the need for deep understanding of traffic patterns and the challenges in dynamic routing environments.
The report’s OCS TAM build is aggressive relative to historical deployments, projecting revenue ramping from approximately $198m in 2024 to approximately $5.7bn in 2028, based on backend Ethernet switch ports scaling and OCS penetration rising from 15% to 25% of ports by 2028, with average selling price per OCS box declining approximately 10% per year. The implied adoption is tied to TPU/GPU ports utilizing OCS rising sharply as AI cluster sizes increase. This ramp is directionally consistent with the industry’s exploration of optical fabrics for AI, including Google’s published work describing “lightwave fabrics” and at-scale optical circuit switching for datacenter and machine learning systems.  Additional academic and industry commentary indicates experience with large-scale deployments of such fabrics for long-lived traffic patterns in tightly-coupled ML clusters, supporting the plausibility of OCS as a meaningful architectural element in certain workloads. 
However, the economic durability of the OCS opportunity is sensitive to 2 structural factors. First, OCS competes with continued advances in packet switching silicon and congestion control in addressing latency and efficiency, potentially limiting penetration if packet switching improves faster than expected. Second, OCS’s greatest value is likely at the intersection of predictable traffic matrices and high cost-of-latency workloads; if workload heterogeneity grows (mixed training/inference, multi-tenant AI clouds) and traffic becomes more dynamic, adoption could remain concentrated among a small number of hyperscalers capable of deep software/hardware co-design. The report itself notes Google’s long period of limited deployment, implying that even when technical feasibility exists, operationalization and standardization can be multi-year bottlenecks.
SCALE-OUT: CO-PACKAGED OPTICS ON-BOARD
Scale-out CPO is motivated by limitations of traditional pluggable transceivers: cost (noted as approximately 10% of data center cost), power consumption, failure contribution, lane scaling constraints (traditional transceivers limited to 16 lanes), theoretical ceilings on per-laser speeds (noted as >400G not available in certain assumptions, implying a 6.4T ceiling), and front-panel/backplate density constraints. Moving optics onto the board reduces electrical path lengths and can reduce latency/power by eliminating some OEO conversion penalties, while enabling higher density.
The report frames scale-out CPO as economically disruptive. While total optical content can remain high or increase, the packaging of value changes. Approximately half of the transceiver BOM is described as moving onto the switch in a CPO architecture, and the report’s base case assumes approximately 15% penetration, modeling this as approximately $5bn of revenue shifting from transceiver modules to on-board optical engines/lasers by 2028. It also cites analysis suggesting CPO architectures can eliminate approximately 7% of total data center cost by reducing cabling and transceiver needs, implying that hyperscalers have a direct incentive to push adoption if operational risks can be mitigated.
The adoption friction closely mirrors scale-up CPO: higher failure impact, lack of hot-swap, challenging troubleshooting that can require downtime, low yields, and immature standards. The net implication for merchant component suppliers is ambiguous and depends on positioning in the new BOM. Suppliers with differentiated lasers/photonic integration may retain substantial content, while suppliers primarily leveraged to transceiver module assembly could face structural headwinds if integration shifts toward switch vendors and tightly controlled ecosystems.
OUT-OF-BAND MANAGEMENT VIA PASSIVE OPTICAL NETWORKS
Out-of-band management (OOBM) is typically implemented via low-speed switches and copper cabling separate from the production network to enable monitoring, troubleshooting, and device recovery. The report frames OOBM as a non-trivial cost line (estimated 2-5% of the Ethernet switching market, and likely 2-3% for AI racks given rack cost inflation). Applying 2-3% to a stated approximately $150bn cloud networking market (switching plus transceivers) yields a $3-4bn opportunity, consistent with the report’s $3bn 2028 TAM estimate.
Ciena’s DCOM is highlighted as a PON-based redesign co-developed with Meta, reducing active network elements and replacing Ethernet aggregation with passive optical technology. Claimed benefits include approximately 99% rack space savings and approximately 30% power reduction, directly aligned with the report’s broader “power as constraint” thesis.  Adoption risk is primarily customer concentration and co-design friction: a 2-year co-design cycle with Meta is cited, while future hyperscaler integrations could be shorter (approximately 6 months) but remain non-trivial. Competitive risk is also meaningful because PON expertise exists across multiple vendors (including traditional telco PON competitors), reducing the probability of durable monopoly economics if the opportunity scales.
SCALE-ACROSS: DCI, ZR PLUGGABLES, SYSTEMS, AND MOFN
Scale-across demand is positioned as a second-order effect of power constraints: inability to scale in a single site increases inter-site cluster distribution, raising the fraction of traffic that exits the data center and creating multiplicative demand for DCI bandwidth. Within DCI, the report emphasizes an architectural shift from chassis-based systems to pluggables for shorter distances due to cost and power advantages, with pluggables described as approximately 20% of the cost of systems and ASPs around $2-3k versus $25-40k for systems. ZR optics initially addressed shorter reach (<200km) but extension to ZR+ and higher speeds (including 800G) expands practical applicability, while the report suggests pluggables remain structurally best suited to metro (<1000km) and systems dominate long haul (>1000km).
The supply/demand and competitive dynamic differs meaningfully from transceivers. Pluggables are described as more competitive with more vendors, implying earlier price pressure potential. Systems and longer reach coherent solutions favor fewer incumbents (notably Ciena, Nokia, Cisco). The report also highlights “multi-rail” architectures (notably Ciena’s) that densify optical capacity per rack, citing up to 32x density improvement (128 rails per rack versus 4 rails per rack), which is a mechanism to mitigate space and power constraints by concentrating bandwidth and reducing the footprint of optical line systems.
MOFN is framed as a structurally incremental opportunity where service providers manage or lease optical networks for hyperscalers that cannot or prefer not to own fiber for regulatory or scaling reasons. The report cites third-party estimates implying a non-direct DCI/disaggregated portion of approximately $2bn by 2030, growing at approximately 13% CAGR from 2025-2030, suggesting that value shifts toward managed services and disaggregated architectures can incrementally expand spend even if direct hyperscaler-owned DCI growth slows.
PRICING, SUPPLY CONSTRAINTS, AND MARGIN SENSITIVITY
A central investment tension is the divergence between “bit growth” and “revenue growth,” largely mediated by pricing and mix. Historically, optical component markets experienced 10-15% annual price declines, compressing revenue even as bandwidth shipped increased. The report highlights a recent reversal: 2 years of limited price declines or outright increases, attributed to tight supply in key components (especially electro-absorption modulated lasers, EMLs) and rapid adoption of higher-speed lanes (200G lanes cited as accelerating). This dynamic has outsized earnings implications because higher pricing flows through gross margin more than proportional to revenue.
The EML tightness is identified as a major unknown because the market has traditionally been deflationary. The report points to a clear profitability impact, including Lumentum’s gross margin expansion (800bps+ Y/Y cited) and the broader risk that if supply/demand returns to equilibrium, pricing could revert toward historical deflation. The report also emphasizes laser technology substitution risk. Silicon photonics (SiPh) is presented as potentially delivering approximately 15% power savings and 10-20% cost savings versus conventional approaches, but with yield as the gating factor: InP processes can achieve >90% yields, while SiPh is described as approximately 60% yield today and needing 80-90% to be pricing-competitive. A penetration pathway is suggested (20-30% at 800G with potential 30-40% at 1.6T), implying that incremental SiPh maturity could materially reduce EML demand share at the next major speed node, creating a negative convexity for suppliers most dependent on EML scarcity economics.
The implication for market sizing is material. If transceiver ASPs resume 10-15% annual declines, a “$50bn transceiver market by 2028” outcome requires substantially higher unit volumes and/or mix shifts to offset deflation; conversely, if pricing remains firm, revenue can grow faster than bit demand. The report’s base case implicitly leans toward pricing resilience through 2028, and equity valuations (discussed below) appear to discount a continuation of the current tightness regime.
EQUITY VALUATION, EXPECTATIONS, AND WHAT APPEARS TO BE PRICED IN
The report explicitly notes that the optical complex has re-rated aggressively, citing approximately 300%+ average stock appreciation over the last 12 months across GLW, LITE, CIEN, and COHR (approximately 200% excluding LITE) and approximately $180bn+ of incremental market capitalization added over the last year. Valuations are characterized as “hovering between 30-40x FY28 estimates,” which, in practical terms, implies investors are underwriting a multi-year period of strong end-demand, sustained pricing, and margin durability with limited cyclicality.
Morgan Stanley’s FY28 framework for the 4 highlighted names illustrates how much fundamental improvement is already embedded:
1.LITE: FY28 revenue approximately $5.7bn, operating margin approximately 36%, EPS approximately $16.23, current multiple approximately 39x on FY28 estimates. The report emphasizes that the market is effectively pricing in continued margin expansion and sustained pricing power, with a referenced assumption set implying a move from approximately 10% operating margins in FY25 to 35%+ by FY28.
2.CIEN: FY28 revenue approximately $7.7bn, operating margin approximately 21%, EPS approximately $8.12, current multiple approximately 39x. The value proposition is positioned around AI-driven DCI/scale-across growth and potential incremental inside-the-data-center opportunities (including DCOM), but with limited expected room for multiple expansion at current levels.
3.COHR: FY28 revenue approximately $9.9bn, operating margin approximately 24%, EPS approximately $9.00, current multiple approximately 26x. Relative valuation is lower versus LITE/CIEN in the report’s framework, reflecting a more diversified end-market exposure and volatility concerns, but with AI/datacom upside and leverage from self-help (margin improvement, debt reduction).
4.GLW: FY28 revenue approximately $22.7bn, operating margin approximately 22%, EPS approximately $4.22, current multiple approximately 31x. The report argues GLW offers relatively better longer-duration exposure to the copper-to-fiber transition that is not fully embedded in estimates and would extend the growth runway beyond 2028.
The report’s investment conclusion is not that the optical theme is wrong, but that equities have already priced “perfection,” making performance increasingly dependent on continued upward revisions to AI capex data points and sustained component tightness. It highlights 2 dominant macro sensitivities:
1.Demand sensitivity: optical equities are expected to “work” as long as capex is revised higher, but would be vulnerable to any perceived pause in AI infrastructure spending, particularly given the high multiples applied to FY28.
2.Pricing sensitivity: a reversion to historical deflation (10-15% annual price downs) or a technology mix shift away from scarce components (EML) would likely compress margins and earnings expectations, with the greatest risk to component-heavy names.
The report’s relative caution on LITE is rooted in this second sensitivity: the valuation is described as requiring a multi-year persistence of elevated margins and favorable pricing to justify the implied earnings power. In contrast, GLW is framed as less exposed to pricing deflation in active optics and more levered to a longer-cycle physical transition (more fiber content) that could persist even if component pricing normalizes.
