The likelihood of legacy media discussing a murderer’s race depends on the murderer’s race. This is why BLM martyrs like George Floyd get endless coverage while cases like Iryna Zarutska are forgotten. The bias is clear:

$LITE The Mizuho initiation on Lumentum on 11/18/25 frames the company as a high‑leverage beneficiary of the AI data center optical transition, with the core claim that LITE will monetize 4‑8x bandwidth growth inside AI data centers through leadership in InP EML lasers, early positioning in co‑packaged optics (CPO) lasers for NVIDIA, and MEMS‑based optical circuit switches (OCS) for hyperscalers. The report projects revenue rising from $1.6B in FY25 to $4.2B in FY28, a 37% CAGR, with EPS increasing from $2.06 to $10.86, a 74% CAGR, and gross/operating margins expanding from 34.7%/9.7% to 42.4%/25.4%.  At the time of publication the target price of $290 implied material upside from a $242 share price. That upside has largely been realized: Lumentum now trades near $290 with a market cap above $20B, roughly 50x FY26 EPS, 33x FY27 EPS, and 26x FY28 EPS on Mizuho numbers, and modestly higher multiples on consensus. The investment question becomes whether the out‑year growth and margin structure embedded in this multiple are achievable and whether incremental upside versus both consensus and peers compensates for execution, technology, and customer‑concentration risk. The industry backdrop described in the report is directionally correct and strongly supportive of high‑end optics. AI cluster scaling is pushing network port speeds from 200G/400G to 800G and, over the coming 3‑5 years, to 1.6T and eventually 3.2T, with networking bandwidth roughly doubling every 2 years. The charts on pages 8–10 show total data‑center port counts growing at only high‑single‑digit CAGRs, while ports at 400G and above grow much faster, with 800G AI port counts modeled at a 119% CAGR and 1.6T starting to contribute late in the decade.  Copper links are increasingly constrained by reach and power consumption at these speeds. Optical links—particularly those using InP wideband lasers—offer better power‑per‑bit and reach, and therefore are expected to capture most of the incremental bandwidth. External research on AI networking from Gartner, LightCounting, and NVIDIA’s own CPO announcements is consistent with this direction of travel: high‑speed optical TAM inside data centers is expanding rapidly, and 800G/1.6T optics will be required to sustain AI cluster scaling. There is, however, meaningful uncertainty around how value will be split between pluggable optics, linear pluggable optics (LPO), and co‑packaged optics, as well as around the pace and breadth of OCS adoption beyond early users like Google. Lumentum’s current business mix provides a solid base but is already becoming highly AI‑skewed. FY25 revenue of $1.65B is split 68% components and 32% systems, with components dominated by high‑speed lasers and optical chips and systems encompassing transceivers, OCS, and industrial lasers.  Regional exposure is 61% Asia‑Pacific, 29% Americas, and 10% EMEA, reflecting both telecom heritage and outsourced manufacturing. Key end markets include data‑center interconnect, long‑haul/metro telecom, industrial lasers, and consumer VCSELs for 3D sensing. Google and Ciena together account for more than 31% of revenue, with Apple, Nokia, and Amazon as other important customers.  The waterfall chart on page 4 shows that >84% of projected FY25‑28 revenue growth is tied to AI data‑center optics, primarily EML/CW lasers, CPO content, OCS, and cloud transceivers.  This concentration strongly gears Lumentum’s earnings to AI data‑center capex cycles and to a small number of hyperscalers and NVIDIA. The core of the equity story is Lumentum’s InP EML laser franchise. The report estimates LITE has 50‑60% unit share in the high‑speed 100G/200G‑per‑lane EML market, despite only ~6% share of the overall laser‑diode market, with Broadcom and Coherent as the main competitors.  InP’s share of optical lanes is modeled to increase from 61% in 2024 to 81% by 2029 as data rates move to 200G per lane; the company has demonstrated 15‑45% lower power‑per‑bit versus alternatives at 200G lanes.  Lumentum is currently supply‑constrained in 100G EML due to 800G AI deployments and is increasing EML capacity by 40% over approximately 3 quarters. The FY26 guide and Mizuho’s FY26‑27 revenue ramp assume that this incremental capacity is fully absorbed by AI demand, with laser revenue growing from roughly $1B in FY25 to about $2.2B by FY28, a 31% CAGR and 54% of total revenue.  This requires not only that AI port growth remains extremely strong but also that Lumentum maintains or grows its share in the face of competition from Coherent, Furukawa, and a set of emerging InP and SiPho suppliers, including potential second‑source laser providers in NVIDIA’s CPO ecosystem. The external data around Lumentum’s 200G and 400G InP chip introductions, UHP 1310nm DFB lasers, and US capacity expansions supports the view that technology and manufacturing investments are being made to sustain this share. However, Mizuho’s implicit assumption of largely stable share and sustained 50‑100% ASP uplifts per generation is optimistic; history in optical components suggests aggressive pricing and share shifts once alternative suppliers qualify. The second pillar is CPO, where Lumentum has been publicly named as a key laser supplier to NVIDIA’s Spectrum‑X and Quantum‑X Photonics switches. NVIDIA’s March 2025 GTC press release and Lumentum’s own release describe a silicon‑photonics ecosystem in which Lumentum supplies high‑power, high‑efficiency lasers, while Coherent collaborates on the silicon‑photonics CPO implementation and other partners such as TSMC, Corning, and Foxconn participate across the stack. NVIDIA claims its CPO switches deliver 1.6Tbps per port, 3.5x power efficiency improvements, and 10x resilience with 4x fewer lasers than traditional architectures. Mizuho’s model assumes Lumentum is effectively sole‑sourced for external laser modules (ELS) feeding Quantum‑X and Spectrum‑X, with UHP 400mW DFB lasers priced at $50‑100 each, 8 lasers per ELS, and ~16‑18 ELS per 128‑144 port switch, implying roughly $10k‑20k LITE content per high‑end CPO switch.  On that basis, CPO revenue is projected to rise from a few million dollars today to ~$93M in FY26, $356M in FY27, and $517M in FY28, a 416% CAGR.  These projections appear aggressive relative to the available external evidence. NVIDIA has explicitly stated that CPO‑based systems will be offered alongside, not in place of, pluggable‑based systems, and that adoption will depend on specific customer workloads and power envelopes. Furthermore, NVIDIA’s communications emphasize a multi‑partner ecosystem rather than exclusive dependence on Lumentum. In high‑volume networking silicon, dual‑sourcing is standard practice. A more conservative base case would assume strong early CPO adoption but with Lumentum’s share of CPO laser content in the 30‑60% range over time rather than effectively 100%, and with the CPO mix of total 1.6T ports ramping more gradually. Under such assumptions, the FY27‑28 CPO revenue contribution could be materially lower than Mizuho’s $300‑500M trajectory, which would reduce both the absolute EPS and the mix‑driven margin uplift that the initiation embeds. The third major growth vector is OCS, where Lumentum’s R300 and R64 MEMS‑based optical switches target AI data‑center fabrics for both scale‑out and scale‑up use cases. The report models the OCS data‑center TAM rising from roughly $400M today, almost fully driven by Google’s internal Palomar/Apollo deployments, to $1.9B by 2029, a 44% CAGR, with Lumentum capturing 30‑40% share and generating more than $500M of OCS revenue by FY28.  This implies LITE OCS revenue rising from near zero in FY25 to $71M in FY26, $416M in FY27, and $586M in FY28, with ASPs around $80‑90k and gross margins above 50%.  The technical rationale for OCS is sound. Google’s public Jupiter/Apollo work shows that OCS‑based fabrics can deliver around 30% capex reduction, >40% power savings, and lower flow‑completion times through direct optical connections and elimination of spine switch layers. Lumentum’s R300 product literature and third‑party analysis claim >65% network power reduction and significant latency improvements for large GPU clusters, with sampling at multiple hyperscalers and general availability in 2H25. In addition, LITE is co‑leading the Open Compute Project’s OCS subproject, alongside iPronics, Google, Microsoft, and others, to standardize photonic networking and OCS interfaces. These developments support the thesis that OCS can transition from bespoke, single‑customer systems to a broader, standards‑based market. The key investment issue is not whether OCS is technically attractive but how quickly and broadly hyperscalers will deploy third‑party OCS hardware and what share Lumentum will capture. Google’s current deployments are largely in‑house. There is no public evidence yet that AWS, Azure, or even Google have committed to multiyear, volume purchases of Lumentum’s R300/R64. Competitive offerings exist from Coherent (liquid‑crystal OCS), Huber Suhner’s Polatis line, Telescent, Calient, and others. Standardization under OCP is a double‑edged sword: it likely accelerates overall OCS adoption but tends to compress margins and lower barriers to entry over time. The Mizuho forecast effectively assumes that LITE wins 2‑3 large hyperscaler designs and rapidly achieves several hundred million dollars of annual OCS revenue within 3 years at very high margins. A more neutral stance would treat this as an upside scenario rather than a base case; the base should probably assume slower OCS adoption curves, partial design win share, and margin pressure as additional vendors enter. The cloud transceiver business provides incremental revenue scale but is structurally much less attractive than lasers, CPO, or OCS. Lumentum leverages its laser and DSP assets (augmented by the NeoPhotonics, IPG Photonics, and Cloud Light acquisitions) to sell 400G/800G/1.6T pluggable transceivers into hyperscalers, with estimated ASPs of $0.40‑0.50 per Gb ($400‑800 per module), but at only mid‑teens to, over time, perhaps 30% gross margins.  The company is deliberately capping this business at roughly $1B per year to limit corporate margin dilution. The Chinese optical module ecosystem is a major competitive force: Innolight, Eoptolink, TFC, Luxshare, and others are scaling 400G/800G transceiver volumes rapidly, with some players more than doubling revenue and pricing around $0.35 per Gb. This supply base has a structural cost advantage from localized labor and vertically integrated component supply. Lumentum’s US and global manufacturing footprint, plus its closer alignment with US hyperscalers and NVIDIA and its expected role in secure domestic AI supply chains, partly offsets that disadvantage. Nonetheless, margin expansion assumptions in the report—corporate gross margin rising to 42‑43% even as transceivers reach nearly $1B revenue—depend on OCS and CPO reaching scale and contributing >50% gross margins. If OCS/CPO ramp more slowly or at lower margins, the incremental transceiver mix will act as a headwind to the targeted margin structure. Financially, Lumentum is transitioning from a telecom‑centric, mid‑30s gross margin, high‑single‑digit operating margin business to a model that more closely resembles high‑growth communications semiconductors. Q1 FY26 results showed revenue of $533.8M, up 58% year‑on‑year, with non‑GAAP gross margin of 39.4%, operating margin of 18.7%, and EPS of $1.10, at the high end of guidance. Management guided Q2 revenue to about $650M with further operating leverage as EML and transceivers ramp. Mizuho projects FY26 revenue of $2.63B, up 60% year‑on‑year, followed by $3.58B in FY27 and $4.21B in FY28, with gross margins rising from 40.3% to 42.4% and operating margins from 21.1% to 25.4% over that period.  EPS is modeled at $5.78 in FY26 (essentially in line with street), $8.77 in FY27 (about 7% above consensus), and $10.86 in FY28 (about 16% above consensus), implying more than 5x EPS growth in 3 years.  Free cash flow is projected to ramp from roughly $141M in FY25 to $593M in FY28, as capex normalized after heavy investment in laser and module capacity.  These numbers are internally consistent with the underlying product assumptions, but they embed very high execution and demand certainty: in rough terms, about $2.6B of FY28 revenue (over 60% of total) is expected to come from AI‑specific laser, CPO, OCS, and cloud transceiver businesses that barely existed in FY24. Any delay or disappointment in AI capex, CPO mix, or OCS deployments would cascade quickly into revenue and margin outcomes, given the company’s operating leverage. At around $290 per share, Lumentum trades on roughly 50x FY26 EPS, 33x FY27 EPS, and 26x FY28 EPS using the Mizuho forecasts; on consensus EPS, the FY27 and FY28 multiples are closer to 35x and 28x. The peer group in the report (Coherent, Fabrinet, AVGO, NVDA, Sumitomo Electric, Furukawa, Innolight, Eoptolink, TFC, Luxshare) carries a median FY26 P/E around 25x and average EPS growth of about 40% in 2026 and 24% in 2027, versus Lumentum’s modeled 84% and 25%.  On this basis, the premium multiple relative to the optical peer set is justified only if LITE achieves its out‑year laser/OCS/CPO ramp and sustains above‑peer EPS growth into FY28. Versus broader AI infrastructure leaders, the picture is more mixed. NVIDIA and Broadcom also trade at high forward multiples but are diversified across GPUs, accelerators, NICs, switch ASICs, and software, with much broader customer bases and multiple growth vectors. Lumentum is a more narrowly focused component supplier with heavier customer and product concentration and less control over system design and adoption timing. In addition, the share price has already moved sharply: external analysis indicates the stock has risen more than 120‑140% year‑to‑date and roughly 144% since late July, driven by the Q1 beat, revised guidance, and growing AI enthusiasm. The risk‑reward profile at current levels is therefore more symmetric than at initiation; the Mizuho $290 target now approximates spot, while the average Street target around the low‑$200s sits meaningfully below spot. Several key risks could impair the bullish thesis. Technology adoption risk is significant: if AI networking remains dominated by improved pluggables and LPO platforms at 800G and even 1.6T, with CPO penetration remaining modest, then the upside from UHP CPO lasers will be limited and traditional transceiver vendors will continue to capture most of the optics value. Similarly, if hyperscalers find that software‑defined electrical fabrics or incremental Clos optimizations are “good enough,” OCS adoption could remain confined to niche or internal deployments, leaving Lumentum’s OCS revenues well below the projected $400‑600M scale. Competitive risk is elevated. Coherent, Furukawa, and others have strong InP capabilities, while Chinese module manufacturers are expanding up the stack from transceivers into photonics and potentially CPO‑adjacent solutions. Hyperscalers and NVIDIA are structurally incented to multi‑source critical components, particularly lasers, to avoid supply shocks. Customer concentration and AI cycle risk are acute: Google and Ciena alone already represent more than 30% of revenue, and Mizuho’s estimates assume increased exposure to Google, Microsoft, Amazon, Meta, and NVIDIA through OCS and CPO.  Any pause or digestion phase in AI GPU deployments—whether due to macro, regulatory limits, AI workload profitability, or alternative architectures—would disproportionately impact Lumentum’s growth trajectory. Geopolitical and supply‑chain risks are non‑trivial given manufacturing and demand exposure to China and the broader Asia‑Pacific region, as well as potential US export restrictions on high‑speed optics. Finally, the ramp of new products like OCS and CPO requires flawless execution in manufacturing yield, field reliability, and customer integration; early failures or delays could jeopardize design‑in positions and reputational standing. There are also non‑trivial positive optionalities. If NVIDIA’s CPO platforms see broad adoption across hyperscalers and AI cloud providers and if Lumentum maintains a leading share of external laser modules, CPO revenue and associated margins could exceed the already aggressive Mizuho trajectory. If OCS standardization through OCP accelerates and LITE’s R300/R64 family becomes a de facto standard, the company could effectively assume a quasi‑system role in AI network fabrics rather than remaining merely a component supplier, supporting structurally higher margins. Incremental US or allied‑country subsidies for domestic photonics manufacturing could support both capacity expansion and margins, especially for lasers used in strategic AI infrastructure. Finally, Lumentum remains a strategic asset within the AI optics ecosystem; consolidation dynamics could lead to either accretive acquisitions by LITE (to deepen SiPho or module capabilities) or the company itself becoming a target for larger semiconductor or networking players seeking AI photonics exposure. From an investment perspective, the Mizuho report correctly highlights Lumentum as a pure‑play, high‑beta levered exposure to AI data‑center optics with scarce assets in high‑power InP lasers and emerging OCS technology. The report’s revenue and EPS trajectory, however, should be viewed as a bullish scenario rather than a base case, given the level of unproven OCS and CPO ramp embedded. At approximately $290 per share, the market is already capitalizing a large portion of that bullish path, with the stock trading at a premium to optical peers and at a multiple that assumes sustained high‑double‑digit EPS growth into FY28. The investment implications are that incremental upside from here likely requires confirmation of large, multi‑customer OCS and CPO design wins and evidence that these products can scale to hundreds of millions of dollars of revenue at >50% gross margins. In the absence of such confirmation, the downside risk stems from any disappointment in AI networking capex or signs of share loss or multi‑sourcing at key customers, which could compress both earnings expectations and the valuation multiple. In portfolio construction terms, Lumentum fits best as a satellite position within an AI infrastructure basket, offering outsized sensitivity to AI optic and networking themes but also higher idiosyncratic risk than more diversified names like NVIDIA and Broadcom. The current valuation and recent share‑price acceleration argue for measured position sizing and heightened focus on upcoming catalysts: subsequent quarterly prints that detail EML capacity utilization and mix shift to 200G lanes; explicit disclosure of OCS revenue contributions and customer count; progress updates from NVIDIA on CPO shipments and mix; and developments from the OCP OCS project that signal whether Lumentum’s implementations are becoming reference designs or face intense standardized competition. Confirmation along these axes would support the higher‑growth, higher‑multiple path implicit in the Mizuho initiation; any shortfall would likely necessitate a reset of both estimates and valuation toward the peer group, with significant downside from current levels.