Skip to content
primarysourced Photonics sector Lumentum
LITE
~4 min read · 906 words ·updated 2026-04-29 · confidence 12%

Datacenter optical-transceiver TAM

The datacenter optical-transceiver market sits inside a stack of nested TAMs. From narrowest to broadest:

  1. InP source-laser layer (EML, DML, narrow-linewidth lasers) — the component layer where Lumentum competes directly.
  2. Pluggable transceiver module layer (400G, 800G, 1.6T DR/FR/LR/ZR) — finished modules; this is the layer where Cloud Light’s heritage business operates.
  3. Datacenter optical interconnect layer (transceivers + ROADMs + active optical cables + co-packaged optics modules) — the broadest aggregation reported by industry research firms.

Lumentum captures revenue in layers 1 and 2; Layer 3 is the “headline” TAM most often cited.

Headline TAM forecasts (research-firm aggregator estimates)

⚠ All TAM forecasts are aggregator estimates from LightCounting, Cignal AI, and Dell’Oro Group. They differ in scope (whether they include AOC, telecom, CPO), in geographic coverage, and in their underlying assumptions about ASP trajectories and unit-shipment cadence. Cite + flag, do not treat as ground truth.

LightCounting — datacenter optical transceiver

YearTAMNotes
2024~$10.5BAI-cluster transceivers ~$5B subset
2025~$16.5B60% YoY (per LightCounting July 2025 newsletter)
2026~$26B60% YoY again; AI-cluster subset >$10B
2027Growth moderatesPost-AI-rush digestion period
2028Returns to double-digitLightCounting’s expectation per January 2025 forecast

Source: LightCounting July 2025 newsletter and LightCounting January 2025 newsletter. ⚠

Industry consensus (cross-firm)

PeriodAggregator-consensus TAMAllocation flag
2026$20–25B (datacenter only) / $26B+ (incl. some adjacent)⚠
2028 (AI baseline)$40–60B⚠
2028 (bear / AI digestion)$30–40B⚠

The wide range at 2028 reflects three swing factors:

  • ASP trajectory: 1.6T module ASP ramp vs. 800G commoditization
  • CPO substitution mix: how fast does CPO displace pluggable volume in the AI-cluster spine layer
  • AI capex slope: linear vs. accelerating vs. plateau scenarios

InP source-laser sub-component layer share allocation

This is the question that bridges TAM to Lumentum’s actual revenue. Approximate dollar-share decomposition of an 800G DR8 / FR4 transceiver (illustrative — ⚠ inferred):

Sub-component layerShare of finished module BOM (approx)
InP source laser (EML/DML chips)~10–15%
DSP/PHY silicon (Marvell, Broadcom, MaxLinear)~20–25%
TIA + driver~5–10%
Optical sub-assembly + packaging + WDM filters~20–30%
Test, qual, yield loss~10%
Module assembly marginresidual

⚠ These BOM allocations are inferred from industry-trade-press teardowns and supplier conference calls. Actual proportions vary materially by module form factor (1.6T modules have higher EML share; CPO modules collapse the BOM stack).

If the 2026 datacenter optical transceiver TAM is ~$26B and the InP source-laser layer captures ~12% of finished-module BOM on average, the InP source-laser merchant TAM is ~$3.0B in 2026. Of that, Lumentum and Coherent together hold ≥80% merchant supply share — which puts Lumentum’s addressable component TAM at ~$1.2–1.5B. Lumentum’s actual EML chip revenue (a subset of total Cloud & Networking revenue) was framed by management as on track to “more than double from end of CY2024 through end of CY2025” (Q3 FY2025 earnings call commentary).

Adding finished-module revenue (Cloud Light heritage) layered on top of the chip-layer revenue gets to Lumentum’s broader Cloud & Networking trajectory: the FY2025 segment revenue of $1,410.8M, the H1 FY2026 trajectory pointing to ≥$2.0B annualized, and management’s $1.25B / $2.0B quarterly run-rate milestones.

CPO transition impact on TAM mix

The 2028+ co-packaged optics transition is a TAM mix shift more than a TAM destruction event:

  • CPO modules consume more InP laser dies per bandwidth-unit than pluggables (more lanes per module; tighter optical-power budgets).
  • CPO modules consume fewer packaged-transceiver units per port (one CPO engine replaces multiple pluggables).
  • Net effect on InP source-laser TAM: likely neutral-to-positive in dollar terms; ASP/lane may compress, but unit volume lifts.
  • Net effect on transceiver-assembly TAM (Cloud Light layer): negative if CPO disrupts the pluggable-module business.

This is a structural reason why Lumentum’s positioning at the source-laser layer is more durable than its positioning in the pluggable-module assembly layer. CPO timing risk applies to the Cloud Light segment more than to the EML chip business.

Geographic and regulatory overlay

US export controls on advanced semiconductors and datacom components to China constrain the addressable share of TAM that flows through Lumentum/Coherent for China-destination demand. Innolight, Eoptolink, and Hisense pluggable-transceiver revenue serves Chinese hyperscalers and is partially supplied by domestic-China InP fabs (HG Genuine, Accelink, Sumitomo via licensing routes). The “merchant TAM excluding China” is the more relevant denominator for Lumentum’s competitive analysis. See regulatory landscape for detail.

Sources