Co-packaged optics (CPO) market

Co-packaged optics is the next architectural transition in datacenter optical interconnect. Rather than pluggable transceivers seated in front-panel cages, CPO co-locates optics directly with the switch ASIC inside the package — reducing electrical-channel loss, simplifying SerDes, lowering power-per-bit, and enabling higher-radix scale-up topologies. NVIDIA framed CPO publicly via the Spectrum-X / Quantum-X Photonics announcement (GTC 2025), with 2027 first commercial volume on pluggable-form-factor CPO (scale-out networks) and 2028 scale-up CPO at NVLink scale (NVIDIA press release on Spectrum-X CPO networking switches).

For Coherent specifically, CPO is structurally significant on two axes: (1) source-laser merchant supply (high-power CW DFB lasers feeding silicon-photonics CPO engines — duopoly with Lumentum), and (2) potential vertical-integration into CPO module assembly via Finisar heritage transceiver-assembly capability.

CPO commercial-volume timing

Milestone	Timing	NVIDIA-named partner status
First Spectrum-X CPO switches in NVDA portfolio	CY2026 (limited volume)	Coherent + Lumentum named source-laser suppliers
Spectrum-X CPO scale-out commercial volume	CY2027	Coherent + Lumentum source-laser scale
Quantum-X NVLink CPO scale-up commercial volume	CY2028	Same
Hyperscaler CPO ramps beyond NVDA-direct	CY2028+	Microsoft, Meta, Google CPO procurement begins
CPO mix-share crossover vs pluggables (>50%)	CY2030+	Industry framing per LightCounting ⚠

The 2027 / 2028 timeline is industry-consensus and has held since the GTC 2025 announcement. Coherent management commentary on Q1 and Q2 FY2026 calls reinforces alignment to the published roadmap; no public timeline-slippage commentary as of April 2026.

NVDA Spectrum-X / Quantum-X dependencies

The CPO commercial path is dependent on NVIDIA’s switch roadmap as the primary demand signal. NVIDIA will be the largest CPO consumer through 2028; hyperscaler-direct CPO procurement (e.g., Meta/Microsoft buying CPO modules separately from switches) is a 2028+ phenomenon. The structural implications:

Source-laser supply duopoly is preserved into CPO transition — NVDA’s bilateral $4B equity investment in March 2026 explicitly references “future access and capacity rights for advanced laser and optical networking products” (NVIDIA-Coherent press release). Both Coherent and Lumentum are named CPO source-laser partners.
CPO chip-volume per unit of bandwidth is higher than pluggables — CPO modules consume more InP chips per gigabit because the integration density is higher per package. This is InP-volume positive even as transceiver form factor changes.
The CPO engine is silicon-photonics-based — Coherent’s in-house silicon-photonics process is small-volume; NVDA-favored CPO engine vendors (Broadcom, TSMC SiPh consortium, possibly Coherent’s own SiPh) supply the engine into which Coherent’s InP source lasers feed via flip-chip bonding.

Coherent’s CPO position

Coherent’s CPO-relevant capabilities:

Layer	Coherent capability	Strategic role
InP source lasers (high-power CW DFB)	Production scale; 6-inch InP fab Sherman TX	Duopoly merchant supplier; NVDA-named partner
Silicon photonics	In-house process, small volume; OFC 2026 demonstrations	Could supply CPO engine to NVDA or to own internal CPO transceiver
CPO transceiver/module assembly	Finisar heritage assembly + advanced packaging	Vertical-integration path beyond just chip supply
Heterogeneous integration / advanced packaging	Internal capability; OFC demos of integrated stack	Differentiator vs Lumentum (Cloud Light is module-only)

The vertical-integration breadth is the structural difference from Lumentum: Coherent has the option to supply CPO at multiple value-stack layers (InP chip, SiPh engine, finished CPO module) where Lumentum is primarily a chip supplier into NVDA’s CPO module assembly. Whether Coherent realizes this option depends on (a) NVDA’s degree of openness to merchant CPO modules vs in-house assembly, (b) Coherent’s silicon-photonics capacity scaling on time, and (c) hyperscaler-direct CPO procurement emerging in 2028+.

Alternative-architecture risks

Three architecture transitions could displace CPO penetration in the 2026–2030 window:

Linear pluggable optics (LPO)

LPO removes the DSP from a pluggable transceiver — the lossy electrical channel between the switch ASIC and the front-panel cage is “linear” (no signal-integrity restoration in the module). Power, cost, and latency all improve relative to standard pluggables; CPO’s value proposition is partially relative to non-LPO pluggables. The risk: LPO captures pluggable-pluggable improvement faster and cheaper than the industry can ramp CPO, slowing CPO penetration and elongating the pluggable-form-factor lifecycle.

LPO is a Marvell-DSP-friendly architecture (Marvell’s coherent-DSP business loses revenue but gains other share — see MRVL). Coherent benefits from LPO indirectly as the source-laser layer is unchanged.

Onboard optics (OBO)

OBO places optics on the switch line-card (one step short of inside-the-package). It’s a simpler integration path than full CPO, less aggressive on power/density, but doesn’t require switch-ASIC vendors to redesign packaging. OBO may capture some of the workload that the industry expected CPO to take, especially if CPO timelines slip.

OBO consumes chips at intermediate density between pluggable and CPO — for the merchant InP supply layer, OBO is a wash vs CPO.

Pure NVDA-internal CPO (in-source risk)

The opposite-end risk: NVDA decides to vertically integrate the CPO source-laser layer or the silicon-photonics engine, displacing merchant supply. The March 2026 bilateral $4B equity investment is structurally consistent with NVDA preferring merchant supply at least through 2028 — but the relationship could pivot if NVDA scales CPO at hyperscaler magnitude and finds the merchant-supply bottleneck unsupportable. This is the load-bearing bear-case scenario for CPO.

CPO TAM scenarios

Scenario	CY2027 CPO TAM ($B)	CY2028 CPO TAM ($B)	CY2030 CPO TAM ($B)	Coherent CPO revenue (FY2028)
Bull	$3	$7–10	$20+	$700M–$1B
Base	$1–2	$4–6	$10–15	$350–600M
Bear (timeline slip)	<$1	$1–3	$5–8	$150–300M

⚠ All TAM figures here are aggregator-estimator-house ranges or analyst extrapolations; CPO TAM is the single most uncertain TAM in the 2025–2030 photonics outlook. The directional shape (CPO grows from <$1B in 2026 to $5–10B by 2028 to $20B+ by 2030) is industry-consensus.

Confirmation paths and counter-signals

Confirmation paths (CPO bull case validated):

Spectrum-X CPO switches ship at NVDA-announced milestones in CY2026
Coherent and Lumentum revenue commentary references CPO source-laser shipments by FY2027
NVDA-Coherent follow-on agreements specifically reference CPO scope
CPO module assembly capacity additions at Coherent (Sherman TX or other site)
Hyperscaler CPO procurement RFPs visible by 2028

Counter-signals (bear case validated):

NVDA pushes Spectrum-X / NVLink CPO timeline by 12+ months — major thesis-breaking
LPO captures more 1.6T / 3.2T share than expected, displacing CPO transition timing
Marvell or Broadcom CPO-engine vendor wins outsized share, excluding Coherent silicon photonics
NVDA in-sources CPO source lasers (would be visible as NVDA fab announcements)

Cross-link

AI capex cycle — top-down demand
Datacenter optics TAM — TAM allocation framework
InP EML duopoly — source-laser supply share
02_technology CPO roadmap — Coherent’s process-level CPO position
02_technology silicon photonics — in-house SiPh capability
05_financials capex cycle — CPO capacity-build dollars
LITE — CPO market — duopoly-partner view
MRVL — DSP/LPO and CPO-engine view