Optical interconnect ecosystem for AI networks: ECOC points to a 2028 NPO trial cycle amid tight optical-component supply

Overview

This ECOC meeting summary examines how AI-driven optical-network demand, supply constraints and evolving NPO/CPO architectures could shape the optical-equipment and component ecosystem. Morgan Stanley sees durable demand and a more credible path to initial NPO trials in 2028, while emphasizing that technology formats, supply availability and company positioning remain fluid.

Core views

Morgan Stanley reports that the principal ECOC themes were consistent with prior OFC discussions: optical content in AI networks is expanding faster than suppliers can produce it, many vendors are sold out, and the ultimate form of CPO/NPO remains unsettled. The notable change versus OFC is greater confidence that NPO will emerge in parts of scale-up networks from 2028. The report attributes this shift partly to the Open CPX MSA, which aims to establish common optical-engine parameters and support a broader ecosystem. An open ecosystem could reduce hyperscalers' concern about vendor lock-in compared with reliance on a single closed solution. The report stresses that an initial 2028 NPO product should not be mistaken for a final technical architecture. Laser type, laser location and other implementation choices may change in subsequent generations, and scaling, production and deployment challenges remain substantial. The underlying rationale for moving optics closer to electronics is to reduce power, cost and latency while improving bandwidth density, capacity and reliability in high-speed AI networks. Morgan Stanley therefore remains constructive on the opportunity for optical suppliers, but cautions that current competitive positions may not be durable as architectures evolve and customers pursue the most efficient solutions. Supply conditions remain a central near-term support and constraint. Most component and system suppliers were described as sold out for the next 12–18 months, with InP substrates—citing AXTI and Sumitomo as examples—seen as the gating input. Constraints also persist in EMLs, CW lasers, pump lasers and gold boxes. Additional substrate availability or improved yields could create upside, while Cisco and Nokia reportedly retain some availability that may allow strategic supply releases. Morgan Stanley does not expect this to drive meaningful line-system share shifts, with Ciena retaining the majority of line-system share. On NPO/CPO adoption, the report says that hyperscalers are broadly working toward NPO despite one view that the technology is inevitable but not imminent. Coherent's PhotonLink was highlighted as a potential response to customers' difficulty assembling an open solution from more than six vendors: it offers an end-to-end route while allowing customers to choose solution components. Coherent said it has a long-term agreement with a customer for an NPO solution expected to begin ramping next year. Morgan Stanley also notes that Lumentum has discussed the prospect of an NPO-related long-term-agreement announcement in coming months. Because first-generation NPO/CPO designs are expected to use external lasers, the report expects high-power laser demand to rise sharply in coming years and sees Lumentum and Coherent likely to remain capacity constrained. Optical circuit switching is expected to expand beyond Google, whose production ramp is expected to move to Lumentum in coming years. Morgan Stanley says nearly every other hyperscaler or neocloud is trialing OCS, but expects subsequent deployments to be smaller than Google's. One customer reportedly cancelled an implementation after fully qualifying a solution, illustrating that broader adoption is not assured. The report still views OCS as margin accretive for Lumentum and Coherent, with Lumentum expected to capture most near-term value because it supplies the largest implementation. Morgan Stanley prefers Keysight because its test and measurement exposure can benefit from optical growth regardless of the eventual implementation path; it places Corning in a similar architecture-agnostic beneficiary category. The report sees Lumentum and Coherent as beneficiaries of scarce InP lasers and external-laser demand. It argues that Ciena has multiple paths to success but needs the InP supply chain to loosen to materially outperform its stated target, while Cisco's demand conditions remain strong and its greater supply-chain control supports cloud-customer revenue opportunities. The report found no new indication of Cisco weakness in its meetings.

Analysis framework

Morgan Stanley synthesizes takeaways from meetings hosted at ECOC in Spain on September 21–22, comparing the discussion with prior OFC feedback. It assesses demand, component availability, technology-roadmap timing, customer adoption patterns and the likely implications for suppliers across alternative optical architectures.

Methodology notes

• Industry AnalysisSupply-demand framework

  Optical-component supply-demand assessment

  The report links strong AI-network demand with suppliers being sold out for 12–18 months and identifies InP substrates as the capacity bottleneck affecting several component categories.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI-network architecture and optical supply-chain transmission

  The report traces how hyperscaler requirements for lower power, cost and latency could change NPO/CPO designs and flow through to external lasers, substrates, optical vendors and testing suppliers.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Keysight Technologies (KEYS)

  Preferred beneficiary of optical growth across alternative implementation paths.

  Strengths

  Less dependent on the eventual optical architecture because it benefits as solutions are trialed.

  Comparison

  Morgan Stanley prefers KEYS and places Corning in a similar architecture-agnostic beneficiary category.

  Risks

  The report notes that the form of optical solutions may evolve over time.

• Corning (GLW)

  Architecture-agnostic optical-growth beneficiary.

  Strengths

  Expected to benefit regardless of the eventual implementation path.

  Comparison

  Grouped with Keysight as a beneficiary across alternative implementations.

• Lumentum Holdings (LITE)

  Beneficiary of external-laser demand and the largest reported OCS implementation.

  Strengths

  Expected to capture most near-term OCS value and remain capacity constrained as InP lasers stay scarce.

  Weaknesses

  Its opportunity remains exposed to architecture choices and OCS deployment scale.

  Comparison

  Morgan Stanley expects LITE and COHR to receive substantial attention, while LITE is expected to capture more near-term OCS value.

  Risks

  Later OCS customer deployments may be smaller than Google's, and one qualified deployment was reportedly cancelled.

• Coherent (COHR)

  Potential beneficiary of NPO/CPO adoption and external-laser demand.

  Strengths

  PhotonLink and a broad portfolio may suit customers seeking an end-to-end yet open NPO/CPO solution; the company noted an NPO long-term agreement expected to ramp next year.

  Comparison

  Along with Lumentum, Coherent is positioned for high-power external-laser demand in first-generation NPO/CPO products.

  Risks

  NPO architecture, deployment timing and customer implementation remain uncertain.

• Ciena (CIEN)

  Line-system leader with multiple potential paths to benefit from optical demand.

  Strengths

  Morgan Stanley says Ciena retains the majority of line-system share.

  Weaknesses

  Material outperformance of its stated target depends on more InP supply-chain loosening.

  Comparison

  Cisco and Nokia may release available supply strategically, but Morgan Stanley does not expect meaningful share shifts.

  Risks

  Persistent InP constraints could limit the ability to materially outperform the target described by the company.

• Cisco Systems (CSCO)

  Networking supplier with some supply availability and cloud-revenue opportunity.

  Strengths

  More supply-chain control gives continued opportunities to grow cloud-customer revenue; meetings did not indicate a reason for current weakness.

  Comparison

  Cisco and Nokia reportedly have some availability, while Ciena retains majority line-system share.

• Nokia (NOK)

  Supplier with reported near-term supply availability.

  Strengths

  Available supply may allow strategic near-term releases.

  Comparison

  Morgan Stanley does not expect Cisco or Nokia availability to cause major share shifts.

Key data

• NPO timing2028Consensus timing for NPO in parts of scale-up networks and initial product trials.
• Supplier capacity visibility12–18 monthsMost component and system suppliers were reportedly sold out for this period.
• ECOC meeting datesSeptember 21–22, 2026Dates of the meetings summarized in the report.
• Open-solution assembly complexity6+ vendorsThe report notes that customers may lack the in-house expertise to assemble an open CPO/NPO solution from this many vendors.

Impact & implications

The report views the expansion of optical content in AI networks as broadly supportive for the sector, but expects benefits to vary with the technical path chosen and component availability. It favors architecture-agnostic exposure through Keysight and identifies external-laser demand as supportive for Lumentum and Coherent, while noting that supply-chain constraints and evolving designs could redistribute longer-term competitive advantages.

Risks

• NPO/CPO scaling, production and implementation challenges remain substantial, and the first-generation architecture may not be the lasting format.
• Current competitive positioning may not persist as customers optimize for power and cost efficiency and technology choices change.
• InP substrate shortages remain the key bottleneck across several optical-component categories.
• OCS adoption beyond Google may occur at smaller scale, and one customer reportedly cancelled an implementation after qualification.

What to watch

• Evidence that NPO trials begin in 2028 and how the Open CPX MSA develops the open optical-engine ecosystem.
• Coherent's expected NPO long-term-agreement ramp next year and any NPO-related long-term-agreement announcement from Lumentum.
• InP substrate availability and yield improvements, which could ease constraints across lasers and other optical components.
• The scale and conversion of OCS trials among hyperscalers and neocloud customers beyond Google.