Morgan Stanley ECOC Minutes: NPO Consensus Shifts to 2028 Deployment, Optical Component Supply Chain Sold Out for 12–18 Months
nashnova research
Morgan Stanley's on-the-ground ECOC notes deliver three calls: the industry consensus on near-package optics (NPO) has locked onto first products in 2028, the optical component supply chain is sold out for 12–18 months, and test leader Keysight is the most route-agnostic beneficiary.
NPO went from "if" to "when" — what changed?
Six months ago at OFC, the industry was still debating whether NPO was necessary. At this ECOC, consensus has shifted to first products entering trials by 2028.
This means → the catalyst is Open CPX MSA — an industry body setting unified specs for optical engines (modules that move light-signal transceivers closer to the chip). Hyperscalers no longer fear vendor lock-in to Nvidia's closed approach.
Morgan Stanley cautions: the first-generation product is unlikely to be the final architecture — laser type, laser placement may all change in later generations. "Today's competitive landscape does not lock in long-term winners."
In plain terms = the debate is no longer "should optics move closer to the chip?" but "whose design ships first?" — and first to ship may not be the endgame winner.
Supply chain sold out for 12–18 months — where is the bottleneck?
Morgan Stanley says most optical component and system suppliers are sold out for the next 12–18 months. The tightest chokepoint is indium phosphide (InP) substrates — the base material for laser chips — with AXT and Sumitomo Electric named specifically.
This reflects a shortage traceable to February 2025: China imposed export controls on indium phosphide, and China produces roughly 70% of the world's refined indium — InP's upstream feedstock. Six-inch InP wafer prices have risen about 250% since controls began, to roughly $5,000 per wafer.
Lumentum CEO Michael Hurlston has said the InP supply-demand gap exceeds 30%, and the company's key AI optical component capacity is "effectively sold out for the foreseeable future" — a shortage he called "potentially worse than memory." Nvidia has taken stakes in both Lumentum and Coherent to secure supply.
In plain terms = this is not one part running short. The base material for making lasers is constrained — nearly every link from raw material to finished device is at full capacity. Whoever secures more substrates has upside.
Cisco and Nokia say they still have supply — what does that mean?
Cisco and Nokia indicated at ECOC that they still have some delivery capacity, giving them room for tactical moves around near-term supply allocation.
Morgan Stanley does not expect meaningful share shifts, however — Ciena still holds the majority of line-system share.
This means → having supply does not equal gaining share. Existing customer relationships and system-integration barriers make it hard to convert a short-term supply advantage into lasting market-structure change.
Optical circuit switching (OCS) — Google in a league of its own, followers much smaller?
Google's use of OCS — a technology that switches data-center traffic with light beams instead of electrical signals — is well documented. Production ramp will shift to Lumentum over the coming years.
Driven by Google's success, nearly all other hyperscalers are trialing OCS. Morgan Stanley tempers expectations: followers' use cases are likely far smaller than Google's, and one customer even canceled deployment after completing full qualification.
This means → OCS is a margin-accretive opportunity for Lumentum and Coherent, but near-term most of the value accrues to Lumentum as Google's primary supplier. It is not yet an industry-wide growth driver.
Who does Morgan Stanley favor most — and why is Keysight called "route-agnostic"?
Keysight (KEYS.US, Overweight): regardless of which CPO/NPO implementation path wins, optical growth drives test demand — Morgan Stanley calls it a "route-agnostic" beneficiary. Corning (GLW.US) is placed in the same "wins either way" category.
Coherent (COHR.US): its PhotonLink integration platform — offering end-to-end delivery for hyperscalers that want to assemble open solutions from six-plus vendors but lack in-house integration capability — is seen as particularly well-matched to current demand. The company has signed a long-term agreement (LTA) with one customer for an NPO solution, ramping next year.
Lumentum (LITE.US): did not meet with Morgan Stanley at the conference, but previously indicated it expects to announce an NPO-related LTA in coming months. Price target $945.67.
Ciena (CIEN.US): has multiple paths to success, but meaningfully outperforming its established targets requires InP supply chain loosening further over the next few years.
The one-line summary — supply chain timing is the key catalyst?
Morgan Stanley maintains an In-Line (market-perform) view on the telecom and networking equipment sector. This note did not change any ratings.
In plain terms = the logic chain is clear: NPO consensus formed → demand certainty rose → but supply chain is stretched tight → whoever secures capacity first realizes the thesis first. When supply loosens is the single most important catalyst for all the stock-level logic above.
This means → for investors, the near-term signal is supply-side marginal change (InP substrate capacity, yield improvement); the medium-term signal is whether NPO's first products hit the 2028 timeline on schedule. These two windows set the pace at which the stock theses above can be realized.
市场有风险，内容仅供研究参考，不构成投资建议。