AI optical communication advanced packaging: glass bridges are used for mode field matching, and NPO commercialization may come earlier than CPO

Overview

This report is a summary of Nomura's 70th Global AI Trend Tracker AI expert call held on July 21, 2026, with the theme of glass bridges and advanced packaging. The expert discussion covered the technical principles and manufacturing routes of glass bridges, the impact on FAU, edge coupling versus surface grating coupling, glass carriers and ABF/glass substrates, NPO versus CPO, and 2.5D packaging. The overall conclusion is that glass bridges have clear application value in high-density optical communications, but are still constrained by loss, process stability, consistency, and coupling precision; NPO may achieve large-scale commercialization earlier than CPO, while glass core substrates are still at a relatively early stage.

Core views

The core views include: first, the essential function of a glass bridge is to adjust the mode field of an optical chip or PIC to match the mode field of an optical fiber, so it is more of a connection and conversion structure rather than a complete replacement for the FAU. Second, in glass bridge manufacturing, the ion-exchange route has a higher level of maturity, while the laser modification route faces issues of low efficiency and high optical loss. Third, edge coupling is suitable for high bandwidth and wavelength-division multiplexing, but surface coupling may be more favorable for multi-channel loss consistency in NPO/CPO multi-coupler scenarios. Fourth, glass core substrates have long-term potential, but insufficient TGV processing yield, glass brittleness, and reliability validation mean commercialization is still far off. Fifth, because NPO uses pluggable electrical interfaces, it is easier to maintain and deploy than CPO, and its commercialization timeline may therefore come earlier.

Analysis framework

The report uses an expert interview summary approach to qualitatively break down key advanced packaging technology paths: it first explains the optical connection principles and manufacturing methods of glass bridges, then compares their impact on FAUs, coupling methods, 2.5D packaging materials, and NPO/CPO architecture choices, and finally assesses industry progress and bottlenecks in conjunction with company case studies.

Methodology notes

• technology path comparisoncomparison of glass bridge manufacturing routes

  ion exchange vs. laser modification

  Ion exchange uses sodium, silver, and other ions to confine light propagation within the waveguide and is more mature; laser modification is less efficient and has higher optical loss.

• packaging architecture comparisonedge coupling and surface grating coupling

  trade-off among optical efficiency, bandwidth, and multi-channel consistency

  Edge coupling usually has higher light-emission efficiency and greater optical bandwidth, making it suitable for high-bandwidth couplers; surface coupling can improve multi-channel loss consistency through two-dimensional arrays, making it suitable for scenarios with a large number of couplers.

• commercialization timing assessmentcomparison of NPO and CPO commercialization

  maintainability, interface form, and deployment difficulty

  Compared with traditional optical modules, NPO has advantages in power consumption and bandwidth upgrades, and is also easier to maintain and commercialize than CPO, so experts judge that its large-scale commercialization timeline may come earlier than CPO.

Asset mapping & comparison

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

• Corning (GLW US)

  glass bridge technology case

  Strengths

  Uses the relatively mature ion-exchange technology, and its glass materials and optical interconnection capabilities are relevant to the glass bridge direction.

  Weaknesses

  The report does not provide the scale, orders, or earnings contribution of its glass bridge business.

  Comparison

  Compared with the laser modification route, the ion-exchange route has greater advantages in maturity and loss control.

  Risks

  There is still uncertainty around the pace of industry adoption, loss control, and process consistency for glass bridges.

• NVIDIA (NVDA US)

  surface coupling application case

  Strengths

  In NPO/CPO multi-coupler scenarios, surface-coupling two-dimensional arrays are conducive to multi-channel loss consistency.

  Weaknesses

  Vertical gratings in surface grating coupling have narrower bandwidth and relatively weaker coupling precision.

  Comparison

  Edge coupling is more suitable for high bandwidth and wavelength-division multiplexing, while surface coupling places more emphasis on multi-channel consistency.

  Risks

  There are high requirements for multi-channel coupling consistency, packaging yield, and optical loss control.

• Intel (INTC US)

  pioneer in glass core substrates

  Strengths

  Mentioned by the expert as a pioneer in developing GCS.

  Weaknesses

  The report notes that reliability test data have not yet been released.

  Comparison

  Some Korean companies are also developing related solutions, and Chinese companies such as AKM Meadville and BOE have also made sampling or cooperation progress.

  Risks

  Commercialization of GCS is still constrained by TGV processing yield, glass brittleness, and reliability validation.

• BOE (000725 CH)

  participant in glass core substrate development

  Strengths

  The expert said BOE is cooperating with Corning and is in the development stage.

  Weaknesses

  The report does not explain the mass-production timeline, customer validation progress, or revenue contribution.

  Comparison

  AKM Meadville is already sampling GCS products, while Intel and Korean companies are also advancing.

  Risks

  If GCS commercialization is delayed or reliability validation falls short, related industry contributions may be postponed.

• NPO

  advanced optoelectronic packaging direction

  Strengths

  Compared with traditional optical modules, it has advantages in power consumption and bandwidth upgrades, and is easier to maintain and commercialize than CPO.

  Weaknesses

  It still needs to solve challenges related to the number of couplers, loss consistency, and packaging integration.

  Comparison

  Experts believe the timeline for large-scale commercialization of NPO may be earlier than that of CPO.

  Risks

  The commercialization pace depends on adoption by system vendors, packaging yield, optical loss control, and maintenance cost.

• CPO

  advanced optoelectronic packaging direction

  Strengths

  A long-term architecture aimed at higher integration and bandwidth density.

  Weaknesses

  Maintenance and commercialization are more difficult than for NPO.

  Comparison

  NPO uses pluggable electrical interfaces and is easier to maintain, while CPO places more emphasis on deep co-packaging.

  Risks

  Coupling consistency, repair and replacement difficulty, ecosystem maturity, and cost may affect the pace of deployment.

Key data

• conference date2026-07-21The report shows that AI Expert Call #70 was held on July 21, 2026, and ProductionComplete was 2026-07-21 12:04 UTC.
• themeGlass bridge and advanced packagingCovers glass bridges, FAU, edge coupling, surface grating coupling, glass carriers, glass core substrates, NPO, CPO, and 2.5D packaging.
• main application of glass bridgeshigh-density optical communication scenariosThe expert pointed out that glass bridges are mainly used in high-density optical communications and achieve mode field matching through a bridging structure.
• glass bridge manufacturing routeion exchange is more mature; laser modification is less efficient and has higher lossThe report mentions that Corning's glass bridge uses ion-exchange technology.
• example of CPO fiber count6.4T CPO requires 32 optical fibers at 200G per laneUsed to illustrate that NPO/CPO require a large number of couplers, and therefore have very high requirements for coupling consistency.
• commercialization status of GCSstill far from commercializationAlthough it has advantages such as suppressing deformation, reducing electrical loss, and supporting higher bandwidth, insufficient TGV yield, glass brittleness, and lack of reliability data remain challenges.
• Nomura global equity research rating distributionBuy 58%; Neutral 39%; Reduce 3%This data comes from the report disclosure page and does not represent the report's rating on the related companies.

Impact & implications

The implication for the industry chain is that upgrades in AI computing networks continue to tighten bandwidth and power-consumption constraints in optical communications, making optoelectronic interconnection, coupling consistency, and material stability in advanced packaging key competitive points. In the short to medium term, more attention should be paid to NPO, the choice between edge and surface coupling solutions, and the introduction of glass bridges in high-density scenarios; over the longer term, focus should be on the yield, reliability, and commercialization milestones of glass core substrates.

Risks

• Loss control, process stability, and consistency of glass bridges remain core technical bottlenecks.
• High-density optical communications impose high requirements on coupling precision and deformation control, and mass-production yield remains uncertain.
• TGV processing yield for glass core substrates is low and the glass material is highly brittle, so commercialization may come later than market expectations.
• Reliability test data for GCS have not yet been fully disclosed, and industry validation is insufficient.
• The commercialization pace of NPO and CPO depends on adoption by system vendors, the maturity of the packaging ecosystem, and maintenance costs.
• The report is an expert conference summary and does not provide quantifiable revenue, profit, or order forecasts, so investment judgment needs to be combined with subsequent company disclosures.

What to watch

• Customer adoption and mass-production validation progress of Corning's ion-exchange glass bridge solution.
• The actual path chosen by AI system vendors such as NVIDIA between surface coupling and edge coupling in NPO/CPO.
• The large-scale commercialization timetable, interface standards, and supply chain maturity of NPO relative to CPO.
• Progress by Intel, Korean vendors, AKM Meadville, and BOE in GCS samples, reliability testing, and mass-production yield.
• TGV processing yield, improvements in glass substrate brittleness, and packaging warpage control metrics.
• Changes in demand for glass carriers in 2.5D packaging as AI chip packaging complexity increases.