Glass bridge and advanced packaging: NPO commercialization may precede CPO, while glass core substrates remain at an early stage

Overview

This report is the 70th AI expert call note in Nomura Global AI Trend Tracker, focused on glass bridges and advanced packaging. Experts discussed advanced packaging technologies in optoelectronic solutions, including the technical principles, manufacturing methods, loss control, and process stability of glass bridges, analyzed their impact on FAUs, and compared paths such as edge coupling, surface grating coupling, glass interposers, ABF glass substrates, NPO, CPO, and 2.5D packaging.

Core views

The core view is that glass bridges have clear application value in high-density optical communications, but are still constrained by loss, process consistency, coupling precision, and material warpage; edge coupling is better suited for high bandwidth and wavelength-division multiplexing, while surface coupling is better suited for NPO/CPO scenarios requiring a large number of consistent channels; glass core substrates have significant long-term potential, but are still far from commercialization; NPO, because it uses pluggable electrical interfaces and is easier to maintain and commercialize than CPO, may achieve large-scale adoption earlier.

Analysis framework

The report uses expert interviews and technology path comparison, breaking down the functional relationships among glass bridges, FAUs, coupling methods, interposers and substrates, NPO, and CPO by application stage, and illustrates industry progress with cases including Corning, NVIDIA, Intel, AKM Meadville, and BOE.

Methodology notes

• Expert interviewAI Expert Call

  Derive technology path and commercialization judgments through industry expert calls

  The report is not a traditional company deep dive driven by financial models, but rather summarizes technical principles, bottlenecks, supply chain participants, and commercialization timelines based on the views of experts in advanced packaging and optical interconnects.

• Technology comparisonComparison between edge coupling and surface grating coupling

  Compare the differences between two optical coupling solutions in efficiency, bandwidth, multi-channel consistency, and applicable scenarios

  Edge coupling offers higher light-emission efficiency and greater optical bandwidth, making it suitable for high bandwidth and wavelength-division multiplexing; although surface grating coupling has narrower vertical grating bandwidth, its 2D array is more conducive to multi-channel loss consistency.

• Commercialization stage assessmentNPO versus CPO path comparison

  Assess the deployment order of optical interconnect packaging paths from the perspectives of power consumption, bandwidth upgrade, maintenance difficulty, and interface form

  Experts believe that compared with traditional optical modules, NPO has advantages in power consumption and bandwidth upgrades, and compared with CPO, its pluggable electrical interface makes it easier to maintain and commercialize, so its scaling timeline may precede that of CPO.

Asset mapping & comparison

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

• Corning (GLW US)

  A company related to glass bridge technology; the report states that its glass bridge uses ion-exchange technology.

  Strengths

  Ion-exchange technology is relatively mature and suitable for forming waveguides in glass and confining light propagation.

  Weaknesses

  The report does not provide Corning's mass-production scale, cost structure, or customer progress for this business.

  Comparison

  Compared with the laser modification route, ion-exchange technology has higher maturity and may offer more controllable optical loss.

  Risks

  Glass bridges as a whole still face challenges in loss control, process stability, coupling precision, and warpage control.

• NVIDIA (NVDA US)

  Mentioned in the report as an application case for surface coupling.

  Strengths

  In NPO/CPO scenarios requiring a large number of couplers, the 2D array of surface coupling is beneficial for multi-channel loss consistency.

  Weaknesses

  The vertical grating bandwidth of surface grating coupling is relatively small, and coupling precision is comparatively weaker.

  Comparison

  Compared with edge coupling, surface coupling sacrifices some bandwidth and efficiency but is better suited to packaging solutions with extremely high requirements for multi-channel consistency.

  Risks

  High-channel-count packaging imposes high requirements on manufacturing consistency, loss control, and reliability.

• Intel (INTL US)

  The report states that it is a pioneer in developing glass core substrates.

  Strengths

  It holds an early R&D position in advancing technologies such as GCS for advanced packaging substrates.

  Weaknesses

  The report notes that no reliability test data has yet been released.

  Comparison

  Compared with other companies still in development or sampling stages, Intel is described by the expert as a pioneer in GCS development.

  Risks

  The brittleness of glass leads to low TGV processing yield, and insufficient reliability validation may delay commercialization.

• BOE (000725 CH)

  The report states that BOE is cooperating with Corning and is in the development stage.

  Strengths

  It has industrial synergy opportunities to participate in development related to glass core substrates.

  Weaknesses

  It is still in the development stage and has not yet translated into clear mass production or commercial revenue.

  Comparison

  Unlike AKM Meadville's sampling progress, BOE is described mainly in terms of cooperative development with Corning.

  Risks

  GCS commercialization is still distant; TGV yield, material brittleness, and reliability data are key uncertainties.

• AKM Meadville (unlisted)

  The report states that it has already sent samples of glass core substrate products.

  Strengths

  Among Chinese companies, progress in sending GCS product samples has already emerged.

  Weaknesses

  It is unlisted, and the report does not provide commercialization, customer validation, or mass-production data.

  Comparison

  Compared with companies still in the development stage, sample delivery indicates it is closer to the customer validation stage, but it is still not equivalent to mass production.

  Risks

  The GCS path as a whole is still affected by low yield, glass brittleness, and the lack of reliability data.

Key data

• Expert call date2026-07-21The report states that the call was held on July 21, 2026, and the report production completion time was 2026-07-21 12:04 UTC.
• 6.4T CPO fiber count example32 fibersThe expert gave the example that under 200G per lane, 6.4T CPO requires 32 fibers, and therefore has very high requirements for coupling consistency.
• Nomura Buy rating distribution58%As of June 30, 2026, 58% of companies in Nomura Group Global Equity Research were assigned a Buy rating.
• Nomura Neutral rating distribution39%As of June 30, 2026, 39% of companies were assigned a Neutral rating.
• Nomura Reduce rating distribution3%As of June 30, 2026, 3% of companies were assigned a Reduce rating.

Impact & implications

The implication for investment research is that upgrades in AI computing infrastructure rely not only on GPUs and switching chips, but will also continue to drive technology iteration in optical interconnects, advanced packaging, glass materials, substrates, and high-precision coupling processes. In the short to medium term, it is more worthwhile to focus on NPO, surface coupling consistency, glass bridge loss control, and glass interposers for 2.5D packaging; although glass core substrates have long-term potential, their commercialization progress still depends on TGV yield, reliability validation, and supply chain maturity.

Risks

• Insufficient loss control in glass bridges may affect the efficiency of optical communication links.
• Insufficient process stability and consistency may limit mass production of high-density packaging.
• High requirements for coupling precision and material warpage may increase manufacturing difficulty and cost.
• TGV processing yield for glass core substrates is low, and the brittleness of glass materials brings reliability risks.
• GCS still lacks publicly available reliability test data, and commercialization timing may lag market expectations.
• CPO is difficult to maintain and commercialize, and the pace of scaled deployment remains uncertain.

What to watch

• Customer adoption and mass-production progress of Corning's ion-exchange glass bridge technology in high-density optical communications.
• Changes in the choice between surface coupling and edge coupling in NPO/CPO by AI hardware platforms such as NVIDIA.
• The evolution of requirements for channel count, loss consistency, and coupling precision in 6.4T and higher-speed optical interconnect solutions.
• Reliability testing, sample delivery, and mass-production milestones for glass core substrates at Intel, Korean manufacturers, AKM Meadville, and BOE.
• The large-scale commercialization timeline of NPO relative to CPO, as well as its actual performance in power consumption, bandwidth upgrades, and maintainability.
• The application scope of glass interposers in 2.5D packaging for warpage control of molding materials.