Corning GlassBridge: Its Glass Materials Dominance Is Extending From Packaging Substrates to the Fiber Coupling Layer
TL;DR On June 24, 2026, Corning unveiled GlassBridge, which uses wafer-level ion-exchange glass waveguides to couple fiber directly into silicon photonics chips. The market rushed to declare that "FAUs will be replaced," but the real story isn't who rises or falls: GlassBridge is Corning's second move to extend its materials dominance in the "electrical layer" of glass substrates (TGV, low warpage) into the "optical layer".
1. Why now: Corning brings fiber coupling onto the glass table
The most tedious and most critical part of co-packaged optics (CPO) is "how to attach a fiber array to a photonic integrated circuit (PIC) with low loss and at volume." The conventional approach is the FAU (fiber array unit), which relies on high-precision physical micro-assembly to align and bond fibers. Corning's move aims to replace this manual micro-assembly with a wafer-level process.
GlassBridge is a fiber-to-PIC platform: wafer-level ion-exchange (IOX) glass waveguides, passive alignment, pitch of 30 µm or more, and a coupling loss target below 2 dB. The industry sees it as an evolutionary replacement for the iFAU in CPO architectures. And it's not a surprise attack: word had been circulating since September 2025, it was already part of Corning's US$10 billion photonics business plan, and it builds on Meta's US$6 billion AI fiber order with Corning in January this year.
2. One piece of glass, carrying both electrons and photons
Glass substrates should be viewed as two layers. The electrical layer relies on through-glass vias (TGV), low dielectric loss and low warpage to support large packages; we covered that race in detail in Glass substrates are no longer slideware: why the TGV race arrives all at once in 2026. The optical layer is what GlassBridge adds. In its ECTC 2026 paper, Corning disclosed that it screened 200 formulations to create a specialty glass with a silver-ion diffusivity of about 5E-22 m²/s; refractive index changes by less than 1.5% over five years at 110°C, bend loss drops from 4 dB/cm to 0.01 dB/cm, and a 16-channel fan-out converting 250 µm to 50 µm pitch loses only 0.62 dB fiber-to-fiber. Electrical and optical layers live on the same piece of glass.
3. The most expensive part of CPO is exactly what it targets
Whether this positioning makes sense, just listen to potential customers. At its 2026 Tech Day, Lightmatter discussed packaging economics: silicon photonics packaging plus test accounts for 60-70% of total cost (versus only 20-30% for conventional CMOS), and "the main challenge is the fiber"; "every customer's FAU requirements are different"; the current approach "doesn't scale, so we're pushing fiber attach to wafer level"; and "for volume production you need to align in seconds, not minutes; FAU is one of the industry's problems right now."
Put those four statements together and the cost black hole of CPO volume production is "fiber coupling plus FAU": expensive, fragmented and manual. GlassBridge's wafer-level passive alignment aims squarely at that hole.
That's the summary of this article.
STT's full analysis, covering why GlassBridge is "the second move of a materials hegemon," exactly which stage takes the hit, how the 2028-2030 timeline plays out, and why what really holds it back is a commercial hurdle rather than a technical one, is in the paid section.
👉 Subscribe to STT premium and read the full version here: Read the full analysis on vocus (in Chinese)
Comments