The synthesis page for “how you package a photonic IC.” Full decision brief:.
The core idea — two axes, not one
“Packaging a PIC” is constantly conflated with “which PIC platform wins.” They are different axes:
- Platform axis — what the PIC is made of (where light is generated/modulated/detected): Silicon Photonics, InP Photonics, SiN, TFLN, polymer, glass. Covered by Photonics Material Class War /.
- Packaging axis — how you connect the PIC: a 4-layer stack, each layer its own contest. This page.
Silicon Photonics is the platform; Glass Photonics, Photonic Wire Bonding, Hybrid Bonding (Cu-Cu) and Co-Packaged Optics are competing answers to the packaging axis. They are mostly complementary, not rival: a SiPh die does the electro-optics, and the packaging stack gets light + power on and off it.
Why packaging is a sub-industry
Silicon’s high index contrast makes the on-chip mode ~0.5 µm vs a fibre’s ~10 µm — a ~20× mode mismatch that no glue can bridge. That, plus silicon’s lack of a native laser and its two-photon absorption at high power, generates the whole stack below. Commercially this is where the value is: packaging is ~25% of a datacom pluggable’s value, ~50% of early-CPO value; Yole sizes photonics packaging at $4.5B (2025) → $14.4B (2031), ~21.5% CAGR ().
The four layers
| Layer | Problem | Competing approaches | Glass’s role |
|---|---|---|---|
| 1. Fibre→chip coupling | bridge the ~20× mode mismatch | edge/grating coupling · polymer Photonic Wire Bonding · glass waveguide · fibre arrays (PROFA) | competes directly — the inorganic, reliability-winning option |
| 2. Laser / light-source attach | silicon can’t lase | external laser · flip-chip · Heterogeneous Photonic Integration · monolithic InP Photonics · micro-LED | absent |
| 3. Die-to-die / die-to-substrate | fine pitch, no optical misalignment | flip-chip bump · Hybrid Bonding (Cu-Cu) · 2.5D interposer (Glass Interposers) | glass-core substrate (primes’ game) |
| 4. System co-packaging | optics next to ASIC | Co-Packaged Optics vs pluggable vs LPO/LRO | sets timing for all above |
Layer 1 is the contested, venture-relevant layer and the one the Glass Photonics cohort lives in. See the cohort screen in.
The investable spine
Companies / organisations using
Connected ideas
Sources
Frontier (open questions)
- Which coupling approach wins volume CPO, and does it differ by platform?
- Does the active-optical-alignment tooling gap become a standalone equipment market?
- Do all fabless coupling specialists get absorbed (Teramount→Molex pattern), leaving captive + equipment only?