Source

@AlmaCap114204 attaches a TODAY vs FUTURE diagram of the photonics value chain — the cleanest material-transition mapping we've captured for the optical thesis. Held positions: $TSEM, $SOI, $ALMU, $LPKF.

TODAY architecture:

• Wafer: Soitec SOI for SiPho ($SOI.PA) + AXTI InP for lasers (China-made)
• Chip: TSEM SiPho on SOI + separate laser chip (InP, China)
• Substrate: Organic ABF (Ibiden, Shinko, Unimicron — Asian, no US-listed proxy)
• Package: GPU + SiPho + external pluggable laser

FUTURE architecture:

• Wafer: Soitec SOI (still needed) + ALMU QD laser on silicon (US-made)
• Chip: TSEM fabs SiPho with ALMU laser inside on Soitec SOI
• Substrate: Glass (replaces ABF); LPKF drills the TGV (Through-Glass Via) holes
• Package: GPU + chiplets + on-chip light

Why future wins: glass is flat (more chiplets fit), ALMU laser eliminates China dependency, light on-chip lowers power, faster chip-to-chip data.

Stated TODAY problems: ABF warps at large package sizes, InP wafers China-made, external lasers waste power, chiplet density capped.