Theses›Semiconductor manufacturing›Photonic foundry capacity

Photonic foundry capacity

By 2028, photonic IC foundry capacity binds production volume of co-packaged optics and photonic compute, with fewer than 5 commercial multi-project 300mm photonic foundries operating globally.

  • Closing
  • $2.7B by 2035

The call, a draftScarcity is winning: optical-chip capacity rises more than 80% in 2026 and still falls 5-15% short of demand. That view is already consensus, the capacity sits with listed foundries, and the window is late.

D1D2D3D4R1R2R3C1foragainst D1D2D3D4R1R2R3C1foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the drivers.
  1. D1Optical-chip capacity rises more than 80% in 2026 and still falls 5-15% short of demand.
  2. D2Lumentum put its laser supply shortfall at more than 30% in May 2026, and NVIDIA has pre-allocated capacity at the main laser suppliers.
  3. D3Barium titanate needs processing above 400C, which no shared CMOS line allows, so exotic materials need their own dedicated lines.
  4. D4Light is more sensitive than electrons to sidewall roughness, so designs prototyped on e-beam often fail to transfer to volume foundries.
  5. R1Tower is tripling silicon photonics capacity by mid-2026, GlobalFoundries bought Advanced Micro Foundry and UMC licensed imec's 300mm process, adding supply.
  6. R2TSMC targets COUPE volume in 2026 and Samsung a turnkey co-packaged optics service by 2029, the closest any large foundry has come to opening up.
  7. R3Small designers still get generous 300mm silicon photonics attention, with Samsung and UMC lined up as second sources behind GlobalFoundries.
  8. C1AI co-packaged optics demand passes 1B photonic chiplets a year.31 Dec 2028
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today1234567predictionswatch 20272028202920302031today1234567predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1No more than 5 commercial multi-project 300mm photonic foundries are operating at end-2027; a credible 6th entering pressures the <5-by-2028 scarcity thesis.Dec 2027 · open
  2. 2Whether a sixth open 300mm photonic foundry is operating by end-2027.Dec 2027 · open
  3. 3AI co-packaged optics demand exceeds 1B photonic chiplets per year by 2028.Dec 2028 · open
  4. 4EU-located commercial photonic foundry capacity remains under 100,000 wafer-starts/year through 2028.Dec 2028 · open
  5. 5Whether TSMC's COUPE line or Samsung's planned service opens to third-party designs.Dec 2029 · open
  6. 6Fewer than 5 commercial multi-project 300mm photonic foundries operate globally in 2030.Dec 2030 · open
  7. 7Whether fewer than 5 open 300mm photonic foundries are operating.Dec 2030 · open
Expressionslisted, private, technologies
Private · 23 private companies, 11 of them in the US.

IMEC · AIM Photonics · CompoundTek · OpenLight Photonics · Aurrion · Quintessent · HyperLight · Intel Foundry Services (Intel IFS / Silicon Photonics) · POET Technologies · Luxtera · Bright Photonics · SiPhox · VLC Photonics · VPIphotonics · Elytra · Ayar Labs · Anello Photonics · EDWATEC · Lightelligence (Shanghai Xizhi) · Mixx Technologies · nanoPHAB · Photon IP · X-Celeprint

37 more private companies tracked, not shown

Technologies
What changedthe record, newest first
  • 23 Jun 2026TSMC is taking its silicon-photonics engine to volume production in 2026 and Samsung is targeting a turnkey service by 2029, the closest yet to a large foundry entering. The capacity-scarcity bet still holds if those stay captive platforms for their anchor customers; it breaks only if they open the line to everyone.
  • 17 Jun 2026A foundry partner expects silicon photonics to reach ~70 percent of its revenue within two years and is exiting other lines — demand is concentrating onto the few photonics-capable foundries, exactly the capacity squeeze the thesis predicts.
  • 1 Jun 2026A pattern is forming across photonics deals: the binding production constraint is not the number of fabs but process-incompatibility on shared CMOS lines, and the missing design/PDK layer on the dedicated lines that do exist.
Reading
Show the reading