Skip to content
primarysourced Photonics sector GF
GFS
~5 min read · 1,048 words ·updated 2026-04-29 · confidence 50%

02 — Technology

Section role

This section answers what does GlobalFoundries actually fab in the photonics layer of the AI value chain, and how is the foundry positioned versus the other commercial silicon-photonics platforms. It traces the Fotonix monolithic silicon-photonics process (the platform under which the 45SPCLO and 9WG PDK variants are sold), the Advanced Micro Foundry (AMF) Singapore acquisition (closed 2025-11-17), the customer-PDK status with Lightwave Logic, NLM Photonics, Marvell, and Ayar Labs, and the broader specialty-process portfolio that surrounds the photonics line at the Malta NY (Fab 8), Dresden, and now Singapore facilities.

GlobalFoundries is not a leading-edge logic foundry — TSMC and Samsung occupy 5/3/2 nm — but in specialty processes (RF-SOI, FDX FD-SOI, BCD power, SiGe BiCMOS) and especially silicon photonics the company has built a defensible position. The November 2025 AMF acquisition reframed GF as the largest pure-play silicon-photonics foundry by revenue (GF press release), and the Lightwave Logic GDSFactory PDK release of 2026-03-16 made GF the first commercial 300mm SiPh foundry to ship a productized electro-optic-polymer modulator design flow (Stocktitan).

Scope discipline

The KB scope is deliberately narrow on the photonics layer. The full GF process roadmap (40LP, 28SLP, 22FDX, 12LP, 45RFSOI, 130BCDLite, 90 SiGe BiCMOS, etc.) is treated as cross-context — it explains GF’s revenue mix, capex cycle, and how the foundry funds its photonics R&D, but it is not the load-bearing analyst question for this site. The load-bearing questions are:

  1. How does Fotonix 45SPCLO + 9WG stack up technically against Tower PH18, TSMC SiPh, Intel SiPh, and IBM/IMEC research foundries?
  2. How does the AMF Singapore capability complement Fotonix — is it additive (200mm vs 300mm, different design rule space, A*STAR research access) or redundant?
  3. Which photonics customers are committed at which stage, and what is the volume / revenue translation timeline?
  4. What is the EO-polymer (LWLG, NLM) and TFLN integration path on Fotonix, and on what timeline does it move from PDK to volume tape-out?

What’s here

  • Fotonix process overview — the 45SPCLO node specifics: monolithic 300mm SOI CMOS+photonics, 9-waveguide PDK variant, integrated photonic device library (modulators, photodetectors, grating couplers, edge couplers, polarization splitter-rotators, phase shifters), 300 GHz-class RF-CMOS integration, 9-level BEOL metallization, far-BEOL V-groove fiber attach, and the 0.5 Tbps/fiber data-rate hero number.
  • Fotonix PDK customers — disclosed Fotonix licensees: Lightwave Logic (Mar 2026 GDSFactory PDK live), NLM Photonics (Mar 2026 sampling on AMF HP process; Fotonix follow-on in development), Marvell (Inphi-lineage SiPh + Polariton-acquired POH integration path), Ayar Labs (TeraPHY optical-I/O chiplets, UCIe-compliant, volume ramp 2025-2027), plus Lightmatter / Ranovus / PsiQuantum / Xanadu / Enosemi.
  • AMF Singapore acquisition — Nov 17 2025 deal mechanics, AMF’s ASTAR / IME spin-out heritage, 200mm vs Fotonix 300mm comparison, structural impact on POET’s prior SilTerra/AMF triangulation (a cross-company implication flagged in the POET KB), plans to scale AMF to 300mm “as market needs grow”, and the new Center-of-Excellence partnership with ASTAR for 400 Gb/s next-generation materials research.
  • Process roadmap — Fotonix 1.0 → next-gen roadmap (the 2024 “next-generation Fotonix” enhancement covering modulator topologies, packaging, PDK compact models, and in-house test capability), AMF process modernization, EO-polymer (LWLG) and TFLN (NLM, third-party) integration scheduling, 1.6T → 3.2T transceiver-generation alignment.
  • Monolithic vs chiplet — GF’s monolithic CMOS+photonics positioning vs the chiplet-based path (UCIe/CoWoS at TSMC, Intel EMIB), tradeoff analysis on yield, cost, performance per watt, and time-to-market.
  • Papers and conferences — bibliography of GF SiPh papers and conference talks 2023-2026 (OFC, ECOC, IEEE Photonics, CLEO, Photonics West), with DOIs where available.
  • Patents overview — GF’s photonics-relevant patent portfolio at a high level (counts, key claim families, the AMF-acquired IP package, freedom-to-operate context vs LWLG / NLM material-science patents).

Reading order

Discipline reminders

GF communicates technology updates through three channels: (a) GF Newsroom press releases, (b) OFC / ECOC / IEEE Photonics / CLEO conference papers (often co-authored with customer engineering teams), and (c) co-disclosure with named customers (LWLG GDSFactory PDK announcement, NLM Photonics press release, Ayar Labs UCIe chiplet launch). Aggregator sources (TrendForce, SDxCentral, EE News Europe) summarize but rarely add new primary fact — every claim in this section should resolve to a GF press release, an investors.gf.com filing, an Optica / IEEE Xplore DOI, or a customer co-disclosed press release.

The November 2025 AMF acquisition closed without disclosed terms — purchase price, employee count, AMF customer list, and integration timeline are all not in the public 8-K. Where this section makes claims about the deal mechanics it cites the GF press release verbatim and flags inferred items.

Cross-section pointers

  • overview — corporate history, IPO, Mubadala control structure (relevant because a sovereign-funded foundry has different strategic patience than a public-pure-play).
  • customers photonics — per-customer engagement detail.
  • competitors — head-to-head vs Tower PH18, TSMC SiPh, Intel SiPh, IBM, IMEC.
  • overview — capex cycle, photonics segment disclosure (or lack thereof).
  • overview — bull / bear / catalyst frame.

Confidence legend

  • ✓ — verified via primary source (SEC filing, GF press release, customer co-disclosed release, peer-reviewed paper, conference proceedings DOI)
  • ◐ — partial / aggregator-derived / secondary citation
  • ⚠ — inferred or estimate (no primary source available; flagged for follow-up audit)

Cross-company consistency notes

  • LWLG KB (platform overview) treats Fotonix as the carrier for Perkinamine polymer; this section’s coverage must align with the LWLG-side dates and PDK status (GDSFactory PDK live 2026-03-16; validation tape-outs scheduled for 2026 at 200G/400G per lane).
  • POET KB (foundry relationships) explicitly identifies AMF Singapore as a foundry alternative to GF Fotonix that POET did not select (POET went to SilTerra Malaysia); the November 2025 GF-AMF acquisition collapses one branch of POET’s optionality and is a cross-company implication worth surfacing.
  • MRVL KB (foundry relationships) treats GlobalFoundries as a “specialty / RF-analog” foundry only (<5% of Marvell wafer spend) — but the photonics layer is the live integration story post-Polariton acquisition, and Inphi-lineage products that pre-date Marvell’s TSMC concentration may still ship from GF Fotonix. The two KBs must be reconciled as the Marvell-Polariton-Fotonix integration matures.