Skip to content
GFS
~7 min read · 1,724 words ·updated 2026-04-29 · confidence 42%

GF photonics customer roadmaps

Updated: 2026-04-29 Status: ✓ Roadmap milestones from primary-source customer disclosures; volume-ramp dates flagged ⚠ where forward-looking and speculative. Cross-references: Customers — photonics · process roadmap · fotonix pdk customers


1. Roadmap framing

The photonics-customer roadmap is the forward-indicator timeline for Fotonix utilization. Each customer’s first-silicon-to-volume timeline depends on three critical-path items:

  1. GF process readiness — the Fotonix PDK rev that supports the customer’s modulator architecture (slot waveguide for LWLG, AMF HP for NLM, plasmonic-slot for Polariton-Marvell)
  2. Customer reliability validation — Telcordia GR-468 85/85, mechanical shock, vibration, etc., which can take 1000-2000 hours per validation cycle
  3. End-customer (transceiver / hyperscaler / chiplet-host) qualification — requires Tier-1 customer engineering validation, often 6-12 months from first silicon

This file synthesizes per-customer roadmaps with dependencies flagged.


2. Lightwave Logic — roadmap

MilestoneDateStatusDependency
LWLG-AMF partnership2024-05-13✓ donesub-1V drive at 200 Gbps PAM4 demonstrated
Telcordia GR-468 85/85 1000h2025-07-15✓ pass1.6% avg loss; >11 samples
GF Fotonix PDK live (GDSFactory)2026-03-16✓ doneSlot-waveguide process modification
Validation tape-outs at 200G/400G per lane2026-Q2-Q4⚠ plannedCustomer designs in flight
First customer volume revenue (Stage 4)2027⚠ plannedTier-1 customer commitment + GF wafer slot booking
2,000h length-swept reliability2026-04-22✓ passper Blum at PIC Intl 2026

Critical-path dependencies:

  • GF slot-waveguide process modification must reach production rev (currently in optimization per Mar 2026 disclosure)
  • Tier-1 customer Stage-3 graduation to Stage 4 — LWLG has 4 Fortune Global 500 in Stage 3 (per LWLG KB); the Stage-3-to-Stage-4 conversion timing is the load-bearing variable
  • LWLG polymer reliability — Telcordia 85/85 cleared at 1000h in Jul 2025 and 2000h in Apr 2026; further endurance qualification continues as customer-specific requirement

LWLG dual-foundry path (Tower PH18 + GF Fotonix + SilTerra) is a structural backstop — losing GF Fotonix volume would be partially offset by Tower or SilTerra capacity.


3. NLM Photonics — roadmap

MilestoneDateStatusDependency
Selerion-BHX material announced2025-06-04✓ doner33 = 1000 pm/V
SOH PIC sampling on AMF GP2026-03-16✓ doneSampling shipped
OFC 2026 demonstration2026-03-17✓ doneShowcase at OFC LA
AMF HP PDK 6.0 GA release2026-Q3-Q4 ⚠PendingGF process development complete
First NLM commercial product samples2027⚠ plannedCustomer NRE + Tier-1 Customer commitment
Volume revenue (NLM customer Stage 4 equivalent)2028+⚠ plannedCustomer transceiver qualification

Critical-path dependencies:

  • AMF HP PDK 6.0 release is the gating item; in development per the 2026-03-16 announcement
  • NLM customer commitment — NLM has been less public about its specific Tier-1 customers than LWLG (which has the Stage 1-4 framework); ⚠ follow-up audit needed
  • Selerion-HTX reliability — composite material reliability data parallel to LWLG Perkinamine but less public; Telcordia and aging data ⚠ to verify

NLM’s roadmap is structurally one cycle behind LWLG — LWLG cleared 1000h reliability July 2025 and has commercial PDK live; NLM is sampling. NLM’s higher claimed r33 (1000 pm/V vs LWLG ~200 pm/V) is the offsetting marketing pitch but doesn’t yet translate to commercial timeline advantage.


4. Marvell — roadmap

MilestoneDateStatusDependency
Inphi-lineage 100/200/400 G PAM4 SiPh transceiversProduction✓ ongoingVolume on Fotonix
Marvell Ara 1.6T DSP (TSMC 3nm)2025✓ winner of Leading EDGE awardTSMC capacity
Marvell Ara X / Spica / Petra family2026 H2 - 2027⚠ in developmentTSMC 3/2 nm capacity + Fotonix photonic die
Polariton acquisition closing2026 H2 ⚠Announced 2026-04-22; expected H2 closeRegulatory + integration
Marvell-Polariton POH integration on Fotonix2027-2028 ⚠Pre-PDK; process modification phaseGF slot-waveguide-plasmonic process modification + Marvell IP integration
Volume Polariton-modulator-equipped transceiver2028+ ⚠SpeculativeAll upstream items

Critical-path dependencies:

  • Polariton acquisition close (regulatory; H2 2026 per Marvell IR)
  • Fotonix process modification for plasmonic-slot waveguides — different from but adjacent to LWLG slot-waveguide modification
  • Marvell-customer transceiver qualification for the post-Polariton 3.2T product class

The Marvell roadmap is the highest-impact single roadmap for GF photonics utilization. Marvell’s volume scale + DSP-and-photonic-die-co-ownership creates a multi-year design pipeline that GF needs to keep on Fotonix rather than allowing migration to TSMC SiPh + CoWoS hybrid.

Cross-reference: polariton acquisition confirms 12-18 month integration timeline; the Fotonix-process-modification critical path aligns with that.


5. Ayar Labs — roadmap

MilestoneDateStatusDependency
TeraPHY 4 Tbps optical chipletPre-2025✓ doneProduction
TeraPHY 8 Tbps UCIe optical chiplet announcement2025-03-31✓ doneUCIe Co-Packaged Optics Standard adoption
”Thousands of engineering samples” shipping2025 ongoing✓ ongoingManufacturing / packaging / test process maturation
First commercial offerings2026-2028⚠ rampStrategic backer (NVIDIA / Intel / Microsoft) commitment to specific scale-up architecture
AI-fabric volume ramp2027-2028⚠ plannedHyperscaler / GPU vendor architectural commitment

Critical-path dependencies:

  • UCIe optical-chiplet ecosystem maturity — UCIe spec adoption across host chiplets (CPU, GPU, AI ASIC); rate of UCIe-equipped host design wins is the load-bearing variable
  • Strategic backer scale-up architecture commitment — NVIDIA, Intel, Microsoft, AMD are all backers but a specific platform commitment to TeraPHY for AI-fabric scale-up
  • Fotonix wafer capacity — at volume, capacity becomes binding for Ayar Labs
  • External-laser supply — SuperNova 16-wavelength laser source is an Ayar Labs proprietary; supply scaling

The Ayar Labs roadmap is the highest-uncertainty roadmap for GF — large commercial upside if UCIe optical wins broadly, but the architecture remains a bet vs the dominant CPO and pluggable approaches.


6. Lightmatter — roadmap

MilestoneDateStatusDependency
Mars / Envise / Passage productization2024-2025✓ ongoingProduction on 9WG
45CLO products in development2026-2027 ⚠In devMigration to higher-density CMOS+photonics
Volume AI-inference customer wins2026-2027⚠ ongoingCustomer acceptance vs GPU / TPU

Critical-path dependencies:

  • AI-inference market acceptance of photonic-electronic accelerators vs incumbent GPUs / TPUs / Cerebras
  • 45CLO PDK mature for AI-accelerator workloads — Lightmatter migration requires Fotonix process at customer-acceptable maturity
  • Customer-design pipeline — Lightmatter’s customer base is less public than Marvell / Ayar Labs; ⚠ follow-up audit

7. PsiQuantum, Xanadu — quantum roadmaps

PsiQuantum and Xanadu are pre-commercial; their roadmaps are tied to:

  • Photonic quantum computer system-level milestones (qubit count, gate fidelity)
  • Government / hyperscaler quantum-program funding cycles
  • Fotonix low-loss waveguide process refinement for sub-dB-per-cm waveguide losses required for quantum coherence

⚠ Volume timeline for quantum-relevant photonic-IC fabrication is 2027-2030+ class; not near-term GF revenue impact.


8. NVIDIA, Broadcom, Cisco — strategic-customer roadmaps

These customers’ Fotonix roadmaps are less publicly granular:

  • NVIDIA — optical interconnect roadmap tied to Quantum-X / Spectrum-X switch platforms; specific Fotonix milestones ⚠ not detailed
  • Broadcom — Tomahawk-Optical CPO ramping 2024-2025-2026; specific Fotonix vs internal-fab share ⚠ not detailed
  • Cisco — internal-IP-dominated; merchant SiPh share less material to GF revenue

⚠ Specific Fotonix usage roadmaps for these customers needs ongoing tracking.


9. Aggregate roadmap dependency map

GF process readiness itemCustomers blocked / advanced
Fotonix slot-waveguide modification (in opt.)LWLG, possibly NLM (if NLM moves to Fotonix Malta)
Fotonix plasmonic-slot waveguide modification (TBD)Marvell-Polariton
AMF HP PDK 6.0 GA (TBD)NLM Photonics, AMF inherited customers
Fotonix CPO substrate maturity (in dev.)Broadcom Tomahawk-Optical, NVIDIA, Marvell
9WG-to-45CLO migration smoothLightmatter, others on 9WG
Photonic test capacity scalingAll photonics customers — capacity as volume builds

The two highest-impact GF process items for the 2026-2027 commercial inflection are:

  1. Slot-waveguide process modification production rev — gates LWLG (and possibly NLM) volume scaling
  2. AMF HP PDK 6.0 GA — gates NLM volume scaling

Beyond 2027, the plasmonic-slot waveguide modification for Marvell-Polariton is the next gating item.


10. Volume revenue inflection — synthesis

YearInflection eventRevenue implication
2026 H2Validation tape-outs LWLG, NLM HP PDK 6.0 GANRE revenue + sample shipments
2027 H1First commercial-volume LWLG modulator transceivers + Marvell post-Polariton integrationVolume wafer revenue ramp begins
2027 H2 - 2028Volume-ramp on multiple modulator architectures simultaneouslyMaterial Fotonix utilization growth
2028+3.2T transceiver class + Polariton POH volume + Ayar Labs UCIe AI-fabric volumeLargest single Fotonix capacity expansion required

⚠ All timelines are forward-looking and dependent on the dependency items flagged above.


11. Open audit items

  1. NLM Photonics specific Tier-1 customer commitments — less public than LWLG’s Stage 1-4 framework.
  2. Marvell-Polariton-Fotonix integration milestones — first design tape-out date.
  3. Ayar Labs strategic backer specific architecture commitment — which AI-fabric program is the volume customer.
  4. NVIDIA, Broadcom, Cisco specific Fotonix volume-roadmap milestones.
  5. Lightmatter customer-design pipeline — public detail is sparse.
  6. GF photonic test capacity scaling capex / timeline — binding constraint as volumes build.

12. Cross-references

Sources