Article published Mar 31, 2026. Prices below use latest available snapshots.
Note: This is a four-wave framework covering optical networking 2025-2030+. LITE is used as the representative ticker (Wave 1 EML laser monopoly, NVDA $2B non-cancelable deal), but the thesis spans many names by wave: COHR, LITE, AAOI, MRVL, FN (Wave 1); AVGO, MTSI/MACOM, CRDO, SMTC (Wave 2); TSEM, ALAB (Wave 3); NVDA, POET, SIVE, LWLG (Wave 4). Server engagement: 26 reactions, 11 replies (highest at time of capture).
The Four-Wave Framework
Frank maps optical networking as four overlapping technology waves. This is MORE structured than our agents' single-pass chain mapping:
| Wave | Phase | Timeline | Key Tech | Key Companies |
|---|---|---|---|---|
| Wave 1 | 800G Transceivers | Peak 2025-2027 | EML lasers, PAM4 DSP | COHR, LITE, AAOI, MRVL, FN |
| Wave 2 | LPO/Active Cables | 2026-2028 | Remove DSP, cut power 30-50% | AVGO, MTSI (MACOM), CRDO, SMTC |
| Wave 3 | 1.6T Transceivers | 2026-2029 | ~84M units at peak | TSEM (Tower Semi), ALAB (Astera Labs) |
| Wave 4 | CPO / 3.2T | 2028-2030+ | Co-packaged optics, 3.5x power efficiency | NVDA, POET, SIVE, LWLG |
This is the critical insight our agents missed: The market isn't static. What's a bottleneck TODAY (Wave 1: EML lasers) gives way to different bottlenecks TOMORROW (Wave 3: SiPh foundry, Wave 4: CPO). Frank maps the SEQUENCE of bottleneck shifts.
Cross-Reference with Our Agent Findings
Names ALL 3 of our agents found + Frank found:
- COHR ✅ — Frank has it as Wave 1 leader
- LITE ✅ — Frank confirms "200G EML laser monopoly"
- MRVL ✅ — Frank adds: Celestial AI acquisition = "most complete optical + compute + networking play"
- FN ✅ — Frank has it in Wave 1
- AVGO ✅ — Frank has it in Wave 2 (switch ASICs)
- AXTI ✅ — Frank has it in upstream (InP wafers)
- MACOM/MTSI ✅ — Frank has it in Wave 2 (analog ICs) — our agents' biggest new discovery is CONFIRMED
Names Frank found that our agents MISSED:
| Ticker | What | Wave | Why We Missed It |
|---|---|---|---|
| LWLG | Lightwave Logic — electro-optic polymer platform | Wave 4 (CPO) | Pre-revenue, prototype stage. Agents focused on current bottlenecks, not future tech. |
| TSEM | Tower Semiconductor — SiPh foundry | Wave 3 (1.6T) | @aleabitoreddit mentioned it! Our agents didn't trace to foundry level for silicon photonics. |
| ALAB | Astera Labs — signal retimers | Wave 3 | New IPO, agents may not have had data. |
| SMTC | Semtech — linear drivers | Wave 2 (LPO) | Analog IC layer — agents found MACOM but not Semtech. |
| SIVE | Sivers Semiconductors — DFB laser arrays | Wave 4 | Swedish, @aleabitoreddit had it as a 2x winner. |
| POET | POET Technologies — optical interposer | Wave 4 | Agent Alpha found this (1/3 confidence). Frank confirms. |
| IQE | Epitaxy wafers | Upstream | Agent Gamma found it. Frank has it but ranked "weakest fit" in chat debate. |
Critical new data from Frank:
NVIDIA invested $2B EACH in THREE companies: COHR, LITE, MRVL. Not just LITE — all three got $2B non-cancelable commitments. This is $6B total locked into the optical chain.
Marvell acquired Celestial AI (Dec 2025) for co-packaged optics. MRVL is positioning as the "system integrator" of the optical stack — consuming COHR components + LITE lasers.
Tower Semi (TSEM) is building 1.6T SiPh modules for NVIDIA protocols. This is the foundry bottleneck for the NEXT generation. Our agents found current-gen bottlenecks (EML lasers) but not the next-gen foundry constraint.
Market sizing: Optical transceiver market heading to $42.5B by 2032. Hyperscaler capex $600B+ in 2026, ~75% for AI infrastructure.
LWLG polymer modulators embedded in Tower's PH18 PDK — if this works, it changes the Wave 4 landscape. Pre-revenue risk but NVIDIA pipeline.
What This Means for Our Research
Upgrade our framework from "single chain" to "wave progression"
Our supply chain tracer maps the chain as it exists TODAY. Frank maps how the chain EVOLVES over 4 waves. We should add a "Wave 2/3/4" section to our trace output — what bottlenecks emerge NEXT when the current ones are resolved?
New tickers to add to watchlists
- ALAB (Astera Labs) — Wave 3 signal retimers. NEW.
- SMTC (Semtech) — Wave 2 linear drivers. NEW.
- TSEM (Tower Semi) — Wave 3 SiPh foundry. @aleabitoreddit had it. ADD.
- LWLG — Wave 4 polymer modulators. HIGH RISK pre-revenue. WATCH only.
- SIVE — Wave 4 DFB laser arrays. @aleabitoreddit had it as 2x winner. WATCH.
Our agents validated against Frank's months of research
The convergence is remarkable — our 3 agents running for 5 minutes each found 70-80% of what Frank found over months. The 20-30% we missed is mostly FUTURE tech waves (LWLG, POET, SIVE) and the NVIDIA $6B three-company deal detail. This suggests our agent protocol is good at tracing CURRENT chains but needs a "what comes next?" step.