Thesis. The steepest growth in optics — revenue up 109% with gross margin rebuilt from 18.5% to 50.4% — anchored by NVIDIA committing $2B and taking equity to lock up laser supply. The catch is that the biggest drivers, optical circuit switching and optical scale-up, are still promises with dates on them, and the multiple has the furthest to fall if a date slips.
Reviewed 19 September 2026, after Q4 FY2026 results. Next review after Q1 FY2027 earnings on 10 November 2026.
What it does
Lumentum makes the light. Specifically indium phosphide lasers — EML and CW — which are the light sources inside optical transceivers, the parts that turn electrical signals into light so data can travel down fibre between racks.
It sells those laser chips to the companies that assemble finished transceivers, and it also sells finished transceivers itself (800G moving to 1.6T). It makes telecom gear, and it makes optical circuit switches — devices that route light directly without converting it back to electricity, which is how very large AI clusters get wired together.
Cloud and AI is now more than 60% of revenue. CEO is Michael Hurlston, in post since February 2025.
Silicon photonics cannot emit light. However clever the chip, something still has to generate the laser beam — and that is what Lumentum sells. TSMC has named lasers as one of four remaining bottlenecks to co-packaged optics.
Bull case
- Q4 FY2026, reported 11 August: revenue $1,006.3M, up 109.3%. Full-year revenue $3.01B
- The margin recovery is the real story. Non-GAAP gross margin 50.4% and non-GAAP operating margin 36.6%, against a gross margin of 18.5% as recently as FY2024. Non-GAAP EPS $3.23
- It hit its long-term target model a full quarter early, on management's own framing
- Guidance accelerates again. Q1 FY2027 revenue $1.225–1.275B, non-GAAP operating margin 39.5–40.5%, non-GAAP EPS $4.05–4.35
- NVIDIA put money in. A $2B commitment in March 2026 with billions committed to laser purchases, plus an equity investment. A customer buying equity to secure supply is about as strong a signal as exists
- The scarcity is confirmed from three directions. The customer (NVIDIA equity), the foundry (TSMC naming lasers a bottleneck), and a competitor (Broadcom investing to build EMLs itself because it can't get enough)
- Contracted work is visible. Optical circuit switch backlog above $400M, plus an incremental multi-hundred-million co-packaged optics order deliverable in H1 CY2027
- FY28 target of $8B revenue
Bear case
- Priced for the supercycle. If a timeline slips, the de-rate is violent. This is the highest-beta way to own optics
- The big drivers haven't hit the numbers yet. Management's own list — OCS volume from Q4 CY26, optical scale-up in 2027, scale-out broadening, 1.6T mix — is all still ahead
- Concentration. Smaller and more concentrated than Coherent, and effectively tied to NVIDIA's roadmap
- Broadcom is building into the exact layer the thesis rests on. It is already a top-three EML supplier and said in September it is investing in EMLs and CW lasers for its own demand
- Optics has a boom-bust history, and Chinese module pricing has broken margins in this industry before
- Co-packaged optics reshapes the category — it could expand the laser market or route around parts of the product line
- A new-ish CEO owns very aggressive public targets, which is a governance risk as much as an operational one
- The balance sheet carries complexity. Q4 showed a $7.16B GAAP net loss from a $7.8B one-time charge on converting notes to equity. Non-cash, but investors have to look past it
Major customers
Individual customer percentages are not disclosed.
- NVIDIA — multi-year partnership with billions committed to laser purchases, plus an equity investment taken in March 2026 as part of roughly $4B split with Coherent
- Transceiver module makers — Lumentum sells EML laser chips to the firms that assemble finished transceivers, including Chinese module manufacturers
- Hyperscalers — directly, for finished transceivers
- Cisco and systems vendors — components and modules
- Industrial and consumer — the legacy laser business in 3D sensing and materials processing, now a smaller share
Read-through. Lumentum sits upstream of much of the photonics list. FN manufactures for Lumentum — historically around 15% of Fabrinet's revenue — so Lumentum's shipment volumes flow directly into Fabrinet's numbers. Lumentum supplies the laser chips that go into transceivers built by AAOI and others, so its supply gap is their constraint.
NVDA platform roadmaps drive the demand curve. It competes with AVGO, COHR and Mitsubishi in EML supply — and Broadcom investing to build its own is a signal about scarcity before it is a competitive threat.
TSM expects silicon photonics to pass 50% of the transceiver market by 2027, with co-packaged optics in mass production from the second half of 2026. Silicon photonics still needs an external laser.
The metric that decides it: the EML supply gap, above 30% and widening. If it narrows, scarcity pricing is ending — that is the signal that breaks the thesis.
What would change the view
- The EML supply gap — widening is the thesis, narrowing breaks it
- Optical circuit switch ramp evidence from Q4 CY2026
- Gross margin durability above 40% now that it has reached 50.4%
- Depth of the NVIDIA relationship beyond the initial commitment
- Module pricing against Innolight and the Chinese assemblers
- Whether Broadcom's in-house EML investment starts to close the gap
Update log
- 19 Sep 2026 — Q4 FY2026 figures added: revenue $1,006.3M (+109.3%), non-GAAP gross margin 50.4%, non-GAAP EPS $3.23, Q1 FY27 guidance $1.225–1.275B. These had been flagged as missing at the previous review. Thesis updated to carry the 50.4% margin rather than 44.2%.
- 7 Sep 2026 — Reviewed. Noted the $7.8B non-cash charge on note conversion as noise rather than signal, and Broadcom's EML investment plus TSMC naming lasers a bottleneck as confirmation of scarcity. Actual Q4 figures not captured.
- 25 Jul 2026 — Full review.
Research and education only — not investment advice.
Made with Bullet