Jefferies: Optical Transceiver Market to Triple by 2027, ~30% Supply Gap for 1.6T
Taylor Wilson
A Jefferies expert call projects the global optical transceiver market will triple by 2027, yet 1.6T modules face a persistent ~30% supply shortfall in both 2026 and 2027 — bottlenecked by upstream chips.
How big can this market get?
The global optical transceiver market is projected to double in 2026 and triple by 2027.
800G shipments in 2026: roughly 40–42 million units against demand of over 45 million — a ~10% gap. By 2027 shipments rise to 80 million.
1.6T is tighter: 2026 shipments ~18 million vs. demand ~26 million; 2027 shipments ~55 million vs. demand over 75 million — a ~30% shortfall both years.
This means → even at full production ramp, 1.6T supply won't catch demand until at least late 2027.
Why can't supply keep up?
The bottleneck sits in three upstream chip categories: DSP — the digital signal processor that converts optical signals into data a server can read; EML — the electro-absorption modulated laser chip that turns electrical signals into light; and CWL — the continuous-wave laser chip that provides a stable light source.
High-end DSPs (especially the 3 nm chips for 1.6T) are dominated by Broadcom and Marvell. 200G EML is controlled by Lumentum, Broadcom, and Sumitomo Electric; China has no mature supplier yet.
In plain terms = Chinese vendors can build the transceiver module itself, but the hardest chips inside it are still in overseas hands.
Where does China's localization effort stand?
CWL chips show the most progress. Yuanjie Technology leads, with four other vendors active.
Dongshan Precision is expected to begin mass-producing 200G EML in H2 2026 — the first domestic volume milestone for EML.
Passive optical components (isolators, filters, lenses, etc.) are already ~85% Chinese by global share — a solved problem.
Electronic chips (TIA, Driver) and silicon photonics — using silicon to carry light signals — are advancing but still need more time.
800G, 1.6T, 3.2T — which technology wins each generation?
800G: EML remains the dominant approach.
1.6T: Silicon photonics is expected to take over 60% share — roughly 15% lower power and cheaper, needing only 2–4 CWL chips versus 8 for EML.
3.2T: EML is likely to reclaim dominance because silicon photonics cannot reach the required frequency. 3.2T samples are expected in Q4 2026; small-scale commercial shipments in Q4 2027.
This means → the technology roadmap is not a straight line — the lead approach may change with every speed generation.
Regardless of route, what is the one indispensable material?
Both EML and CWL chips require indium phosphide (InP) substrates. Silicon-photonics and co-packaged optics (CPO — mounting the transceiver directly inside the chip package) also rely on CWL chips.
Thin-film lithium niobate (TFLN) may emerge as a modulation material at the 3.2T stage, but it still needs an external InP-based CWL chip to generate light.
In plain terms = no matter which technology converts electricity into light, InP substrates remain a non-negotiable foundation.
What is the next thing to watch?
Whether the 1.6T gap narrows on schedule hinges on domestic DSP and EML chip progress.
This is the single most important proof point for the 2026–2027 transceiver supply chain: can Dongshan Precision ship 200G EML on time, and can domestic 3 nm DSP keep pace?
This reflects a competitive pivot — the battle in optical transceivers has shifted from "who can build the module" to "who can make the chips inside it."
Content is for reference only, not financial advice.