POET Technologies and Sivers Semiconductors work as a pair trade, not an either/or. POET is the architecture bet: its optical engines and interposer platform plug into Celestial AI, which Marvell just spent $2B on and which feeds Nvidia; if that stack becomes standard, it is a 10x story. Sivers is the bottleneck beneath it, with InP laser arrays inside Ayar Labs' SuperNova. Every CPO roadmap, pluggable and Celestial engine needs a light source, and that source is Sivers. x.com 

POET Technologies is the upstream pick-and-shovel for Celestial AI's Photonic Fabric, which Marvell just acquired for about $5B. POET supplies the light source that powers it, and Marvell has the hyperscaler relationships with AWS, Google and Meta to deploy it at scale. Revenue is still early, but once Marvell integrates Celestial and ramps production orders, the market should recognize POET's value; Q3 or Q4 2026 could be the inflection, and patience is required. x.com 

China's already-shipping CW lasers make Sivers' Chinese pluggable pipeline depend on price and qualification speed. Yuanjie says its 70/100mW chips, the classes Sivers lists, are in continuous large-volume delivery. It guides Q3 revenue up 265-293% and expects new capacity from end-2026 to early 2027, before Sivers' Q4 2027 Glasgow expansion. Yuanjie's 120/300/400mW parts remain in customer validation; Western hyperscaler qualification is probably where Sivers still has time. x.com  x.com  x.com  … Show more

Optics and photonics should be among AI's biggest winners as faster GPU networks need optical connections that suppliers cannot expand quickly enough. Lumentum says some components are sold out through early 2029; Sivers is investing $30M to target more than 100M lasers a year, with operations expected in Q4 2027. Potential US restrictions on Chinese transceivers could reshape supply in 2027. Applied Optoelectronics and Sivers are the favorites. x.com  x.com 

Aeva's move into NPO/CPO is highly bullish for Sivers, again at the center of photonics through its CW-DFB lasers. Sivers is regarded as Aeva's laser supplier, and Aeva signed a joint development deal with an optical engine maker for a major hyperscaler, deploying in 2H27 and ramping in 2028. Per @ParadisLabs, Sivers partner GlobalFoundries expects silicon photonics revenue to more than double in 2026, with NPO ramping in 2027 and CPO in 2028. Jane Street closed its Sivers short. x.com  x.com  … Show more

Sivers Semiconductors is expensive because its valuation rests on a $1.2 billion opportunity pipeline, an unreliable KPI. At 15% conversion next year, it trades near 22x 2027 sales versus Lumentum's 14x and Coherent's and AAOI's 5x, and about 13x 2029 sales versus 3-9x for incumbents with contracted LTAs. The InP laser shortage is Lumentum's and Coherent's asset; Sivers gains access only in H2 2027 if all goes to plan and needs the shortage into 2028. It is a swing trade, not a conviction hold. … Show more

POET Technologies is uninvestable because its management has lost the trust of the big players. In April its CFO leaked information that led to the cancellation of Celestial AI and Marvell contracts. Being early in interposers and packaging does not make it a co-packaged optics winner, as others will catch up. Its TFLN partner QCI trails Hyperlight and is stuck on a 6-inch line, with no 8-inch wafer demo yet to show it can scale. x.com  x.com 

Sivers is less exposed to tighter AI financing because its 700M SEK raise funds a buildout largely reliant on external foundries. It is qualifying with Ayar, Jabil, GlobalFoundries and O-Net for potential tier-one hyperscaler pluggable, CPO and NPO deployments. Higher rates may strain neoclouds needing billions for GPUs, but optical content and aggregate demand should still rise even if hyperscaler AI capex falls, especially as new players start qualification. x.com 

POET Technologies and Sivers, as co-packaged optics pure plays, lose because CPO adoption has moved right by a full generation. SemiAnalysis reports Nvidia's flagship Kyber NVL144 rack is delayed over 12 months to 2028 and its NVL72x2 backup was cancelled, while optical scale-up won't arrive until Feynman. Packaging and interconnect hit physical walls in the same cycle, and the entire thesis for these names rests on CPO timing. x.com 

Sivers Semiconductors' LiDAR design win is closer to generating revenue now that its customer has placed initial production orders. Shipments begin in Q4 2026; automotive and industrial programs target 2027 production. Sivers estimates $53M-$138M in lifetime revenue potential, not firm backlog; the initial order value is undisclosed and the program was announced in March. Repeat orders and shipment growth must convert potential into revenue; the market also wants profitability. x.com  x.com