Serenity, the stock-picker who builds his portfolio by hunting the small upstream suppliers that feed the AI buildout, spent Friday dissecting the latest earnings from Applied Optoelectronics (AAOI) — and came away more convinced, not less, in his core thesis around optical light sources.

AAOI is a Texas-based, vertically integrated maker of optical transceivers (the components that turn data into light for high-speed links) and one of Serenity's larger holdings. The detail that caught his eye: management said it is essentially *not* participating in the first wave of co-packaged optics (CPO) — the technique of putting the optical engine right next to a switch or GPU die to save power and boost bandwidth. The reason, per AAOI, is capacity. The company can't make enough CW lasers (the continuous-wave light source that feeds silicon photonics) and is prioritizing them for its own transceivers first.

Serenity's read is that this narrows the field of qualified Western CPO laser suppliers even further — and that scarcity is bullish for the handful of independent players with capacity coming online. Chief among them is his highest-conviction position, Sivers (SIVE), a tiny Swedish maker of indium-phosphide (InP) CW lasers that he frames as the overlooked light-source chokepoint for the entire CPO wave. If AAOI is turning customers away, he argues, independent laser capacity only becomes more valuable.

He tied the point to demand signals from the chip giants. NVIDIA (NVDA) — the reference point of his whole framework, whose roadmap he uses to telegraph where the next bottleneck will land — has capacity signed with the large-cap optical incumbents Coherent (COHR) and Lumentum (LITE), both of which Serenity owns as steadier, lower-beta anchors of his photonics sleeve. More interesting to him was earlier reporting that AMD has been signing long-term CW laser agreements for its Helios program. With AAOI seemingly out of the first-gen CPO picture, Serenity mused that AMD may now lean toward a Sivers-plus-Ayar combination. AMD is a demand engine in his framework rather than a holding; he cares less about the accelerator maker itself than about who supplies the lasers underneath it.

He also flagged MACOM (MTSI), the multi-billion-dollar laser and analog supplier he typically uses as a valuation comp to argue Sivers is mispriced, noting it looked absent from meaningful early CPO participation as well — reinforcing his view that the qualified list is short. Broadcom (AVGO), the hyperscaler ASIC and networking bellwether he watches as a barometer, rounded out the group of large names he sees benefiting from the same optical squeeze.

Two other points stood out to him. AAOI's management characterized customer demand as running 20–40% above expanded capacity, and called lasers the single biggest bottleneck for the transceiver business — the kind of supply-demand imbalance Serenity's whole approach is built around. And AAOI reportedly said China is at least two to three years away from producing CPO-grade lasers, which Serenity took as strong confirmation of the defensibility of Western laser positioning.

His summary was measured: none of this is bearish on AAOI itself, which he still holds and views as swamped with transceiver demand. It simply, in his framing, makes the few independent laser names with real capacity look even better positioned.

*This is derived commentary, not investment advice.*