Serenity, the stock-picker whose portfolio is reconstructed from his public posts, spent Wednesday reinforcing his central bet — that the light sources feeding AI data-center optics are the next great supply bottleneck — with fresh evidence rather than new positions.

A channel check from the show floor. The day opened with what he described as a crowdsourced report from a community member at CIOE 2026 in Shenzhen, China's big optics trade show. The takeaways, as he relayed them: Sivers (SIVE), the tiny Swedish maker of indium-phosphide (InP) lasers that is his largest and highest-conviction holding, is actively engaged with Chinese pluggable-transceiver makers and has raised laser prices amid tight supply — with 70mW parts tied to 800G links reportedly the scarcest. InP is the compound-semiconductor material used for high-speed lasers; Serenity has long argued Sivers sits at the light-source chokepoint for co-packaged optics (CPO), the shift toward putting optical engines next to the switch or GPU die. Rising prices on constrained supply are exactly the scarcity dynamic his thesis predicts.

"Lasers as the next HBM." He seized on a Citi TMT note framing lasers and optical fiber as the "next HBM" — a nod to the memory-packaging bottleneck that drove pricing power in prior cycles. Serenity noted he has been talking up lasers and the architectural shift toward continuous-wave (CW) sources ahead of consensus since 2025. He laid out the ecosystem as a reference map: on lasers, Lumentum (LITE), Applied Optoelectronics (AAOI), Coherent (COHR), MACOM (MTSI) and Sivers; on fiber, Corning (GLW). LITE is his quality large-cap incumbent (roughly a 5% weight, tied to Google TPU optics); AAOI is his Made-in-America vertically integrated transceiver bet; COHR a lower-beta "they do everything" photonics compounder; MTSI a valuation benchmark he uses to argue Sivers is cheap; GLW the incumbent in glass substrates and optical fiber. His read: the 2027–2028 photonics supercycle will rhyme with what memory did in 2025–2026, especially as optical content per GPU/ASIC climbs.

Mapping the macro to capex. Later, Serenity worked through Anthropic's economic-scenario paper, which sketches sharply higher GDP growth after 2027 and posits that most incremental gains accrue to capital rather than labor. His interpretation: 2026–2027 are buildout years and 2028–2030 the payoff, a timeline he mapped onto the heavy AI capital spending — and near-term free-cash-flow pressure — at Amazon (AMZN), Meta (META) and Alphabet (GOOGL). He treats these hyperscalers as the demand anchor for his upstream bets: their spending on compute, power, networking and robotics is what ultimately pulls through his laser and photonics names. He remains bullish on Meta and Alphabet as owned positions and views Amazon favorably, though he holds no stated Amazon stake.

Nothing was bought or sold today; the notable move was conviction reinforcement, with real-world price hikes and a sell-side thesis both landing on the same conclusion he has held for months.