Serenity AI Portfolio Explodes: Not Buying AI Applications, But the Optical Interconnection Bottleneck

The Serenity holdings are quite interesting:
AXTI, AAOI, LITE, SIVE, AEHR, OSS, TSEM, INTC, SIMO, SNDK
On the surface, it seems very diversified, including optical modules, InP substrates, lasers, silicon photonics foundry, testing equipment, SSD controllers, storage, and edge AI.
But if broken down by the AI data center supply chain, the logic is actually very clear:
→ AI computing power continues to expand
→ Data exchange between nodes skyrockets
→ Optical interconnection, CPO, silicon photonics, laser sources, InP substrates become new bottlenecks
→ Mass production still requires foundry, testing, burn-in, and high-speed storage support.
So this is not an ordinary AI concept stock group, but a portfolio built around the questions of “how to connect, how to test, how to store” in AI data centers.
Among them:
◾️ AAOI / LITE: Optical modules, lasers, data center optical interconnection;
◾️ AXTI: InP / GaAs / Ge substrates, upstream material bottlenecks;
◾️ SIVE: CPO, silicon photonics, laser array, highly flexible direction;
◾️ TSEM: Silicon photonics foundry platform;
◾️ AEHR: Wafer-level testing and burn-in;
◾️ SIMO / SNDK: AI data center SSD controllers and storage;
◾️ OSS: Edge AI / rugged computing;
◾️ INTC: Platform semiconductor assets.
◾️ Serenity is not truly buying AI applications, but the hardware bottlenecks after AI computing power expansion.
Optical modules are surface-level, InP and lasers are upstream, silicon photonics foundry and testing are for mass production validation, SSDs and controllers are supporting data flow expansion.
The most critical validation points are three:
◾️ Can CPO / silicon photonics enter mass production?
◾️ Can the demand for 800G / 1.6T optical interconnection continue to expand?
Are there orders, revenue, and gross margin validations in upstream materials and testing segments?
In one sentence:
This portfolio bets on the next phase of bottlenecks in AI data centers, shifting from GPU itself to optical interconnection, CPO, InP, lasers, testing, and high-speed storage.