Serenity, the small-cap stock-picker whose reconstructed views this site tracks, spent Monday celebrating a thesis he says is finally showing up in the trade press — and flagging where he thinks the next supply squeeze forms.

The indium phosphide squeeze he predicted. The stock-picker's day centered on a report describing the largest-ever price hike for indium phosphide (InP) substrates, the compound-semiconductor material used to make the high-speed lasers that light up optical links inside AI datacenters. According to the reporting he cited, InP substrate prices have now risen for four straight quarters, with the latest increase topping 10%, while prices for the epitaxial wafers built on those substrates are heading toward a third consecutive hike. One supplier line he found striking: even buyers with money on the table may simply be unable to get material.

That plays directly into his two flagship holdings. AXT (AXTI) is his highest-conviction name — a US maker of InP substrates and feedstock he frames as the multi-layer chokepoint (a critical upstream supplier with few substitutes) for AI photonics, where he argues the company holds roughly 40% of the InP chain in a duopoly with Japan's Sumitomo and gains pricing power as China export controls hobble rivals. He noted he called this bottleneck last year, and said he actually expects steeper hikes to come. IQE (IQE), the UK-listed merchant epiwafer foundry he holds as a turnaround bet, benefits one layer downstream, since those rising epiwafer prices flow through the reactors he says the company has sitting in latent capacity. He observed the Taiwan-centric coverage credited local names like VPEC and Landmark, but read the broader trend as validation of his own work.

A new bottleneck on his radar: MLCCs. The more forward-looking part of his day was a fresh thesis on multilayer ceramic capacitors, the tiny passive components stuffed by the thousands into every server, phone and car. He argued that surging AI-server demand for high-end MLCCs is crowding out capacity for consumer and general-purpose parts — citing channel checks that found common broad-use capacitors going out of stock and some makers halting orders. He drew an explicit parallel to the memory dynamic behind his "supercycle" thesis, where AI-driven demand for premium DRAM/HBM starved the legacy market and lifted prices; Micron (MU), his core US memory long, is the anchor of that view, and he expects a similar crowding-out to hit broad MLCCs in late 2026 or early 2027. He disclosed a position in Japan's Taiyo Yuden (6976), reasoning that a supplier with large overall MLCC share but relatively little AI-server exposure would benefit most as the shortage spills into consumer grades.

One skeptical aside. On the day's big headline — a reported $7 billion acquisition of AI-routing startup OpenRouter by payments firm Stripe — Serenity was measured, arguing the orchestration technology looks easy to replicate with little moat, even if OpenRouter brings a valuable userbase, dataset and momentum. He asked readers for their own verdict rather than staking a position.

*This is derived commentary for informational purposes only and is not investment advice.*