Serenity, the small-cap stock-picker whose reconstructed positions are tracked on this site, spent Monday pointing to a fresh Nomura research report that he reads as broad confirmation of his dominant investment theme: the upstream materials that feed AI photonics.

At the center of that theme is indium phosphide, or InP — the compound-semiconductor material used to make the high-speed lasers and optical components that carry data inside AI datacenters. As the industry moves toward co-packaged optics (CPO), which places the optical light source directly next to switch and GPU chips to save power and lift bandwidth, demand for InP substrates and epiwafers is expected to climb. Serenity's core argument has long been to "buy what's upstream of NVDA" — to hunt the small, hard-to-replace suppliers, or chokepoints, that the entire buildout quietly depends on.

The report and the price hikes. According to Serenity, the Nomura note names two of his flagship holdings, AXT Inc. (AXTI) and IQE plc (IQE), as the leading players in the space. He highlighted a set of pricing moves cited in the report: 2-inch InP substrate prices rising roughly 42–76%, 3-inch InP substrates up about 78%, 2-inch EML epiwafer prices up 50–75%, and 3-inch CW epiwafer prices climbing more than 40%. (EML lasers serve the fastest 800G/1.6T links, while CW — continuous-wave — lasers are the light source for silicon photonics and CPO.) He also noted that prices are falling inside China, consistent with his view that export controls are handing pricing power to non-Chinese suppliers.

Why it matters to his book. AXT is Serenity's highest-conviction name and, by his framing, the multi-layer InP chokepoint — controlling a large share of the supply chain from raw material through refining, crucibles and finished substrates. He treats it as his most legendary call, having tracked it from around $12 toward $140. IQE, meanwhile, is the largest independent merchant epiwafer foundry by reactor count and his high-risk turnaround bet: he argues its distressed valuation ignores latent InP reactor capacity that could re-rate the equity if the company restructures and pivots toward photonics. The sharp substrate and epiwafer price increases feed directly into both theses, since they suggest genuine supply shortages rather than temporary demand blips.

The Sandisk parallel. Serenity drew an explicit comparison to Sandisk (SNDK), the NAND flash maker he holds as a template for what he calls a "supercycle" — a structural, AI-driven shortage in which producers repeatedly raise prices above expectations. His personal expectation, he said, was that InP prices would keep climbing the way NAND has, as optical demand ramps. He was measured about it, framing the outcome as something the next roughly two years of the demand curve will ultimately decide.

The takeaway from his commentary today was less a new trade than a checkpoint: an outside research house naming his top holdings and documenting the exact upstream price pressure his thesis predicted. Serenity presented it as validation "on all fronts" — pricing, demand and supply — while acknowledging the thesis still has to play out over time.