LandMark orders surpass 2026 revenue; capacity to expand sixfold

The real signal isn't LandMark's 5x profit jump—it's that the AI optical interconnect bottleneck has migrated upstream to epitaxy. Epitaxy wafers are the physical substrate for EML/DFB chips, capping silicon-photonics output. As transceivers scale to 1.6T, required epitaxy area and material purity per device rise in lockstep, while MOCVD growth rates hit hard physical ceilings. Sixfold expansion is simultaneously constrained by tube count, InP substrate allocation, and yield ramp. Whoever locks upstream locks the entire chain's production cadence. Compliance risk: MOCVD tooling (Veeco, Aixtron) and high-purity InP source material remain geographically concentrated. Taiwan, China and Japan still dominate InP single-crystal substrates; any supply-chain friction triggers costly dual-track qualification regimes. Competitively, IQE, Sumitomo, and domestic epitaxy players face a brutal 12-month window—a six-month order backlog effectively walls them out of the same customer base. The rational response is differentiation toward CPO thin-film lithium niobate routes, or vertical integration into substrate supply. Over 18–24 months, silicon-photonics epitaxy will shift from 'optional vendor' to 'strategic resource.' Hyperscalers will qualify second sources; once sixfold expansion lands, pricing power migrates from per-wafer spot to multi-year capacity contracts. Epitaxy wafers are becoming quasi-infrastructure.