Radiant targets next-gen CPO above 6.4T, eyes 2028 mass production

Radiant's CPO entry is not a pivot—it is a structural reallocation of the optical interconnect value chain. Backlight module competencies (micro-lens arrays, precision optical alignment, high-volume SMT) map directly onto CPO optical engine assembly. At 6.4T, the bottleneck shifts from ASIC to the optical engine (laser + waveguide + fiber array), pressuring silicon photonics foundries to deliver integrated PIC+EIC wafers and forcing Broadcom and NVIDIA to qualify multiple optical engine sources. On compliance, US BIS restrictions on >200G/lane photonic ICs are bifurcating the supply chain; if Radiant depends on imported EML lasers, the 2027 sampling milestone carries material-availability risk. Competitively, NVIDIA is accelerating in-house CPO via TSMC's COUPE platform, while Innolight and Eoptolink will be compelled to vertically integrate into optical engines. Radiant's lateral entry is fundamentally a margin-structure play: backlight module gross margins (15-25%) dwarf commodity transceiver margins (8-12%), and CPO optical engines command premium pricing. 12-24 month outlook: by mid-2026, expect 2-3 Chinese CPO optical engine suppliers to clear hyperscaler qualification. CPO eliminates the DSP/retimer, cutting BOM 30-40% and power 50%+. The traditional pluggable-module business model is being structurally dismantled, not merely disrupted.