Zhongji InnoLight revenue and profit more than double in 3Q26

Zhongji InnoLight's Q3 print—revenue up 172%, net profit up 174%—is not a cyclical beat; it is a structural repricing of optical interconnects from commodity components to compute-critical infrastructure. The 1.6T ramp is the real tell. It forces synchronous SerDes upgrades in switch ASICs (Tomahawk 6, Spectrum-X) and compresses the CPO migration timeline by roughly 18 months. Upstream, InP epi-wafer and TFLN modulator capacity is now the binding constraint, not module assembly. On risk: US BIS controls have not explicitly targeted 1.6T modules, but EDA tooling for EML chip design and InP substrate supply are experiencing gray-zone restrictions. Zhongji's margin expansion is tethered to hyperscaler order books; if NVIDIA or Broadcom define the CPO architecture, Chinese module makers risk being relegated to the assembly tier. Competitively, Coherent and Lumentum will accelerate 1.6T qualification in Q4, squeezing Fabrinet's ODM share. More critically, Marvell is bundling PAM4 DSPs with optical engines—shifting the profit pool from the module layer to the silicon layer. Zhongji must close the in-house DSP gap or face a margin cliff. Outlook: within 18 months, 3.2T CPO prototypes will enter NVIDIA's Rubin validation cycle, dissolving the boundary between module maker and chip maker. Without vertical integration in silicon photonics and DSP, Zhongji's 2027 earnings curve risks a sharp post-peak contraction.