AI Training Clusters Demand 800G and 1.6T Interconnects
AI training cluster operators require interconnect speeds between individual servers and switches far exceeding what traditional enterprise data center networking ever demanded, pulling demand toward 800G modules and pushing early roadmap discussions toward 1.6T generation products. Broadcom and Coherent have both expanded 800G module shipments considerably as hyperscale operators build out AI training infrastructure at unprecedented pace. This shift is pulling manufacturing investment toward high-speed production lines that cost considerably more to build but achieve volume economics that lower speed categories cannot match given AI infrastructure spending levels reflecting sustained data center and telecommunications infrastructure investment.
Market Impact: Adds 8 million high-speed ports annually.








