The interconnection between chips lies at the heart of the global AI computing infrastructure, which is currently undergoing a monumental shift from electrical to optical signals. On March 16 (local time), NVIDIA introduced the Feynman chip, marking the first instance of optical communication technology being applied to chip interconnections. This innovation cuts down communication energy consumption in AI data centers by more than 70%. Analysts note that, with the clarification of overseas technological routes and the backing of domestic industrial policies, leading optical module companies listed on the A-share market, which are deeply integrated into the global computing supply chain, as well as manufacturers with CPO technology reserves, are poised to outperform in the new wave of 'computing-electricity synergy' infrastructure development.
