Nvidia's Huang Declares AI Infrastructure Cycle Will Span Decades at June 24 Shareholder Meeting

According to Nvidia's annual shareholder meeting on June 24, CEO Jensen Huang highlighted that the current AI infrastructure buildout cycle will span decades and involve major investments in power grids and networks. The company approved all 10 nominated board members.

Huang stated that AI infrastructure has transitioned from experimental to production stages, comparing the industry to a five-layer structure encompassing energy, chips, systems, infrastructure, models, and applications. He emphasized that Nvidia's Blackwell architecture delivers the lowest-cost token generation and highest token throughput in AI inference workloads, establishing the company's leadership as inference demands increasingly outpace training demands. Revenue potential, he noted, will depend on AI infrastructure performance per watt.