Huang Renxun: In the future, AI models can run continuously on laptops, workstations, and hardware such as DGX Spark to build an AI intelligent agent that is online 24 hours a day.

In response to the industry trend of AI reasoning power gradually sinking to terminal devices, NVIDIA CEO Huang Renxun proposed that smartphones have already formed a distributed computing power architecture, with some calculations running on the local terminal and the remaining computing power processed in the cloud. This architecture will also become the mainstream operating mode for AI Agents in the future. He stated that tasks that can be processed locally should be run on terminal devices first, as this can reduce costs, decrease response latency, and achieve a more customized user experience. In the future, AI models can run continuously on laptops, workstations, and hardware such as DGX Sparks, creating an AI intelligent agent that is online all the time. Huang Renxun pointed out that the era of artificial intelligence will enter a decoupled distributed computing architecture, where AI computing power loads will be deployed in the cloud, enterprise intranets, and various terminal devices, achieving seamless interconnection and collaboration. For end users, they do not need to know the actual location of the computing power, they only need to enjoy the optimal AI services and user experience.