NVIDIA CEO Jensen Huang stated in Stanford University's course 'CS153: Cutting-Edge Systems' that the public's understanding of AI is overly narrow, viewing it merely as a new application within computers, such as image recognition, text generation, or programming assistance. He emphasized that AI is driving a fundamental shift in computing paradigms, moving from a 'pre-recorded' model to a 'generative' model—one that generates content in real time, perceives context, and responds to intent, rather than merely executing preset instructions. This transformation impacts every layer of the full stack, including development methodologies, team organization, computer architecture, network storage, cloud services, and application types. Huang argued that AI represents not just a tool upgrade but a platform revolution, with its potential depending on the synergy of underlying infrastructure, such as the integration of energy, chips, cloud services, models, and application layers. Using NVIDIA's chip architecture evolution as an example, he explained how each generation is designed to anticipate future computing bottlenecks, such as the Hopper architecture supporting pre-training, the Grace Blackwell architecture enabling inference and token generation, and the Vera Rubin architecture facilitating agent-based computing. Additionally, he stressed that the core function of AI factories is to produce tokens, with each token representing a monetizable unit of intelligence—the raw material for code, answers, designs, actions, and services.