NVIDIA's Jensen Huang on Building the Dynamo of the Intelligence Age

He described the recent shift as AI moving beyond producing novel responses to reasoning, planning, and using tools (like browsers and spreadsheets), which is what makes it economically valuable and pay-worthy. He framed this as the key transition from “cute” generation to systems that “do work.”

He contrasted decades of data-center computing that mainly stores and retrieves pre-recorded files with a new paradigm where outputs are produced freshly for each query and context. He said this implies far more “generators” will be needed because future media and information will be generated rather than retrieved.

He defined NVIDIA’s role as building large systems that generate “tokens” (numbers) that can be arranged into different forms of intelligence, likening this to a factory output. He connected this to the “AI factory” idea: expensive, high-volume equipment that turns energy into computation and token generation.

He argued that as agentic AI scales, many more entities than humans will use computing and networks: agents collaborating with other agents, plus robots, cars, buildings, and manufacturing systems. He characterized this as a new global layer—like energy grids and the internet—that “cocoons” the planet with continuously generated intelligence.

He offered an industrial investment framework as five stacked layers: energy; chips/computers/networking; infrastructure (land/power/shell/operations); model companies; and an applications layer aimed at improving the human condition. He emphasized that focusing only on well-known model companies misses the broader opportunity across structured domains beyond text.