Nvidia CEO Jensen Huang Predicts AI’s Next Leap: From Digital Employees to Self‑Learning Agent Teams

In a sprawling 30,000‑word interview, Nvidia founder Jensen Huang painted a vivid picture of what comes after today’s AI hype. He says we’re moving from the "inference" stage—where models simply answer questions—to an "agentic" era where AI systems act like tiny digital employees. These agents can browse databases, fetch tools, and even spawn sub‑agents, creating whole AI‑powered teams that scale far more easily than human workforces. Huang calls this the "Agentic Scaling Law," a multiplier effect that turns raw computing power into ever‑more capable assistants. To keep up, hardware must evolve too. He highlighted Nvidia’s NVLink 72, which can host trillion‑parameter models on a single GPU, and new rack designs like the Grace Blackwell (built for large language models) and the Vera Rubin (optimized for AI agents that constantly call external tools). The secret, according to Huang, is a blend of specialization and flexibility. Nvidia’s CUDA platform, now at version 13.2, lets engineers rapidly adapt to breakthroughs such as Mixture‑of‑Experts models. By co‑designing software and silicon and listening to industry signals, Nvidia aims to stay ahead of the curve. Ultimately, Huang believes AI will reinvent the computer itself: agents that can research, read manuals, and use everyday tools—just like a person learning on the job—will become the backbone of future computing.

Read more