Jensen Huang introduces the concept of "Agentic AI," a shift from generative AI (content creation) to productive AI that can perform tasks and use tools. This evolution requires a new computing stack, including powerful CPUs like NVIDIA's Vera to "harness" the GPU brains, driving demand for a broader range of hardware.
The conversation highlights a significant trend of enterprises like Eli Lilly and Samsung deploying AI in production on-premise, rather than solely in the cloud. This is driven by the need to process secure, proprietary data locally, at the "point of context."
Both CEOs assert that the demand for AI infrastructure is growing far faster than the supply chain can scale. They identify memory and advanced node semiconductors as key bottlenecks and predict this imbalance will last for a decade or more, even with massive investments in new capacity.
The discussion touches on the complex geopolitical landscape, particularly regarding China and Taiwan. While NVIDIA's H200 is licensed for China, future access is uncertain, and Taiwan remains the epicenter of semiconductor manufacturing, creating a significant concentration risk.
The conversation concludes by looking beyond the data center to a future of "personal AI" on devices like PCs and "physical AI" in robotics, autonomous vehicles, and hospitals. This requires distributed, localized intelligence that can't rely on a remote cloud connection.
Keep pulling the thread on Jensen Huang.