Enterprises like Eli Lilly and Samsung are increasingly deploying AI infrastructure on-premise rather than relying solely on the cloud. This move is driven by the need to maintain control over proprietary data, enhance security, and have intelligence located at the point of action.
The conversation highlights a paradigm shift from generative AI (creating content) to agentic AI (performing tasks). These AI agents will function as digital workers, requiring a 'harness' (run on CPUs) to manage their actions and a 'brain' (run on GPUs) for reasoning, driving massive, sustained demand for both types of processors.
Despite the AI hardware supply chain more than doubling annually, demand is growing even faster. Both CEOs identify memory and advanced node semiconductors as the primary bottlenecks, predicting that demand will outstrip supply for a decade or more.
The discussion looks beyond the current data center buildout to the next waves of AI deployment. "Physical AI" aims to bring IT to the $90 trillion industrial economy through robotics and automation, while "Personal AI" will run local models on PCs for 'unmetered intelligence' and context-aware assistance.
The conversation touches on the complex geopolitical landscape, including US-China relations and Taiwan's critical role. While the US is building domestic capacity, Taiwan is expected to remain a global tech epicenter, and access to the Chinese market remains a point of cautious optimism contingent on government policy.
Keep pulling the thread on Jensen Huang, Michael Dell.