NVIDIA's supply chain risks - CEO explains | Jensen Huang and Lex Fridman

Jensen Huang addresses the complexity of managing NVIDIA's supply chain during unprecedented growth, emphasizing that no company in history has grown at their scale while continuing to accelerate. He describes his hands-on approach of regularly communicating with hundreds of CEOs across the entire IT and infrastructure industries, both upstream suppliers and downstream customers. Huang explains how he proactively educates these partners about NVIDIA's business conditions, growth drivers, and future direction to help them make informed investment decisions. He provides specific examples of successfully convincing memory manufacturers to invest in HBM memory and LPDDR5 for data centers, despite initial skepticism about using these technologies outside their traditional applications. The conversation reveals how NVIDIA's architectural shift from DGX-1 to MVLink-72 required fundamental changes in manufacturing approach, moving supercomputer integration from data centers into the supply chain itself. Each rack now contains 1.3-1.5 million components from 200 suppliers, requiring suppliers to significantly increase their power capacity for building and testing. Huang describes his process of flying to meet partners personally, explaining the technical and business rationale through first principles reasoning, and requesting billions in capital investments from each supplier. When asked about specific bottlenecks like ASML's EUV lithography or TSMC's advanced packaging, Huang expresses confidence rather than worry, stating that clear communication of needs and mutual understanding with suppliers allows him to trust their execution.