Jensen Huang, CEO of NVIDIA, discusses the necessity of extreme co-design in computing, driven by the need to distribute workloads for greater efficiency than simply adding more computers. He emphasizes NVIDIA's approach to company organization as a mechanism to produce output, reflecting the environment it exists within. Huang recounts the strategic decision to put CUDA on GeForce, despite financial strain, to build an install base, highlighting the importance of install base in defining an architecture. He also shares his methodology for envisioning and manifesting the future, shaping the belief systems of his team and the broader industry. Huang touches on the evolution of scaling laws in AI, from pre-training to agentic scaling, and addresses concerns about power and supply chain bottlenecks, as well as the potential of AI to transform various industries and human roles.
Sign in to continue reading, translating and more.
Continue
